Noise of Polyphase Electric Motors

DK3193_half-series-title 11/2/05 8:35 AM Page A Noise of Polyphase Electric Motors Copyright © 2006 Taylor & Franci...

Author: Jacek F. Gieras | Chong Wang | Joseph Cho Lai

133 downloads 1262 Views 22MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

DK3193_half-series-title

11/2/05

8:35 AM

Page A

Noise of Polyphase Electric Motors

Copyright © 2006 Taylor & Francis Group, LLC


11/2/05

8:35 AM

Page B

ELECTRICAL AND COMPUTER ENGINEERING A Series of Reference Books and Textbooks

FOUNDING EDITOR Marlin O. Thurston Department of Electrical Engineering The Ohio State University Columbus, Ohio

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

14. 15. 16. 17. 18. 19. 20.

Rational Fault Analysis, edited by Richard Saeks and S. R. Liberty Nonparametric Methods in Communications, edited by P. Papantoni-Kazakos and Dimitri Kazakos Interactive Pattern Recognition, Yi-tzuu Chien Solid-State Electronics, Lawrence E. Murr Electronic, Magnetic, and Thermal Properties of Solid Materials, Klaus Schröder Magnetic-Bubble Memory Technology, Hsu Chang Transformer and Inductor Design Handbook, Colonel Wm. T. McLyman Electromagnetics: Classical and Modern Theory and Applications, Samuel Seely and Alexander D. Poularikas One-Dimensional Digital Signal Processing, Chi-Tsong Chen Interconnected Dynamical Systems, Raymond A. DeCarlo and Richard Saeks Modern Digital Control Systems, Raymond G. Jacquot Hybrid Circuit Design and Manufacture, Roydn D. Jones Magnetic Core Selection for Transformers and Inductors: A User’s Guide to Practice and Specification, Colonel Wm. T. McLyman Static and Rotating Electromagnetic Devices, Richard H. Engelmann Energy-Efficient Electric Motors: Selection and Application, John C. Andreas Electromagnetic Compossibility, Heinz M. Schlicke Electronics: Models, Analysis, and Systems, James G. Gottling Digital Filter Design Handbook, Fred J. Taylor Multivariable Control: An Introduction, P. K. Sinha Flexible Circuits: Design and Applications, Steve Gurley, with contributions by Carl A. Edstrom, Jr., Ray D. Greenway, and William P. Kelly



21. 22. 23. 24. 25. 26.

27.

28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44.

11/2/05

8:35 AM

Page C

Circuit Interruption: Theory and Techniques, Thomas E. Browne, Jr. Switch Mode Power Conversion: Basic Theory and Design, K. Kit Sum Pattern Recognition: Applications to Large Data-Set Problems, Sing-Tze Bow Custom-Specific Integrated Circuits: Design and Fabrication, Stanley L. Hurst Digital Circuits: Logic and Design, Ronald C. Emery Large-Scale Control Systems: Theories and Techniques, Magdi S. Mahmoud, Mohamed F. Hassan, and Mohamed G. Darwish Microprocessor Software Project Management, Eli T. Fathi and Cedric V. W. Armstrong (Sponsored by Ontario Centre for Microelectronics) Low Frequency Electromagnetic Design, Michael P. Perry Multidimensional Systems: Techniques and Applications, edited by Spyros G. Tzafestas AC Motors for High-Performance Applications: Analysis and Control, Sakae Yamamura Ceramic Motors for Electronics: Processing, Properties, and Applications, edited by Relva C. Buchanan Microcomputer Bus Structures and Bus Interface Design, Arthur L. Dexter End User’s Guide to Innovative Flexible Circuit Packaging, Jay J. Miniet Reliability Engineering for Electronic Design, Norman B. Fuqua Design Fundamentals for Low-Voltage Distribution and Control, Frank W. Kussy and Jack L. Warren Encapsulation of Electronic Devices and Components, Edward R. Salmon Protective Relaying: Principles and Applications, J. Lewis Blackburn Testing Active and Passive Electronic Components, Richard F. Powell Adaptive Control Systems: Techniques and Applications, V. V. Chalam Computer-Aided Analysis of Power Electronic Systems, Venkatachari Rajagopalan Integrated Circuit Quality and Reliability, Eugene R. Hnatek Systolic Signal Processing Systems, edited by Earl E. Swartzlander, Jr. Adaptive Digital Filters and Signal Analysis, Maurice G. Bellanger Electronic Ceramics: Properties, Configuration, and Applications, edited by Lionel M. Levinson



45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71.

11/2/05

8:35 AM

Page D

Computer Systems Engineering Management, Robert S. Alford Systems Modeling and Computer Simulation, edited by Naim A. Kheir Rigid-Flex Printed Wiring Design for Production Readiness, Walter S. Rigling Analog Methods for Computer-Aided Circuit Analysis and Diagnosis, edited by Takao Ozawa Transformer and Inductor Design Handbook: Second Edition, Revised and Expanded, Colonel Wm. T. McLyman Power System Grounding and Transients: An Introduction, A. P. Sakis Meliopoulos Signal Processing Handbook, edited by C. H. Chen Electronic Product Design for Automated Manufacturing, H. Richard Stillwell Dynamic Models and Discrete Event Simulation, William Delaney and Erminia Vaccari FET Technology and Application: An Introduction, Edwin S. Oxner Digital Speech Processing, Synthesis, and Recognition, Sadaoki Furui VLSI RISC Architecture and Organization, Stephen B. Furber Surface Mount and Related Technologies, Gerald Ginsberg Uninterruptible Power Supplies: Power Conditioners for Critical Equipment, David C. Griffith Polyphase Induction Motors: Analysis, Design, and Application, Paul L. Cochran Battery Technology Handbook, edited by H. A. Kiehne Network Modeling, Simulation, and Analysis, edited by Ricardo F. Garzia and Mario R. Garzia Linear Circuits, Systems, and Signal Processing: Advanced Theory and Applications, edited by Nobuo Nagai High-Voltage Engineering: Theory and Practice, edited by M. Khalifa Large-Scale Systems Control and Decision Making, edited by Hiroyuki Tamura and Tsuneo Yoshikawa Industrial Power Distribution and Illuminating Systems, Kao Chen Distributed Computer Control for Industrial Automation, Dobrivoje Popovic and Vijay P. Bhatkar Computer-Aided Analysis of Active Circuits, Adrian Ioinovici Designing with Analog Switches, Steve Moore Contamination Effects on Electronic Products, Carl J. Tautscher Computer-Operated Systems Control, Magdi S. Mahmoud Integrated Microwave Circuits, edited by Yoshihiro Konishi



72.

73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84.

85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96.

11/2/05

8:35 AM

Page E

Ceramic Materials for Electronics: Processing, Properties, and Applications, Second Edition, Revised and Expanded, edited by Relva C. Buchanan Electromagnetic Compatibility: Principles and Applications, David A. Weston Intelligent Robotic Systems, edited by Spyros G. Tzafestas Switching Phenomena in High-Voltage Circuit Breakers, edited by Kunio Nakanishi Advances in Speech Signal Processing, edited by Sadaoki Furui and M. Mohan Sondhi Pattern Recognition and Image Preprocessing, Sing-Tze Bow Energy-Efficient Electric Motors: Selection and Application, Second Edition, John C. Andreas Stochastic Large-Scale Engineering Systems, edited by Spyros G. Tzafestas and Keigo Watanabe Two-Dimensional Digital Filters, Wu-Sheng Lu and Andreas Antoniou Computer-Aided Analysis and Design of Switch-Mode Power Supplies, Yim-Shu Lee Placement and Routing of Electronic Modules, edited by Michael Pecht Applied Control: Current Trends and Modern Methodologies, edited by Spyros G. Tzafestas Algorithms for Computer-Aided Design of Multivariable Control Systems, Stanoje Bingulac and Hugh F. VanLandingham Symmetrical Components for Power Systems Engineering, J. Lewis Blackburn Advanced Digital Signal Processing: Theory and Applications, Glenn Zelniker and Fred J. Taylor Neural Networks and Simulation Methods, Jian-Kang Wu Power Distribution Engineering: Fundamentals and Applications, James J. Burke Modern Digital Control Systems: Second Edition, Raymond G. Jacquot Adaptive IIR Filtering in Signal Processing and Control, Phillip A. Regalia Integrated Circuit Quality and Reliability: Second Edition, Revised and Expanded, Eugene R. Hnatek Handbook of Electric Motors, edited by Richard H. Engelmann and William H. Middendorf Power-Switching Converters, Simon S. Ang Systems Modeling and Computer Simulation: Second Edition, Naim A. Kheir EMI Filter Design, Richard Lee Ozenbaugh Power Hybrid Circuit Design and Manufacture, Haim Taraseiskey



97. 98. 99. 100. 101. 102.

103. 104. 105. 106. 107. 108.

109. 110.

111. 112. 113.

114. 115. 116. 117. 118. 119.

11/2/05

8:35 AM

Page F

Robust Control System Design: Advanced State Space Techniques, Chia-Chi Tsui Spatial Electric Load Forecasting, H. Lee Willis Permanent Magnet Motor Technology: Design and Applications, Jacek F. Gieras and Mitchell Wing High Voltage Circuit Breakers: Design and Applications, Ruben D. Garzon Integrating Electrical Heating Elements in Appliance Design, Thor Hegbom Magnetic Core Selection for Transformers and Inductors: A User’s Guide to Practice and Specification, Second Edition, Colonel Wm. T. McLyman Statistical Methods in Control and Signal Processing, edited by Tohru Katayama and Sueo Sugimoto Radio Receiver Design, Robert C. Dixon Electrical Contacts: Principles and Applications, edited by Paul G. Slade Handbook of Electrical Engineering Calculations, edited by Arun G. Phadke Reliability Control for Electronic Systems, Donald J. LaCombe Embedded Systems Design with 8051 Microcontrollers: Hardware and Software, Zdravko Karakehayov, Knud Smed Christensen, and Ole Winther Pilot Protective Relaying, edited by Walter A. Elmore High-Voltage Engineering: Theory and Practice, Second Edition, Revised and Expanded, Mazen Abdel-Salam, Hussein Anis, Ahdab El-Morshedy, and Roshdy Radwan EMI Filter Design: Second Edition, Revised and Expanded, Richard Lee Ozenbaugh Electromagnetic Compatibility: Principles and Applications, Second Edition, Revised and Expanded, David Weston Permanent Magnet Motor Technology: Design and Applications, Second Edition, Revised and Expanded, Jacek F. Gieras and Mitchell Wing High Voltage Circuit Breakers: Design and Applications, Second Edition, Revised and Expanded, Ruben D. Garzon High Reliability Magnetic Devices: Design and Fabrication, Colonel Wm. T. McLyman Practical Reliability of Electronic Equipment and Products, Eugene R. Hnatek Electromagnetic Modeling by Finite Element Methods, João Pedro A. Bastos and Nelson Sadowski Battery Technology Handbook, Second Edition, edited by H. A. Kiehne Power Converter Circuits, William Shepherd and Li Zhang



11/2/05

8:35 AM

Page G

120. Handbook of Electric Motors: Second Edition, Revised and Expanded, edited by Hamid A. Toliyat and Gerald B. Kliman 121. Transformer and Inductor Design Handbook, Colonel Wm T. McLyman 122. Energy Efficient Electric Motors: Selection and Application, Third Edition, Revised and Expanded, Ali Emadi 123. Power-Switching Converters, Second Edition, Simon Ang and Alejandro Oliva 124. Process Imaging For Automatic Control, edited by David M. Scott and Hugh McCann 125. Handbook of Automotive Power Electronics and Motor Drives, Ali Emadi 126. Adaptive Antennas and Receivers, edited by Melvin M. Weiner 127. SPICE for Power Electronics and Electric Power, Second Edition, Muhammad H. Rashid and Hasan M. Rashid 128. Gaseous Electronics: Theory and Practice, Gorur G. Raju 129. Noise of Polyphase Electric Motors, Jacek F. Gieras, Chong Wang, and Joseph Cho Lai



11/2/05

8:35 AM

Page i


Jacek F. Gieras United Technologies Corporation Hamilton Sundstrand, Applied Research Rockford, Illinois, U.S.A.

Chong Wang General Motors Corporation Milford, Michigan, U.S.A.

Joseph Cho Lai The University of New South Wales at the Australian Defence Force Academy Canberra, Australian Capital Territory, Australia

Boca Raton London New York

A CRC title, part of the Taylor & Francis imprint, a member of the Taylor & Francis Group, the academic division of T&F Informa plc.


DK3193_Discl.fm Page 1 Friday, November 4, 2005 10:25 AM

Published in 2006 by CRC Press Taylor & Francis Group 6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 © 2006 by Taylor & Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group No claim to original U.S. Government works Printed in the United States of America on acid-free paper 10 9 8 7 6 5 4 3 2 1 International Standard Book Number-10: 0-8247-2381-3 (Hardcover) International Standard Book Number-13: 978-0-8247-2381-1 (Hardcover) Library of Congress Card Number 2005050213 This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials or for the consequences of their use. No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data Gieras, Jacek F. Noise of polyphase electric motors / Jacek F. Gieras, Chong Wang, Joseph Cho Lai. p. cm. -- (Electrical and Computer Engineering ; 129) Includes bibliographical references and index. ISBN 0-8247-2381-3 (alk. paper) 1. Electric motors, Polyphase--noise. 2. Electric Motors, Alternating current--Noise. 3. Electric motors, Alternating current--Design and construction. I. Lai, Joseph Cho. II. Wang, Chong. III. Title. IV. Series. TK2785.G535 2005 621.46--dc22

2005050213

Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com Taylor & Francis Group is the Academic Division of Informa plc.

and the CRC Press Web site at http://www.crcpress.com


DK3193

noise

November 4, 2005

14:37

Preface It has been estimated that more than 65% of the electrical energy produced in developed countries is consumed by electric motors. Electrical motors are the most popular machines of everyday life. They are used either as power machines providing propulsion torque or servo motors operating in a closed loop control with speed or position feedback. Electric motors are embedded in larger systems as their integral part. The noise radiated by electric motors immensely affects the overall noise of the system. Contemporary electric motors are designed with higher magnetic flux density in the air gap than motors manufactured a half century ago. Higher magnetic flux density in the air gap produces higher radial magnetic forces acting on the stator system and, consequently, higher vibration and acoustic noise. With the increased power density of electric motors and more demanding environmental requirements, the prediction of noise at the early stage of design of electrical motors has become a very important issue. Not only electromagnetic, thermal, and economic calculations, but also the level of noise and vibration must be considered, so that the overall performance can be optimized/balanced and specific requirements can be incorporated in the design to avoid large retrofit expenses. However, prediction of noise is more difficult and less accurate than, for example, torque–speed characteristics. This is because only a very small fraction of electrical energy is converted into acoustic energy and correct estimation of some mechanical and acoustic parameters is very difficult. The first book [119] on calculation of noise in electrical motors was published by Jordan in 1950. Details of harmonic field analysis including harmonic torques, noise, and vibration in induction motors are given in the book [87] by Heller and Hamata published in 1977. Analysis of noise in induction machines with emphasis on its reduction is given in monograph [248] by Yang, which was published in 1981. The most comprehensive analysis of noise and vibration in electrical machines is contained in the book [200] by Timar, Fazekas, Kiss, Miklas, and Yang published in 1989. It is also necessary to mention two books on noise and vibration in induction machines published by Russian researchers: Shubov in 1974 [187] and Astakhov, Malishev, and Ovcharenko in 1985 [10], and a book published by Polish researcher Kwasnicki in 1998 [127]. There is no book published so far on noise and vibration of permanent magnet (PM) synchronous

Copyright © 2006 Taylor & Francis Group, LLCxi

DK3193

noise

November 4, 2005

14:37

xii

Preface

motors. The demand on these motors is nowadays in second place, after the demand on induction motors. As most books on noise and vibration analysis in electrical machines were published over two decades ago, recent advances in vibro-acoustic theories and technologies are only accessible in learned journals and have not been captured in a single monograph. These advances include the development and application of numerical methods of noise computation such as the finite element method (FEM), boundary element method (BEM), and statistical energy analysis (SEA) [43, 230] to the prediction of noise in electrical machines. With the increase in the importance of noise analysis and synthesis in the modern approach to the design of electrical motors, the authors have made an attempt to prepare a modern monograph on noise calculation in induction and PM synchronous motors addressing electromagnetic, mechanical, and vibro-acoustic issues. The noise and vibration of switched reluctance motors have not been considered here. The authors have devised the book as both an electrical motor noise textbook and a handbook for electrical machine design engineers, research scientists, and graduate students. The book can also be helpful for multidisciplinary research teams working on noise prediction of systems with electrical motors, e.g., electrical vehicles, industrial electromechanical drives, HVAC (heating, ventilating, air conditioning) systems, marine propulsion systems, airborne apparatus, elevators, office equipment, health care equipment, etc. The authors hope that this book will fill the current gap in modern treatment of the analysis and reduction of noise in polyphase electric motors. Jacek F. Gieras Chong Wang Joseph C.S. Lai


DK3193

noise

November 4, 2005

14:37

Authors Jacek F. Gieras, Ph.D., graduated with distinction in 1971 (Master’s in Engineering) from the Technical University of Lodz, Poland. He received his Ph.D. in Power Electrical Engineering (Electrical Machines) in 1975 and dr hab. degree (corresponding to D.Sc. in the United Kingdom), also in Electrical Engineering (Electromagnetic Field Theory) in 1980 from the University of Technology, Poznan, Poland. From 1971 to 1998, Dr. Gieras pursued his academic career at several universities worldwide including Poland, Canada, Jordan, and South Africa. He was also a Central Japan Railway Company visiting professor at the University of Tokyo (Endowed Chair in Transportation Systems Engineering), Japan; guest professor at Chungbuk National University, Choengju, South Korea; and visiting professor at the University of Rome La Sapienza, Italy. In 1987, he was promoted to the rank of full professor (life title given by the President of the Republic of Poland). Since 1998 Gieras has been involved in industry-oriented research, cutting-edge technologies, and innovations at United Technologies Corporation, recently in the Department of Applied Research, Hamilton Sundstrand, Rockford, Illinois. Dr. Gieras has authored and coauthored nine books, more than 220 scientific and technical papers and ten patents. His most important books of international standing include: Linear Induction Motors, Oxford University Press, 1994, United Kingdom; Permanent Magnet Motors Technology: Design and Applications, Marcel Dekker, New York, 1996, Second Edition, 2002 (coauthor M. Wing); Linear Synchronous Motors: Transportation and Automation Systems, CRC Press, Boca Raton, Florida, 1999 (coauthor Z. Piech); and Axial Flux Permanent Magnet Machines, Springer-Kluwer, Boston, 2004 (coauthors R. Wang and M. Kamper). He is a Fellow of IEEE, full member of the International Academy of Electrical Sciences, a member of numerous steering committees of international conferences, and cited by all Marquis Who’s Whos.

Copyright © 2006 Taylor & Francis Group, LLCxiii

DK3193

noise

November 4, 2005

14:37

xiv

Authors

Chong Wang, Ph.D., received B.S. and M.S. degrees in Acoustics from Nanjing University, China, in 1964 and 1989, respectively. He received his Ph.D. in Mechanical Engineering from the University of New South Wales, Australia, in 1999. From 1989 to 1995, he was with the Institute of Acoustical Engineering, Northwestern Polytechnical University, China, as an assistant lecturer, lecturer, and an associate professor, where he lectured on several acoustics courses and conducted research work in noise and vibration engineering. In 1999, Dr. Wang joined the Institute for Research in Construction, National Research Council, Canada, as a Canadian Government Laboratory Visiting Fellow. Since 2000, he has been employed as a senior engineer with General Motors Corporation, United States. His research experience and interests include vibro-acoustics, room acoustics, FEA/BEA, SEA, and active noise and vibration control. He has published more than 60 papers in journals and conference proceedings in these areas. Joseph Cho Lai, Ph.D., is a professor of mechanical engineering and associate dean (research) at the Australian Defense Force Academy (ADFA) Campus of the University of New South Wales (UNSW). He was appointed in January 2004. Prior to this appointment, he was the head of the School of Aerospace and Mechanical Engineering from 2001–2003. Dr. Lai obtained a B.Sc. (United Kingdom) in Mechanical Engineering with first-class honors from the University of Hong Kong in 1975. He was awarded a Master of Engineering Science in 1978 and a Ph.D. in Mechanical Engineering in 1981 from the University of Queensland, Australia. He lectured in mechanical engineering at the University of Queensland until June 1985 when he joined the Department of Mechanical Engineering at UNSW/ADFA as a lecturer. Dr. Lai’s research interests are in turbulent shear flows and acoustics and vibration control. He has published 192 papers in journals and conference proceedings in these areas. His current work focuses on the aerodynamic propulsive mechanism of flapping wings as applied to microaerial vehicles, vibro-acoustic communication in termites, and noise and vibration control of machines, including electrical machines. He is a member of four International Standard Organization Acoustics Working Groups. He is a Fellow of the Institution of Engineers, Australia, a Fellow of the Australian Acoustical Society, and an Associate Fellow of the American Institute of Aeronautics and Astronautics.


DK3193

noise

November 4, 2005

14:37

Contents 1

2

Generation and Radiation of Noise in Electrical Machines 1.1 Vibration, sound, and noise . . . . . . . . . . . . . . . . 1.2 Sound waves . . . . . . . . . . . . . . . . . . . . . . . 1.3 Sources of noise in electrical machines . . . . . . . . . . 1.3.1 Electromagnetic sources of noise . . . . . . . . . 1.3.2 Mechanical sources of noise . . . . . . . . . . . 1.3.3 Aerodynamic noise . . . . . . . . . . . . . . . . 1.4 Energy conversion process . . . . . . . . . . . . . . . . 1.5 Noise limits and measurement procedures for electrical machines . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 Deterministic and statistical methods of noise prediction 1.7 Economical aspects . . . . . . . . . . . . . . . . . . . . 1.8 Accuracy of noise prediction . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . .

. . . .

. . . .

. . . .

. 9 . 13 . 17 . 18

Magnetic Fields and Radial Forces in Polyphase Motors Fed with Sinusoidal Currents 2.1 Construction of induction motors . . . . . . . . . . . . . . . . . 2.2 Construction of permanent magnet synchronous brushless motors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 A.C. stator windings . . . . . . . . . . . . . . . . . . . . . . . 2.4 Stator winding MMF . . . . . . . . . . . . . . . . . . . . . . . 2.4.1 Single-phase stator winding . . . . . . . . . . . . . . . 2.4.2 Three-phase stator winding . . . . . . . . . . . . . . . . 2.4.3 Polyphase stator winding . . . . . . . . . . . . . . . . . 2.5 Rotor magnetic field . . . . . . . . . . . . . . . . . . . . . . . 2.6 Calculation of air gap magnetic field . . . . . . . . . . . . . . . 2.6.1 Effect of slots . . . . . . . . . . . . . . . . . . . . . . . 2.6.2 Effect of eccentricity . . . . . . . . . . . . . . . . . . .

2.7

2.6.4 Effect of rotor saliency . . . . . . . Radial forces . . . . . . . . . . . . . . . . 2.7.1 Production of radial magnetic forces 2.7.2 Amplitude of magnetic pressure . .

Copyright © 2006 Taylor & Francis Group, LLCxv

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

1 1 1 5 5 7 7 7

. . . . . . .

. . . .

. . . .

. . . .

. . . .

21 . 21 . . . . . . . . . .

23 25 28 28 32 33 36 37 37 40

. . . .

44 45 45 48

DK3193

noise

November 4, 2005

14:37

xvi

Contents 2.7.3 2.7.4 2.7.5

2.8

Deformation of the stator core . . . . . . . . Frequencies and orders of magnetic pressure Radial forces in synchronous machines with slotted stator . . . . . . . . . . . . . . . . . 2.7.6 Frequencies of vibration and noise . . . . . . Other sources of electromagnetic vibration and noise 2.8.1 Unbalanced line voltage . . . . . . . . . . . 2.8.2 Magnetostriction . . . . . . . . . . . . . . . 2.8.3 Thermal stress analogy . . . . . . . . . . . . 2.8.4 FEM model . . . . . . . . . . . . . . . . . .

. . . . . . . 49 . . . . . . . 50 . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

3 Inverter-Fed Motors 3.1 Generation of higher time harmonics . . . . . . . . . . . . . 3.2 Analysis of radial forces for nonsinusoidal currents . . . . . 3.2.1 Stator and rotor magnetic flux density . . . . . . . . 3.2.2 Stator harmonics of the same number . . . . . . . . 3.2.3 Interaction of stator and rotor harmonics . . . . . . . 3.2.4 Rotor harmonics of the same number . . . . . . . . 3.2.5 Frequencies and orders of magnetic pressure for nonsinusoidal currents . . . . . . . . . . . . . . . . 3.2.6 Interaction of stator harmonics of different numbers . 3.2.7 Interaction of switching frequency and higher time harmonics . . . . . . . . . . . . . . . . . . . . . . . 3.2.8 Interaction of permeance and magnetomotive force (MMF) harmonics . . . . . . . . . . . . . . . . . . 3.2.9 Rectifier harmonics . . . . . . . . . . . . . . . . . . 3.3 Higher time harmonic torques in induction machines . . . . 3.3.1 Asynchronous torques . . . . . . . . . . . . . . . . 3.3.2 Pulsating torques . . . . . . . . . . . . . . . . . . . 3.4 Higher time harmonic torques in permanent magnet (PM) brushless machines . . . . . . . . . . . . . . . . . . . . . . 3.5 Influence of the switching frequency of an inverter . . . . . 3.6 Noise reduction of inverter-fed motors . . . . . . . . . . . . 4 Torque Pulsations 4.1 Analytical methods of instantaneous torque calculation . . . 4.2 Numerical methods of instantaneous torque calculation . . . 4.3 Electromagnetic torque components . . . . . . . . . . . . . 4.4 Sources of torque pulsations . . . . . . . . . . . . . . . . . 4.5 Higher harmonic torques of induction motors . . . . . . . . 4.6 Cogging torque in permanent magnet (PM) brushless motors 4.6.1 Air gap magnetic flux density . . . . . . . . . . . . 4.6.2 Calculation of cogging torque . . . . . . . . . . . . 4.6.3 Simplified cogging torque equation . . . . . . . . . 4.6.4 Influence of eccentricity . . . . . . . . . . . . . . . Copyright © 2006 Taylor & Francis Group, LLC

. . . . . . .

. . . . . . .

. . . . . . .

51 54 58 58 58 62 62

. . . . . .

. . . . . .

. . . . . .

65 65 66 67 68 69 70

. . . 70 . . . 70 . . . 71 . . . . .

. . . . .

. . . . .

71 71 71 71 72

. . . 73 . . . 73 . . . 75

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

77 77 78 79 80 80 81 82 84 87 88

DK3193

noise

November 4, 2005

14:37

Contents

4.7 4.8 4.9

5

xvii 4.6.5 Calculation and comparison with measurements . . . . Torque ripple due to distortion of EMF and current waveforms in permanent magnet (PM) brushless motors . . . . . . . . . . Tangential forces vs. radial forces . . . . . . . . . . . . . . . Minimization of torque ripple in PM brushless motors . . . . . 4.9.1 Slotless windings . . . . . . . . . . . . . . . . . . . . 4.9.2 Skewing stator slots . . . . . . . . . . . . . . . . . . 4.9.3 Shaping stator slots . . . . . . . . . . . . . . . . . . . 4.9.4 Selection of the number of stator slots . . . . . . . . . 4.9.5 Shaping PMs . . . . . . . . . . . . . . . . . . . . . . 4.9.6 Skewing PMs . . . . . . . . . . . . . . . . . . . . . . 4.9.7 Shifting PM segments . . . . . . . . . . . . . . . . . 4.9.8 Selection of PM width . . . . . . . . . . . . . . . . . 4.9.9 Magnetization of PMs . . . . . . . . . . . . . . . . . 4.9.10 Creating magnetic circuit asymmetry . . . . . . . . .

. . 92 . . . . . . . . . . . . .

Stator System Vibration Analysis 5.1 Forced vibration . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Simplified calculation of natural frequencies of the stator system . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Improved analytical method of calculation of natural frequencies . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Natural frequency of the stator core . . . . . . . . . . . 5.3.2 Natural frequency of a frame with end bells . . . . . . . 5.3.3 Natural frequency of a stator core–frame system . . . . 5.3.4 Effect of the stator winding and teeth . . . . . . . . . . 5.3.5 Analytical calculation of natural frequencies for a stator core-winding-frame system . . . . . . . . . . . . . . . 5.4 Numerical verification . . . . . . . . . . . . . . . . . . . . . . 5.4.1 FEM modeling . . . . . . . . . . . . . . . . . . . . . . 5.4.2 Comparison of analytical calculations with the FEM . .

6 Acoustic Calculations 6.1 Sound radiation efficiency . . . . . . . . . . . . . 6.2 Plane radiator . . . . . . . . . . . . . . . . . . . . 6.2.1 Infinite plates . . . . . . . . . . . . . . . . 6.2.2 Finite plates in bending motion . . . . . . 6.3 Infinitely long cylindrical radiator . . . . . . . . . 6.4 Finite length cylindrical radiator . . . . . . . . . . 6.4.1 Acoustically thin shells . . . . . . . . . . . 6.4.2 Acoustically thick shells . . . . . . . . . . 6.4.3 Modal radiation efficiencies of acoustically thick shells . . . . . . . . . . . . . . . . . 6.4.4 Modal averaged radiation efficiency . . . . 6.4.5 Validity of using an infinite length model . Copyright © 2006 Taylor & Francis Group, LLC

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . . . . . . .

94 99 102 102 103 103 104 104 104 104 104 105 105

107 . 107 . 110 . . . . .

112 112 114 115 116

. . . .

117 120 120 122

. . . . . . . .

127 127 129 130 133 141 145 147 149

. . . . . . . . 154 . . . . . . . . 157 . . . . . . . . 161

DK3193

noise

November 4, 2005

14:37

xviii

Contents 6.4.6

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

164 166 167 168 171

Noise and Vibration of Mechanical and Aerodynamic Origin 7.1 Mechanical noise due to shaft and rotor irregularities . . . 7.2 Bearing noise . . . . . . . . . . . . . . . . . . . . . . . . 7.2.1 Rolling bearings . . . . . . . . . . . . . . . . . . 7.2.2 Sleeve bearings . . . . . . . . . . . . . . . . . . . 7.3 Noise due to toothed gear trains . . . . . . . . . . . . . . 7.4 Aerodynamic noise . . . . . . . . . . . . . . . . . . . . . 7.5 Mechanical noise generated by the load . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

175 175 176 176 180 180 181 184

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

187 187 189 189 189 193 196 197 197 197 198 198 198

6.5

7

8

Effects of boundary conditions on the radiation efficiency . . . . . . . . . . . . . . . . . . . . Calculations of sound power level . . . . . . . . . . . 6.5.1 Sound power radiated from a stator . . . . . . 6.5.2 Total sound power of an induction motor . . . 6.5.3 Permanent magnet synchronous motors . . . .

. . . . .

Acoustic and Vibration Instrumentation 8.1 Measuring system and transducers . . . . . . . . . . . . . 8.2 Measurement of sound pressure . . . . . . . . . . . . . . 8.2.1 Choice of microphones . . . . . . . . . . . . . . . 8.2.2 The sound pressure sensor–condenser microphone 8.2.3 Sound level meter . . . . . . . . . . . . . . . . . . 8.2.4 Acoustic calibrator . . . . . . . . . . . . . . . . . 8.2.5 Level recorder . . . . . . . . . . . . . . . . . . . 8.3 Acoustic measurement procedure . . . . . . . . . . . . . . 8.3.1 Effect of the operator on measurement results . . . 8.3.2 Measurement position . . . . . . . . . . . . . . . 8.3.3 Standing waves . . . . . . . . . . . . . . . . . . . 8.3.4 Measurements of ambient sound pressure levels . . 8.3.5 Corrections for background sound during source measurements . . . . . . . . . . . . . . . . . . . . 8.3.6 Polar plots . . . . . . . . . . . . . . . . . . . . . 8.4 Vibration measurements . . . . . . . . . . . . . . . . . . 8.4.1 Theory of operation of vibration-measuring transducer . . . . . . . . . . . . . . . . . . . . . . 8.4.2 Characteristics of piezoelectric accelerometers . . 8.4.3 Other vibration transducers . . . . . . . . . . . . . 8.5 Frequency analyzers . . . . . . . . . . . . . . . . . . . . 8.6 Sound power and sound pressure . . . . . . . . . . . . . . 8.7 Indirect methods of sound power measurement . . . . . . 8.7.1 Determination of sound power in an anechoic/ semianechoic room . . . . . . . . . . . . . . . . . 8.7.2 Reverberation room . . . . . . . . . . . . . . . . . 8.7.3 Juxtaposition principle using a reference sound source . . . . . . . . . . . . . . . . . . . .


. . . . 199 . . . . 200 . . . . 200 . . . . . .

. . . . . .

. . . . . .

. . . . . .

201 207 211 213 214 215

. . . . 215 . . . . 216 . . . . 217

DK3193

noise

November 4, 2005

14:37

Contents 8.8

8.9

xix Direct method of sound power measurement — sound intensity technique . . . . . . . . . . . . . . . . . . . . 8.8.1 Historical perspective . . . . . . . . . . . . . . . 8.8.2 Theoretical background . . . . . . . . . . . . . 8.8.3 Sound intensity probe . . . . . . . . . . . . . . 8.8.4 External noise suppression . . . . . . . . . . . . 8.8.5 Error considerations . . . . . . . . . . . . . . . 8.8.6 Dynamic capability and pressure-intensity index Standard for testing acoustic performance of rotating electrical machines . . . . . . . . . . . . . . . . . . . . 8.9.1 Background . . . . . . . . . . . . . . . . . . . . 8.9.2 Acoustic tests on an induction motor . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .


. . . . . . . . . . . . . . . .

217 217 217 219 221 221 223

. . . . . 224 . . . . . 224 . . . . . 226

9 Numerical Analysis 9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2 FEM model for radial magnetic pressure . . . . . . . . . . . . . 9.2.1 Induction motor . . . . . . . . . . . . . . . . . . . . . . 9.2.2 Permanent magnet synchronous motor . . . . . . . . . . 9.3 FEM for structural modeling . . . . . . . . . . . . . . . . . . . 9.4 BEM for acoustic radiation . . . . . . . . . . . . . . . . . . . . 9.4.1 Governing equation and boundary conditions . . . . . . 9.4.2 FEM . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.3 BEM . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.4 Radiating sphere . . . . . . . . . . . . . . . . . . . . . 9.4.5 Application to the prediction of radiated acoustic power from an inverter-fed induction motor . . . . . . . . . . . 9.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Statistical Energy Analysis 10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 10.2 Power flow between linearly coupled oscillators . . . . 10.2.1 Two coupled oscillators . . . . . . . . . . . . 10.2.2 Three series coupled oscillators . . . . . . . . 10.2.3 Energy exchange between groups of oscillators 10.3 Coupled multimodal systems . . . . . . . . . . . . . . 10.3.1 General SEA equations . . . . . . . . . . . . . 10.3.2 SEA model establishment . . . . . . . . . . . 10.3.3 SEA parameters . . . . . . . . . . . . . . . . 10.3.4 Limitations of SEA . . . . . . . . . . . . . . . 10.4 Experimental SEA . . . . . . . . . . . . . . . . . . . 10.4.1 General theory . . . . . . . . . . . . . . . . . 10.4.2 Recent developments . . . . . . . . . . . . . . 10.5 Application to electrical motors . . . . . . . . . . . . . 10.5.1 Subsystems of a motor structure . . . . . . . . 10.5.2 Internal and coupling loss factors . . . . . . .

. . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . .

231 231 232 233 237 239 243 243 247 248 248

. 250 . 255

. . . . . . . . . . . . . . . .

257 257 259 259 261 264 267 267 269 272 278 281 282 285 290 291 292

DK3193

noise

November 4, 2005

14:37

xx

Contents 10.5.3 Input power to the stator . . . . . . . . . . . . . . . . . . 293 10.5.4 Sound power radiated from the motor structure . . . . . . 295

11 Noise Control 11.1 Mounting . . . . . . . . . . . . . . . . . . . . . 11.1.1 Foundation . . . . . . . . . . . . . . . . 11.1.2 Principles of vibration and shock isolation 11.1.3 Vibration limits . . . . . . . . . . . . . . 11.1.4 Shaft alignment . . . . . . . . . . . . . . 11.2 Standard methods of noise reduction . . . . . . . 11.3 Active noise and vibration control . . . . . . . . 11.3.1 Principles of active noise control . . . . . 11.3.2 Induction motor acoustic noise reduction 11.3.3 Active vibration isolation . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

301 301 301 303 306 306 307 311 311 313 315

Appendix A Basics of Acoustics A.1 Sound field variables and wave equations A.2 Sound radiation from a point source . . . A.3 Decibel levels and their calculations . . . A.4 Spectrum analysis . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

319 319 321 323 325

. . . .

. . . .

. . . .

. . . .

Appendix B Permeance of Nonuniform Air Gap 327 B.1 Permeance calculation . . . . . . . . . . . . . . . . . . . . . . . 327 B.2 Eccentricity effect . . . . . . . . . . . . . . . . . . . . . . . . . . 328 Appendix C Magnetic Saturation

333

Appendix D Basics of Vibration 337 D.1 A mass–spring–damper oscillator . . . . . . . . . . . . . . . . . . 337 D.2 Lumped parameter systems . . . . . . . . . . . . . . . . . . . . . 339 D.3 Continuous systems . . . . . . . . . . . . . . . . . . . . . . . . . 342 Symbols and Abbreviations

347

Bibliography

353


DK3193

noise

November 3, 2005

15:28

1 Generation and Radiation of Noise in Electrical Machines 1.1

Vibration, sound, and noise

Vibration is a limited reciprocating motion of a particle of an elastic body or medium in alternately opposite directions from its position of equilibrium, when that equilibrium has been disturbed. In order to vibrate, the body or system must have two characteristics: elasticity and mass. The amplitude of vibration is the maximum displacement of a vibrating particle or body from its position of rest. Sound is defined as vibrations transmitted through an elastic solid, liquid, or gas with frequencies in the approximate range of 20 to 20,000 Hz, capable of being detected by human ears. Pitch is the perceived tone of a sound which is determined by the sound wave frequency. A sound with a high frequency (short wavelength) has a high pitch, while a sound with low frequency (long wavelength) has a low pitch. Noise is disagreeable or unwanted sound. Distinction is made between airborne noise and noise traveling through solid objects. Airborne noise is the noise caused by the movement of large volumes of air and the use of high pressure. Structure-borne noise is the noise carried by means of vibrations of solid objects.

1.2

Sound waves

A sound wave is generated by a vibrating object and can be defined as a mechanical disturbance advancing with a finite speed through a medium. Sound waves are small-amplitude adiabatic oscillations characterized by wavespeed, wavelength, frequency, and amplitude (Appendix A). In air, sound waves are longitudinal waves, that is, with displacement in the direction of propagation. In other words, the motion of the individual particles of the medium is in the direction that is 1 Copyright © 2006 Taylor & Francis Group, LLC

DK3193

noise

November 3, 2005

15:28

2


parallel to the direction of the energy transport. Transverse waves are those with vibrations perpendicular to the direction of travel of the wave and exist in the elastic medium. Examples of transverse waves include waves on a string and electromagnetic waves. Only transverse waves can be polarized, i.e., can have orientation. Polarized waves oscillate in only one direction perpendicular to the line of travel. For example, the polarization of an electromagnetic wave is defined as the orientation of the electric field vector. The electric field vector is perpendicular to both the direction of travel and the magnetic field vector. Polarized waves can be formed from unpolarized waves by passing them through some polarizing process, e.g., a train of unpolarized waves in a rope can be polarized by passing them through a narrow physical gap. Sound waves cannot be polarized. Unpolarized waves can oscillate in any direction in the plane perpendicular to the direction of travel and have no preferred plane of polarization. All sound waves have common behavior under a number of standard situations and exhibit: • reflection, i.e., the phenomenon of a propagating wave being thrown back from a surface between two media with different mechanical properties; • refraction, i.e., the change in direction of a propagating wave when passing from one medium to another; • diffraction, i.e., the process of spreading out of waves, e.g., when they travel through a small slit or go around an obstacle; • scattering, i.e., the change in direction of motion; • interference, i.e., mutual influence of two waves, e.g., the addition of two waves that come in to contact with each other; • absorption, i.e., the incident sound that strikes a material that is not reflected back; • dispersion, i.e., the splitting up of a wave depending on frequency. Sound amplitude can be measured as sound pressure level (SPL), sound intensity level (SIL), sound power level (SWL), and sound energy density (SED) (Appendix A). A human ear can perceive sound waves of sufficient intensity whose frequencies are approximately within the limits from 16 to 20, 000 Hz (audio-frequency range). There is a minimum sound intensity for a given frequency at which the sound can be perceived by the human ear. The minimum sound intensity is different for different frequencies and is called the threshold of audibility. Figure 1.1 shows the audibility zone for the whole audio-frequency range. The range of the sound intensity that can be perceived by the ear is from 10−12 to 1 W/m2 corresponding to 20 µPa sound pressure. The maximum sound intensity at which the ear feels a pain is called the threshold of pain. Sound amplitudes that are extremely loud (at the threshold of pain) have pressure amplitudes of only 100 Pa. Some environmental noise levels are compared in Figure 1.2. Typical sound power levels for common sounds are also given in Table 1.1


DK3193

noise

November 3, 2005

15:28

Generation and Radiation of Noise in Electrical Machines 150

Zo ne ty

100

Au d

ib

ili

80

50

60

Th

ld

of

Au d

ib

50

10- 3

10- 8

40

re sh o

0 20

Loudness Levels

120 100

100

Sound Intensity (W/m 2 )

10- 2

Threshold of Pain Sound Intensity Levels (dB)

3

20 0

ili

ty

200

500 1000 2000 Frequency, Hz

10- 13 5000 10000 20000

Figure 1.1 Sound intensity and audibility zone as a function of frequency. dB 140 Threshold of Pain 130 120 Jet Airliner 110 100 Heavy City Traffic 90 Beginning of Hearing Damage 80 70 60 50 40

Average Street Traffic Normal Conversation Business Office Living Room

30 20 10 0

Threshold of Audibility

Figure 1.2 Comparison of some environmental noise levels.


DK3193

noise

November 3, 2005

15:28

4

Noise of Polyphase Electric Motors Table 1.1 Typical sound power levels. Source of noise Quietest audible sound for persons under normal conditions Rustle of leaves Soft whisper, room in a quiet dwelling at midnight Voice, low Mosquito buzzing Department store, clothing department Modern elevator propulsion motor Normal conversation Bird singing Large department store Busy restaurant or canteen Voice, conversation 10-kW, 4-pole cage induction motor Normal street traffic Pneumatic tools Alarm clock ringing Buses, trucks, motorcycles Small air compressors Loud symphonic music Lawn mower Your boss complaining Heavy city traffic Air compressor Heavy diesel vehicle Permanent hearing loss (exposed full-time) Car on highway Steel plate falling Magnetic drill press Vacuum pump Hard rock music Jet passing overhead Jackhammer Jolt squeeze hammer Jet plane taking off Saturn rocket


Sound power level, dB(A) 10 15 30 40 45 48 50 55 60 70 75 80

90

92 95 100 105 106 108 110 115 120 122 150 200

DK3193

noise

November 3, 2005

15:28

Generation and Radiation of Noise in Electrical Machines

5

1.3 Sources of noise in electrical machines The frequency of interest for vibration is generally within 0 to 1000 Hz, and for noise is over 1000 Hz. Vibration and noise produced by electrical machines can be divided into three categories: • electromagnetic vibration and noise associated with parasitic effects due to higher space and time harmonics, eccentricity, phase unbalance, slot openings, magnetic saturation, and magnetostrictive expansion of the core laminations; • mechanical vibration and noise associated with the mechanical assembly, in particular bearings; • aerodynamic vibration and noise associated with flow of ventilating air through or over the motor. The load induced sources of noise include: • noise due to coupling of the machine with a load, e.g., shaft misalignment, belt transmission, elevator sheave with ropes, tooth gears, coupling, reciprocating compressor; • noise due to mounting the machine on foundation or other structure. The noise from its source is transmitted through the medium (structure, air) to the recipient (human being, sensor) of the noise. The process of noise generation and transmission in electrical machines is illustrated in Figure 1.3. Basics of acoustics are explained in Appendix A.

1.3.1

Electromagnetic sources of noise

Electromagnetic vibration and noise are caused by generation of electromagnetic fields (Chapter 2). Both stator and rotor excite magnetic flux density waves in the air gap. If the stator produces Bm1 cos(ω1 t + kα + φ1 ) magnetic flux density wave and rotor produces Bm2 cos(ω2 t +lα + φ2 ) magnetic flux densisty wave, then their product is 0.5Bm1 Bm2 cos[ω1 + ω2 )t + (k + l)α + (φ1 + φ2 )] + 0.5Bm1 Bm2 cos[ω1 − ω2 )t + (k − l)α + (φ1 − φ2 )]

(1.1)

where Bm1 and Bm2 are the amplitudes of the stator and rotor magnetic flux density waves, ω1 and ω2 are the angular frequencies of the stator and rotor magnetic fields, φ1 and φ2 are phases of the stator and rotor magentic flux desnity waves, k = 1, 2, 3, . . ., and l = 1, 2, 3, . . .. The product expressed by Equation 1.1 is proportional to magnetic stress wave in the air gap with amplitude Pmr = 0.5Bm1 Bm2 , angular frequency ωr = ω1 ±ω2 , order r = k ±l and phase φ1 ±φ2 . The magnetic stress (or magnetic pressure) wave acts in radial directions on the stator and rotor active surfaces causing the deformation and, hence, the vibration and noise. The slots, distribution of windings in slots, input current waveform distortion, air gap permeance fluctuations, rotor eccentricity, and phase unbalance give


DK3193

noise

November 3, 2005

15:28

6


Figure 1.3 Noise generation and transmission in electrical machines.

rise to mechanical deformations and vibration. Magnetomotive force (MMF) space harmonics, time harmonics, slot harmonics, eccentricity harmonics, and saturation harmonics, produce parasitic higher harmonic forces and torques. Especially, radial force waves in a.c. machines, which act both on the stator and rotor, produce deformation of the magnetic circuit. The stator-frame (or stator-enclosure) structure is the primary radiator of the machine noise. If the frequency of the radial force is close to or equal to any of the natural frequencies of the stator–frame system, resonance occurs, leading to the stator system deformation, vibration, and acoustic noise. Magnetostrictive noise of electrical machines in most cases can be neglected due to low frequency 2 f and high order r = 2 p of radial forces, where f is the fundamental frequency and p is the number of pole pairs. However, radial forces due to the magnetostriction can reach about 50% of radial forces produced by the air gap magnetic field. In inverter fed motors, parasitic oscillating torques are produced due to higher time harmonics in the stator winding currents. These parasitic torques are, in general, greater than oscillating torques produced by space harmonics. Moreover,


DK3193

noise

November 3, 2005

15:28


7

the voltage ripple of the rectifier is transmitted through the intermediate circuit to the inverter and produces another kind of oscillating torque [200].

1.3.2

Mechanical sources of noise

Mechanical vibration and noise (Chapter 7) is mainly due to bearings, their defects, journal ovality, sliding contacts, bent shaft, rotor unbalance, shaft misalignment, couplings, U-joints, gears etc. The rotor should be precisely balanced as it can significantly reduce the vibration. The rotor unbalance causes rotor dynamic vibration and eccentricity which in turn results in noise emission from the stator, rotor, and rotor support structure. Both rolling and sleeve bearings are used in electrical machines. The noise due to rolling bearings depends on the accuracy of bearing parts, mechanical resonance frequency of the outer ring, running speed, lubrication conditions, tolerances, alignment, load, temperature, and presence of foreign materials. The noise due to sleeve bearings is generally lower than that of rolling bearings. The vibration and noise produced by sleeve bearings depends on the roughness of sliding surfaces, lubrication, stability and whirling of the oil film in the bearing, manufacture process, quality, and installation.

1.3.3

Aerodynamic noise

The basic source of noise of an aerodynamic nature (Chapter 7) is the fan. Any obstacle placed in the air stream produces noise. In nonsealed motors, the noise of the internal fan is emitted by the vent holes. In totally enclosed motors, the noise of the external fan predominates. According to the spectral distribution of the fan noise, there is broad-band noise (100 to 10, 000 Hz) and siren noise (tonal noise). Siren noise can be eliminated by increasing the distance between the impeller and the stationary obstacle.

1.4 Energy conversion process Figure 1.4 shows how the electrical energy is converted into acoustic energy in an electrical machine. The input current interacts with the magnetic field producing high-frequency forces that act on the inner stator core surface (Figure 1.5). These forces excite the stator core and frame in the corresponding frequency range and generate mechanical vibration and noise. As a result of vibration, the surface of the stator yoke and frame displaces with frequencies corresponding to the frequencies of forces. The surrounding medium (air) is excited to vibrate, too, and generates acoustic noise. The radiated acoustic power is very small, approximately 10−6 to 10−4 W for an electrical motor rated below 10 kW. It is, therefore, not easy to calculate the acoustic power with reasonable accuracy.


DK3193

noise

November 3, 2005

15:28

8


Electric Power Forces Supply Electromagnetic System

Mechanical System

Displacement

Acoustic Environment

Acoustic Noise

Figure 1.4 Conversion of electric energy into acoustic energy in electrical machines. The stator and frame assembly, as a mechanical system, is characterized by a distributed mass M, damping C, and stiffness K . The electromagnetic force waves excite the mechanical system to generate vibration. The amplitude of vibration is a function of the magnitude and frequency of those forces (Appendix D). The mechanical system can be simply described by a lumped parameter model with N degrees of freedom in the following matrix form ¨ + [C]{q} [M]{q} ˙ + [K ]{q} = {F(t)}

(1.2)

where q is an (N , 1) vector expressing the displacement of N degrees of freedom, {F(t)} is the force vector applying to the degrees of freedom, [M] is the mass matrix, [C] is the damping matrix and [K ] is the stiffness matrix. Equation 1.2

Figure 1.5 Mechanism of generation of vibration and noise in electrical machines.


DK3193

noise

November 3, 2005

15:28


9

can be solved using a structural finite element method (FEM) package. In practice, there are difficulties with predictions of the [C] matrix for laminated materials, physical properties of materials, and errors in calculation of magnetic forces [213].

1.5

Noise limits and measurement procedures for electrical machines

Acoustic quantities (Appendix A) can be expressed in terms of the sound pressure level (SPL) or sound power level (SWL). The sound pressure level is the most common descriptor used to characterize the loudness of an ambient sound level. In general, it is more complicated to measure the sound power level than the sound pressure level. The sound power level measurement is independent of the surface of the machine and environmental conditions. According to the National Electrotechnical Manufacture’s Association (NEMA) [162], the sound pressure level L p A can be related to the sound power level L W A in dB(A), as follows L p A = L W A − 10 log10

2πrd2 S0

(1.3)

where L p A is the average sound pressure level in a free-field over a reflective plane on a hemispherical surface at 1 m distance from the machine, rd = 1.0 + 0.5lm , lm is the maximum linear dimension of the tested machine in meters, and S0 = 1.0 m2 . The noise of electrical machines depends on the type of the machine, its topology, size, design, construction, enclosure, materials, manufacturing, rated power, speed, tolerances, mounting, support, foundation, coupling, bearings, supply, load, etc. Some consequences of noise as, for example, manufacturing, mounting or support are very difficult to predict. In general, the equations for sound pressure level or sound power level as functions of rotational speed n, rated output power Pout , or torque T have the following forms: L p1 = A1 + B1 log10 n

(1.4)

L p2 = A2 + B2 log10 Pout L p3 = A3 + B3 log10 T

(1.5) (1.6)

where A1 , A2 , A3 , B1 , B2 , and B3 are constants. When a motor is tested at no load under conditions specified by [162], the sound power level of the motor shall not exceed values given in Tables 1.2 and 1.3. The enclosures of motors are an open drip proof machine (ODP) type, totally enclosed fan cooled machine (TEFC) type, and weather protected type II machine (WPII) type. The WPII machine is a guarded machine with its ventilating passages at both intake and discharge so arranged that high velocity air and airborne particles blown into the machine by storms or high winds can be discharged without entering


DK3193

noise

November 3, 2005

15:28

10


Table 1.2 Maximum A-weighted sound power levels L w A in dB(A) at no load for motors with rated speeds 1200 rpm and less according to NEMA [162]. Rated power kW hp 0.37 0.5 0.5 0.75 0.75 1.0 1.1 1.5 1.5 2.0 2.2 3.0 3.0 5.0 5.5 7.5 7.5 10 11 15 15 20 17 25 22 30 30 40 40 50 45 60 55 75 75 100 100 125 110 150 150 200 185 250 220 300 260 350 300 400 350 450 370 500 450 600 520 700 600 800 670 900 750 1000 930 1250 1,100 1500 1,300 1750 1,500 2000

900 rpm and less ODP TEFC WPII 67 67 67 67 69 69 69 69 70 72 70 72 73 76 73 76 76 80 76 80 79 83 79 83 81 86 81 86 84 89 84 89 87 93 87 93 93 96 92 95 97 92 95 97 92 95 97 92 98 100 96 98 100 96 98 100 96 99 102 98 99 102 98 99 102 98 99 102 98 101 105 100 101 105 100 101 105 100 101 105 100 103 107 102 103 107 102 103 107 102


901 to 1200 rpm ODP TEFC WPII 65 65 67 67 72 72 76 76 81 81 83 83 86 86 88 88 91 91 96 99 99 99 99 102 102 102 102 102 105 105 105 105 107 107 107

64 64 67 67 71 71 75 75 80 80 83 83 86 86 90 90 94 94 98 100 100 100 100 103 103 103 103 103 106 106 106 106 109 109 109

97 97 97 97 99 99 99 99 99 101 101 101 101 103 103 103

DK3193

noise

November 3, 2005

15:28


11

Table 1.3 Maximum A-weighted sound power levels L w A in dB(A) at no load for motors with rated speeds 1201 to 3600 rpm according to NEMA [162]. Rated power kW hp 0.75 1.0 1.1 1.5 1.5 2.0 2.2 3.0 4.0 5.0 5.5 7.5 7.5 10 11 15 15 20 17 25 22 30 30 40 40 50 45 60 55 75 75 100 100 125 110 150 150 200 185 250 220 300 260 350 300 400 335 450 370 500 450 600 520 700 600 800 670 900 750 1000 930 1250 1,100 1500 1,300 1750 1,500 2000 1,700 2250 1,850 2500 2,250 3000

1201 to 1800 rpm ODP TEFC WPII 70 70 70 70 70 70 72 74 73 74 76 79 76 79 80 84 80 84 80 88 80 88 84 89 84 89 86 95 86 95 89 98 89 100 93 100 93 103 103 105 99 103 105 99 103 105 99 103 105 99 106 108 102 106 108 102 106 108 102 106 108 102 108 111 104 108 111 104 108 111 104 108 111 104 109 113 105 109 113 105 109 113 105 109 113 105 110 115 106 110 115 106


1801 to 3600 rpm ODP TEFC WPII 76 76 76 80 80 82 82 84 84 86 86 89 89 94 94 98 98 101 101 107 107 107 107 110 110 110 110 111 111 111 111 112 112 112 112 114

85 85 88 88 91 91 94 94 94 94 100 100 101 101 102 104 104 107 107 110 110 110 110 113 113 113 113 116 116 116 116 118 118 118 118 120

102 102 102 102 105 105 105 105 106 106 106 106 107 107 107 107 109

DK3193

noise

November 3, 2005

15:28

12


Table 1.4 Expected incremental increase over no-load condition in A-weighted sound power levels L w A , dB(A) for rated load condition for single-speed, threephase, cage induction motors according to NEMA [162] and IEC 60034-9 Standards [93]. Rated power kW hp 1 to 11 1.0 to 15 1 to 37 15 to 50 37 to 110 50 to 150 110 to 400 150 to 500

2p = 2 2 2 2 2

Number of poles 2p = 4 2p = 6 5 7 4 6 3 5 3 4

2p = 8 8 7 6 5

the internal ventilating passages leading directly to the electric parts of the machine itself. The sound power level at rated load should be adjusted according to Table 1.4. The increase in the sound power level under load is mostly due to the change in the air gap magnetic flux density harmonic amplitudes. This effect can be expressed by the following equation [137] Bload 2 L W = 10 log10 (1.7) Bnoload Table 1.5 shows maximum sound pressure level at 1 m from the machine surface according to IEC 60034-9 Standards [93]. Table 1.6 shows maximum sound power level according to IEC 60034-9 Standards [93]. The sound pressure level spectrum is the distribution of effective sound pressures measured as a function of frequency in specified frequency bands. It can also be defined as the resolution of a signal into components, each of different frequency and different amplitude (Figure 1.6). If the sound pressure level spectrum is given in a form of the Fourier series p=

kmax

Pmk sin(ωk t + φk )

(1.8)

k=1

where Pmk is the amplitude of the kth harmonic, ωk = 2π k f is the angular fequency of the kth harmonic, and φk is the phase angle for the kth harmonic, the overall sound pressure level is calculated as a sum of amplitudes squared, i.e., P=

kmax

2 Pmk

W.

(1.9)

k=1

The sound pressure level in dB is then calculated according to Equation A.25. The broad-band noise is the noise in which the acoustic energy is distributed over a relatively wide range of frequencies. The spectrum is generally smooth and continuous. The narrow-band noise is the noise in which the acoustic energy is concentrated in a relatively narrow range of frequencies. The spectrum will generally show a localized “hump” or peak in amplitude. Narrow-band sound may be superimposed on broad-band sound.


DK3193

noise

November 4, 2005

14:37


13

Table 1.5 IEC 60034-9 limits for sound pressure level at 1 m from machine surface, dB(A) [93]. Rated power kW Pout < 1.1 1.1 < Pout < 2.2 2.2 < Pout < 5.5 5.5 < Pout < 11 11 < Pout < 22 22 < Pout < 37 37 < Pout < 55 55 < Pout < 110 110 < Pout < 220 220 < Pout < 400 Rated power kW Pout < 1.1 1.1 < Pout < 2.2 2.2 < Pout < 5.5 5.5 < Pout < 11 11 < Pout < 22 22 < Pout < 37 37 < Pout < 55 55 < Pout < 110 110 < Pout < 220 220 < Pout < 400

1.6

n < 960 rpm ODP TEFC 67 69 72 72 75 75 78 77.5 79.5 78.5 80.5 82 94 85 87 86 88 1900 < n < 2360 ODP TEFC 74 78 82 81 86 83.5 87.5 85.5 89.5 88 94 90.5 93.5 93 96 94 98

960 < n < 1320 ODP TEFC 70 70 74 75 78 78 82 80.5 83.5 82.5 85.5 85 89 87 91 89 92 2360 < n < 3150 ODP TEFC 75 80 83 84 87 86.5 90.5 88.5 92.5 93 96 92.5 93.5 95 98 95 99

1320 < n < 1900 ODP TEFC 71 73 77 81 81 81.5 85.5 83 86 86 88 88.5 91.5 90.5 93.5 92.5 95.5 3150 < n < 3750 ODP TEFC 77 82 85 87 90 90 93 92 95 95.5 98.5 95 98 96 100 98 102

Deterministic and statistical methods of noise prediction

In the efforts to predict the noise emitted from an electric machine, there are two approaches: deterministic and statistical methods. In the deterministic method, shown in Figure 1.7a, the electromagnetic forces acting on a motor structure have to be calculated from the input currents and voltages using an electromagnetic analytical model [254] or the FEM model [226]. The vibration characteristics are then determined using a structural model normally based on the FEM [223, 230]. By using the vibration velocities on the motor structure predicted from the structural model, the radiated sound power level can then be calculated on the basis of an acoustic model.1 The acoustic model may be formulated using either the FEM or boundary-element method (BEM). Generally, for calculating 1 The only commercial FEM software for calculating the electromagnetic field, stator vibration mode shapes, and acoustic noise of electrical machines is JMAG-Studio developed by Japan Research Institute, Ltd., Engineering Technology Division, Tokyo-Osaka-Nagoya, http://www.iri.co.jp/proeng/jmag/e/jmg/.


DK3193

noise

November 3, 2005

15:28

14

Noise of Polyphase Electric Motors Table 1.6 IEC 60034-9 limits for sound power level, dB(A) [93].

Rated power n < 960 rpm 960 < n < 1320 kW ODP TEFC ODP TEFC Pout < 1.1 76 79 1.1 < Pout < 2.2 79 80 2.2 < Pout < 5.5 82 84 5.5 < Pout < 11 82 85 85 88 11 < Pout < 22 86 89 89 93 22 < Pout < 37 89 91 92 95 37 < Pout < 55 90 92 94 97 55 < Pout < 110 94 96 97 101 110 < Pout < 220 98 100 100 104 220 < Pout < 400 100 102 103 106 Rated power 1900 < n < 2360 2360 < n < 3150 kW ODP TEFC ODP TEFC Pout < 1.1 83 84 1.1 < Pout < 2.2 87 89 2.2 < Pout < 5.5 92 93 5.5 < Pout < 11 91 96 94 97 11 < Pout < 22 94 98 97 101 22 < Pout < 37 96 100 99 103 37 < Pout < 55 99 103 101 105 55 < Pout < 110 102 105 104 107 110 < Pout < 220 105 108 107 110 220 < Pout < 400 107 111 108 112

Figure 1.6 Sound pressure level spectrum.


1320 < n < 1900 ODP TEFC 80 83 87 88 91 92 96 94 97 97 99 99 103 103 106 106 109 3150 < n < 3750 ODP TEFC 86 91 95 97 100 100 103 102 105 104 107 106 109 108 112 110 114

DK3193

noise

November 3, 2005

15:28

Generation and Radiation of Noise in Electrical Machines Input Data

Input Data

Electromagnetic Force Model

Electromagnetic Force Model

Force

15

Force Mobility Model Input Power

Structural Model Vibration Velocity Acoustic Model

Sound Power Level (a)

Statistical Energy Model Vibration Power Radiation Efficiency Models

Sound Power Level (b)

Figure 1.7 Flowcharts for noise prediction: (a) deterministic method; (b) statistical method. the noise radiated into a space, the BEM is preferred because only the surface of the motor needs to be discretized and the space does not have to be discretized. Although, the analytical and FEM/BEM numerical approaches seem to work well, there are quite a number of limitations for it to be applied in practice (Section 1.8). In the deterministic approach, sometimes, simplified models can be utilized and analytical calculations can be implemented by writing a Mathcad2 or Mathematica3 computer program for fast prediction of the sound power level spectrum generated by magnetic forces. The accuracy due to physical errors may not be high, but the time of computation is very short and it is very easy to introduce and manage the input data set. The main program consists of the input data file, electromagnetic module, structural module (natural frequencies of the stator system), and acoustic module. The following effects can be included: phase current unbalance, higher space 2 Industry

standard technical calculation tool for professionals, educators, and college students. integrated technical computing environment used by scientists, engineers, analysts, educators, and college students. 3 Fully


DK3193

noise

November 3, 2005

15:28

16


harmonics, higher time harmonics, slot openings, slot skew, rotor static eccentricity, rotor dynamic eccentricity, armature reaction, magnetic saturation. An auxiliary program calculates the torque ripple, converts the tangential magnetic forces into equivalent radial forces, and transfers radial forces due to the torque ripple to the main program. The input data file contains the dimensions of the machine and its stator and rotor magnetic circuit, currents (including unbalanced system and higher time harmonics), winding parameters, material parameters (specific mass, Young modulus, Poisson’s ratio), speed, static and dynamic eccentricity, skew, damping factor as a function of frequency, correction factors, e.g., for the stator systems natural frequencies, maximum force order taken into consideration, minimum magnetic flux density to exclude all magnetic flux density harmonics below the selected margin. The rotor magnetic flux density waveforms are calculated on the basis of MMF waveforms and permeances of the air gap. Magnetic forces are calcualated on the basis of Maxwell stress tensor. The natural frequencies of the stator system are calculated with the aid of equations given in Chapter 5. Those values can be corrected with the aid of correction factors obtained, e.g., from the FEM structural package. Then, using the damping coeffcient as a function of frequency, amplitudes of radial velocities are calculated. The damping factor affects significantly the accuracy of computation. Detailed research has shown that the damping factor is a nonlinear function of natural frequencies. The radiation efficiency factor (Chapter 6), acoustic impedance of the air and amplitudes of radial velocities give the sound power level spectrum (narrow band noise). The overall noise can be found on the basis of Equation 1.9. The overall sound power level calculated in such a way is lower than that obtained from measurements because computations include only the noise of magnetic origin (mechanical noise caused by bearings, shaft misalignment, and fan is not taken into account) and usually, the calculation is done for low number of harmonics of magnetic flux density waves. The FEMs/BEMs, by their nature, are limited to low frequencies. This is because the number of elements required for the model increases by a factor of 8 when the upper frequency of interest is doubled and the number of vibration modes increases significantly with frequency [230]. If the FEMs/BEMs are applied to a large motor for frequencies up to 10,000 Hz, the number of elements and the computing time required will become prohibitive, as discussed in [223]. A method that is particularly suitable for calculations of noise and vibration at high frequencies is the so-called statistical energy analysis (SEA) (Chapter 10), which has been applied with success to a number of mechanical systems such as ship, car, and aircraft structures [140]. This method, however, was applied for the first time to electrical motors in 1999 [43, 223, 230]. The method basically involves dividing a structure (such as a motor) into a number of subsystems and writing the energy balance equations for each subsystem, thus allowing the statistical distribution of energies over various frequency bands to be determined. This method is normally valid for high frequencies where the modal overlap is high [140]. An outline of the calculation procedure using the statistical method is given in Figure 1.7b. The main advantage of the statistical approach is that it does not require all the


DK3193

noise

November 4, 2005

14:37


17

details to be modeled. The accurate distribution of the electromagnetic force might not be so important; only the total force in a frequency band is required. Thus, the electromagnetic force needs not be calculated using a FEM model and an approach such as that adopted by Cho and Kim [31] might be suitable. By considering the motor as a simple cylindrical shell, the input power due to this electromagnetic force can be formulated using an analytical ”mobility” model. By invoking a statistical energy model, this input power can then be distributed as vibrational power to different subsystems which make up the motor. If the sound radiation efficiencies of these subsystems are known, then the sound power due to each subsystem can be calculated. Since a motor structure can be decomposed into simple structural elements such as cylindrical shells, plates, and beams, the radiation efficiencies of these simple structural elements can be determined analytically, as depicted in the radiation efficiencies model in Figure 1.7b.

1.7

Economical aspects

Figure 1.8 shows the distribution of the magnetic flux density in the magnetic circuit of a 4-pole permanent magnet (PM) brushless machine. The magnetic flux density in the stator return path (yoke) is proportional to the magnetic flux density in the air gap and can be a measure of both the noise of electromagnetic origin and machine 1.8795e+000 1.8012e+000 1.7229e+000 1.6446e+000 1.5663e+000 1.4879e+000 1.4096e+000 1.3313e+000 1.2530e+000 1.1747e+000 1.0964e+000 1.0181e+000 9.3975e−001 8.6144e−001 7.8313e−001 7.0481e−001 6.2650e−001 5.4819e−001 4.6988e−001 3.9156e−001 3.1325e−001 2.3494e−001 1.5663e−002 7.8313e−002 0.0000e+000

Figure 1.8 (See color insert following page 236.) Distribution of the magnetic flux density in the cross section of a 4-pole brushless machine with surface PMs, as obtained from the 2D FEM.


DK3193

noise

November 3, 2005

15:28

18


Noise, dB; cost × 200 m.u.

95 90

cost, m.u.

85

noise, dB

80

1 m.u. (monetary unit) = cost of 1 kg of laminations or 4.5 kg of copper

75 70 65 60 55 50 0

0.2

1.2 1 0.8 0.6 0.4 1.4 1.6 Magnetic Flux Density in the Stator Yoke, T

1.8

2

Figure 1.9 Noise level and cost plotted against magnetic flux density (MFD) in the stator yoke for a 200 kW induction motor [2].

cost. Figure 1.9 shows the noise and total cost of an induction machine rated at 200 kW, 50 Hz, 380 V, 1480 rpm [2]. To keep the cost independent of inflation, an arbitrary monetary unit (m.u.) has been used that is equal to the price of 1 kg of steel sheets. Using this unit, the copper wire costs 4.5 m.u./kg and aluminum costs 3 m.u./kg [2]. The minimum of cost is different than the minimum of noise. Therefore, low magnetic flux density (MFD) means low level of noise and, vice versa, increased utilization of the magnetic circuit results in increased noise. The minimum of cost is for the MFD in the stator yoke in the range from 0.6 to 1.0 T. The minimum of noise is for lower MFDs; however, due to increase in dimensions and mass of active materials the cost increases sharply at low MFDs in the stator yoke.

1.8

Accuracy of noise prediction

The results of both analytical and numerical noise prediction may significantly differ from measurements. Forces that generate vibration and noise are only small fraction of the main force produced by the interaction of the fundamental current and the fundamental normal component of the magnetic flux density. The power converted into acoustic noise is only approximately 10−6 to 10−4 of the electrical input power. The accuracy of the predicted, say, sound power level spectrum depends not only on how accurate the model is, but also how accurate are the input data, e.g., level of current unbalance, angle between the stator current and q-axis (in PM brushless machines), influence of magnetic saturation on the equivalent slot opening, damping factor, elasticity modulus of the slot content (conductors, insulation, encapsulation), higher time harmonics of the input current (inverter-fed motor), etc.


DK3193

noise

November 3, 2005

15:28


19

All the above input data are difficult to predict with sufficient accuracy. Listed below are the common problems encountered in the analytical and numerical noise prediction [213, 220]: 1. The most difficult task in analytical calculation of sound power level radiated by electrical machines is the accurate prediction of the natural frequencies of the stator structure. At present, the best method to calculate the natural frequencies of the stator is to use the FEM. This is the only technique that can take into account with reasonable accuracy the end bells, mountings (feets or flanges) and asymmetries due to, e.g., terminal boxes. 2. The calculation of matrices of the mass [M] and stiffness [K ] seems to be obvious in the FEM. However, the physical properties of the materials used in electrical machines design are not known. The anisotropy of laminations, internal stresses caused by manufacturing, and change in stiffness matrix [K ] due to temperature variation (differential thermal expansion of the laminations and housing) are mostly not taken into account. 3. The damping matrix [C] in the FEM is difficult to predict. There are no adequate models available for describing damping in laminated materials and structures composed of different types of materials, e.g. copper, insulation, epoxy, laminations. Practice shows that good values for the damping are absolutely essential for predicting accurate vibrational amplitudes. 4. The force vector {F(t)} has to be found in all points on the inner stator surface. Even the most accurate FEM programs introduce a lot of errors in force calculations [213]. Forces are usually calculated analytically in the preprocessor module on the basis of magnetic flux density harmonics or using a 2D FEM. 5. Because neither the analytical approach nor the FEM/BEM computations guarantee accurate results, the laboratory tests are always very important. 6. The main advantage of the SEA is that it does not require all details to be modeled. 7. The vibration and acoustic noise can be calculated on the basis of modal analysis which is free of calculating the electromagnetic forces [213]. Only flux linkages have to be calculated (Section 9.2.2). 8. The calculated noise level is rather lower than the measured noise level. The calculation is mostly done for low harmonic numbers of the air gap permeance. The measurement gives the total noise level due to all harmonics [220].


DK3193

noise

November 3, 2005

15:28

2 Magnetic Fields and Radial Forces in Polyphase Motors Fed with Sinusoidal Currents In this chapter the magnetic field in the air gap of polyphase electrical machines will be discussed. The polyphase winding, usually three-phase, is located in the stator core slots. It has been assumed that the stator is fed with a balanced system of sinusoidal phase currents. For polyphase machines with rectangular and trapezoidal current waveforms, e.g., permanent magnet (PM) brushless machines, this assumption is equivalent to taking into account only the fundamental stator current harmonics.

2.1 Construction of induction motors An induction motor consists of two main parts: stator core and rotor core both with polyphase windings (Figure 2.1a). The stator and rotor ferromagnetic cores are stacked of thin steel laminations. Windings are located in stator and rotor slots or can be made as slotless windings. The stator polyphase winding, in most cases, the three-phase winding has independent phase windings that are star (Y) or delta () connected and fed with a three-phase a.c. current system. Each phase winding consists of multiturn coils connected in series. Series connected coils can also be groupped in two or more parallel paths. The air gap between the stator and rotor core is small, usually less than 1 mm for modern small and medium power induction motors. The rotor winding can be made of (1) three-phase Y-connected multiturn coil groups the terminals of which are connected to three slip rings or (2) in the form of a cage winding. In the first case the induction motor is called the phasewound motor or slip ring motor. Brushes are connected to a three phase variable resistance (rheostat) and slide on slip rings. The variable resistance is necessary 21 Copyright © 2006 Taylor & Francis Group, LLC

DK3193

noise

November 3, 2005

15:28

22


(a)

(b)

Figure 2.1 Cross sections and magnetic flux lines of three-phase motors: (a) 4-pole induction motor; (b) 8-pole PM brushless motor. for starting to reduce the high inrush current and also for speed control. The maximum resistance is at the first instant of starting, when the stator electromotive force (EMF) E 1 = 0. As the speed increases, the stator EMF increases too and the stator and rotor currents decrease. When the speed reaches its nominal value, the slip rings can be short-circuited. The rheostat used for the speed control must be rated at larger power than that for starting. The speed decreases as the resistance increases. Speed control rheostat is designed for continuous operation while the starting resistance is smaller because it is designed only for the short time duty. In cage induction motors the rotor winding consists of axial bars short circuited by two end rings forming a “squirrel cage” winding. A cage winding made of aluminum alloy or brass is not insulated from the rotor laminated stack. The polyphase stator winding when fed with polyphase a.c. currents produces rotating magnetic field in the air gap. The stator rotating field induces currents in the rotor winding which in turn interact with the stator field to produce the electromagnetic torque. For motoring operation the rotor spins in the same direction as the stator rotating field. The speed of the stator rotating magnetic field f ns = (2.1) p where f is the stator current frequency and p is the number of pole pairs, is called the synchronous speed. The slip between the rotor and the stator magnetic field for the fundamental harmonic is defined as ns − nm s= (2.2) ns where n m = (1 − s) f / p is the rotor mechanical speed. The slip speed n s − n m expresses the speed of the rotor relative to the rotating magnetic field of the stator. The current in the rotor creates its own magnetic field rotating with a speed that is given by sf sn s = . (2.3) p


DK3193

noise

November 3, 2005

15:28

Magnetic Fields and Radial Forces

23

The frequency of the current in the rotor s f is called the slip frequency. Since n s (1 −s )+sn s = n s , the stator and rotor magnetic fields rotate with the same speed. If the stator fundamental field harmonic rotates with the speed given by Equation 2.1, the stator higher space harmonic waves will rotate with the speed nsν = ±

f ns =± . ν νp

(2.4)

The “+” sign relates to the wave rotating in the same direction as the fundamental wave of the magnetomotive force (MMF) (forward-rotating wave) and the “−” sign relates to the wave rotating in the opposite direction (backward-rotating wave). The slip for the νth space harmonic is sν =

±n s ν − n m n s − (n s − n m ) = 1 ∓ ν(1 − s ) =1∓ ±n s ν n s /ν

(2.5)

where s is the slip (Equation 2.2) for the fundamental space harmonic ν = 1 and n s /ν in the denominator is the speed of the rotor νth harmonic that can have either the “+” or “−” sign. The frequency of the current induced in the rotor winding by the magnetic flux of the stator νth harmonic is f 2ν = f sν = f [1 ∓ ν(1 − s )]

(2.6)

The sign “−” in Equations 2.5 and 2.6 is for higher harmonics rotating in the same direction as the fundamental ν = 1 and the “+” sign is for harmonics rotating in the oposite direction.

2.2

Construction of permanent magnet synchronous brushless motors

The stator of a PM synchronous brushless motor is similar to that of an induction motor (Figure 2.1b). The rotor consists of laminated stack with PMs. Rotor laminations are necessary to reduce the rotor core losses due to higher harmonics. PMs can be of surface, bread loaf, interior, or interior double-layer configuration (Figure 2.2). Stators of PM brushless motors can be sinusoidally excited or trapezoidally excited. In the first case the phase windings are fed with sinusoidal waveforms shifted by 360o /m 1 one from another (m 1 is the number of stator phases) and produce rotating magnetic field. The motor works as a synchronous motor and is called sinewave motor. The trapezoidally excited motor also called square wave motor is fed from from waveforms shifted by 360o /m 1 too, but those waveforms are rectangular or trapezoidal (Figure 2.3). Waveforms according to Figure 2.3b are produced when the stator current (MMF) is precisely synchronized with the rotor instantaneous position and frequency (speed). The most direct and popular method of providing the required rotor position information is to use an absolute angular position


DK3193

noise

November 3, 2005

15:28

24


q

q

d

S

N

S

S

S

S

N

q

N (b)

d

q

S

N

N

S

S

N

N

(a) S

d

S

N

d

N

S

N

S

N

N (d)

(c)

Figure 2.2 Rotors of PM brushless motors with: (a) surface PMs; (b) bread loaf PMs; (c) interior (embedded) PMs; (d) interior double-layer PMs.

i1 Phase A

i1

Phase B

Phase C ωt 0

60

12 0

18 24 0 0 (a)

30 0

360°

0

60

12 0

18 0

24 0

ωt 30 360° 0

(b)

Figure 2.3 Basic stator winding waveforms for three-phase PM brushless motors: (a) sinusoidal waveforms; (b) rectangular waveforms.


DK3193

noise

November 3, 2005

15:28


25

encoder mounted on the rotor shaft. Only two phase windings out of three conduct current simultaneously. Such a control scheme or electronic commutation is functionally equivalent to the mechanical commutation in d.c. motors. Therefore, those motors are called d.c. brushless motors.

2.3

A.C. stator windings

In a single-layer winding, only one coil side is located in each slot. The number of all coils is s1 /2 and the number of coils per phase is n c = s1 /(2m 1 ) where s1 is the number of stator slots and m 1 is the number of phases. In a double-layer winding two sides of different coils are accommodated in each slot. The number of all coils is s1 and the number of coils per phase is n c = s1 /m 1 . The number of slots per pole is s1 Q1 = (2.7) 2p where 2 p is the number of poles. The number of slots per pole per phase is q1 =

s1 . 2 pm 1

(2.8)

The number of conductors per coil can be calculated as • for a single-layer winding Nc =

a p aw N 1 a p aw N 1 a p aw N 1 = = nc s1 /(2m 1 ) pq1

(2.9)

• for a double-layer winding Nc =

a p aw N 1 a p aw N 1 a p aw N 1 = = nc s1 /m 1 2 pq1

(2.10)

where N1 is the number of turns in series per phase, a p is the number of parallel current paths, and aw is the number of parallel conductors. The number of conductors per slot is the same for both single-layer and double-layer windings, i.e., Nsl =

a p aw N 1 . pq1

(2.11)

The full coil pitch measured in terms of the number of slots is y1 = Q 1 where Q 1 is according to Equation 2.7. The short coil pitch can be expressed as y1 = β Q 1 =

wc Q1 τ

(2.12)

where w c is the coil pitch (coil span) measured in units of length and τ is the pole pitch. The coil pitch–to–pole pitch ratio is β=


wc . τ

(2.13)

DK3193

noise

November 3, 2005

15:28

26


The distribution factor of a coil group for the fundamental space harmonic ν = 1 is defined as the ratio of the phasor sum E q —to—arithmetic sum q1 E c of EMFs E c induced in each coil and expressed by the following equation: kd1 =

Eq sin(π/2m 1 ) = q1 E c q1 sin[π/(2m 1 q1 )]

(2.14)

24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

The pitch factor for the fundamental space harmonic ν = 1 is defined as the ratio of the phasor sum—to—arithmetic sum of the EMFs per coil side and expressed

U1

W2

U2

W1

V1

V2

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

(a)

V2

U1

W2

V1

U2

W1 (b)

Figure 2.4 Three-phase, 4-pole windings distributed in s1 = 24 slots with full pitch coil groups (q1 = 2): (a) single-layer, (b) double-layer.


DK3193

noise

November 3, 2005

15:28


27

π (2.15) k p1 = sin β 2 The winding factor for fundamental is the product of the distribution factor (Equation 2.14) times pitch factor (Equation 2.15), i.e.

as:

kw1 = kd1 k p1 .

(2.16)

The angle in electrical degrees between the neighboring slots is γ =

360o 2π p≡ p. s1 s1

(2.17)

Figure 2.4 shows simple three-phase, 4-pole windings distributed in s1 = 24 slots with full pitch coil groups. Stator winding factors for higher space harmonics ν = 2m 1 k ± 1 where k = 1, 2, 3, 4, . . . are expressed by similar equations to Equations 2.14, 2.15, and 2.16, i.e., • stator winding distribition factor for the ν-th space harmonic sin[νπ/(2m 1 )] q1 sin[νπ/(2m 1 q1 )] • stator winding pitch factor for the ν-th space harmonic πw νβπ c k p1ν = sin = sin ν 2 2 τ • the resultant winding factor for the ν-th space harmonic kd1ν =

kw1ν = kd1ν k p1ν .

(2.18)

(2.19)

(2.20)

The values of the stator winding distribution factor for higher space harmonics as a function of the number of slots per pole per phase q1 (Equation 2.8) for m 1 = 3 are given in Table 2.1. Table 2.1 shows that for some space harmonics and q1 ≥ 2 the distribution factor is the same as for the fundamental harmonic, i.e., kd1ν = ±kd1 . Those harmonics s1 ν = 2m 1 q1 k ± 1 = k ± 1 (2.21) p where k = 1, 2, 3, . . . , are called stator slot harmonics. The EMF induced by slot harmonics is not reduced. For q1 = 1 all harmonics become slot harmonics. The angle between two neighboring slots (coil sides) for the ν-th space harmonic is ν times larger. For slot harmonics 2π p s1 γν = νγ = k ± 1 = 2π k ± γ (2.22) p s1 where γ for the fundamental space harmonic ν = 1 is according to Equation 2.17. Similarly, the number of rotor slot harmonics is s2 µ=k ±1 (2.23) p where s2 is the number of rotor slots.


DK3193

noise

November 3, 2005

15:28

28


Table 2.1 Winding distribution factor kd1ν according to Equation 2.18 for threephase stator windings. ν 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51

2.4 2.4.1

q1 2 0.966 0.707 0.259 -0.259 -0.707 -0.966 -0.966 -0.707 -0.259 0.259 0.707 0.966 0.966 0.707 0.259 -0.259 -0.707 -0.966 -0.966 -0.707 -0.259 0.259 0.707 0.966 0.966 0.707

3 0.960 0.667 0.218 -0.177 -0.333 -0.177 0.218 0.667 0.960 0.960 0.667 0.218 -0.177 -0.333 -0.177 0.218 0.667 0.960 0.960 0.667 0.218 -0.177 -0.333 -0.177 0.218 0.667

4 0.958 0.653 0.205 -0.158 -0.271 -0.126 0.126 0.271 0.158 -0.205 -0.653 -0.958 -0.958 -0.653 -0.205 0.158 0.271 0.126 -0.126 -0.271 -0.158 0.205 0.653 0.958 0.958 0.653

5 0.957 0.647 0.200 -0.149 -0.247 -0.109 0.102 0.200 0.102 -0.109 -0.247 -0.149 0.200 0.647 0.957 0.957 0.647 0.200 -0.149 -0.247 -0.109 0.102 0.200 0.102 -0.109 -0.247

6 0.956 0.644 0.197 -0.145 -0.236 -0.102 0.092 0.173 0.084 -0.084 -0.173 -0.092 0.102 0.236 0.145 -0.197 -0.644 -0.956 -0.956 -0.644 -0.197 0.145 0.236 0.102 -0.092 -0.173

∞ 0.955 0.636 0.191 -0.136 -0.212 -0.087 0.073 0.127 0.056 -0.050 -0.091 -0.041 0.038 0.071 0.033 -0.051 -0.058 -0.027 0.026 0.049 0.023 -0.022 -0.042 -0.02 0.019 0.037

Stator winding MMF Single-phase stator winding

Figure 2.5 shows a flat model of an electrical machine with smooth stator and rotor cores. The stator winding consists of one coil per double pole pitch. The relative magnetic permeability of the stator and rotor cores µr → ∞. Coils are fed with d.c. current I and connected in series to create a single-phase winding. For a single coil with Nc turns and span w c = τ the Ampere’s circuital law can be written as · dl = 2H g = Nc I H (2.24)


DK3193

noise

November 3, 2005

15:28

Magnetic Fields and Radial Forces t

29

t

t

t

Stator

g

Rotor Fc(a )

ν=1

Fc

ν=3 p 2

NcI

-

5 p 2

p

p

3

2

2

p

3p

a

2p

Figure 2.5 Magnetic field and MMF of a single-phase full-pitch winding. where H is the magnetic field intensity in the air gap g. The MMF of a single coil regarded as a magnetic voltage drop across a single air gap is Nc I 2 Introducing the relative permeance (per unit area) of the air gap as µ0 g = , H/m2 g Fcg = H g =

(2.25)

(2.26)

the flat-topped value of a rectangular wave of the magnetic flux density induced by coils in the air gap is B g = µ0 H =

µ0 Nc I = g Fcg . g 2

(2.27)

In Equations 2.26 and 2.27 the air gap is assumed to be smooth (no stator and rotor slots) and the magnetic circuit is assumed to be unsaturated, i.e., the relative magnetic permeability of the stator and rotor cores µr → ∞. Similar to the magnetic flux density wave, the MMF wave is also rectangular (Figure 2.5) and can be resolved into Fourier series as ∞ π Fc (x) = (2.28) Fcmν cos ν x τ ν=1,3,5,...


DK3193

noise

November 3, 2005

15:28

30


where the higher space odd harmonics ν = 1, 3, 5, . . . and the peak value of the νth harmonic π π 4 4 0.5τ Fcmν = Fcg sin ν . (2.29) Fcg cos ν x d x = τ 0 τ νπ 2 The variation of the MMF Fc (x) with the linear coordinate x can also be expressed as a variation of the MMF with the angular coordinate α. The geometrical angle α=

1π 2π x= x pτ s 1 t1

(2.30)

where 2 pτ = s1 t1 , t1 is the stator slot pitch, and νπ x/τ is replaced by νpα. Considering an a.c. machine with sinusoidal excitation, the d.c. current I should be replaced by a sinusoidal current √ i(t) = 2I cos(ωt) (2.31) where ω = 2π f is the stator current angular frequency, f is the input frequency of the sinusoidal current, and I is the stator r ms current. Putting Equations 2.25 and 2.31 into Equations 2.28 and 2.29 √ π 2 2 Fcmν = Nc I sin ν (2.32) νπ 2 and Fc (x, t) =

π Fcmν cos(ωt) cos ν x . τ ν=1,3,5,... ∞

(2.33)

Equation 2.33 shows that the MMF of a phase winding consists of an infinite number of harmonic MMFs changing in time according to cosinusoidal law cos(ωt) and changing in space also according to cosinusoidal law cos ν πτ x . A doublelayer winding with short coil span w c < τ , i.e., β < 1 can be represented as two windings with full coil spans shifted by (1−β)τ one from another. The magnitude of the ν-th harmonic of the MMF per phase is

π π Fmν = 2Fqν cos ν(1 − β) = 2q1 Fcmν kd1ν cos ν(1 − β) (2.34) 2 2 where the magnitude of the MMF of a group of coils according to Equation 2.14 is Fqν = q1 Fcmν kd1ν (2.35) and the winding distribution factor kd1ν for the νth space harmonic is given by Equation 2.18. Putting Equation 2.32 into Equation 2.34 and regarding that νπ

π νβπ sin cos ν(1 − β) = sin = k p1ν 2 2 2 νπ νπ = 0; sin = ±1 for ν = 1, 3, 5, . . . cos 2 2


DK3193

noise

November 3, 2005

15:28


31

the magnitude of the MMF per phase (Equation 2.34) becomes √

π π 2 2 Nc I kd1ν sin ν cos ν(1 − β) Fmν = 2q1 νπ 2 2 √ 2 2 = 2q1 Nc kd1ν k p1ν I νπ

(2.36)

where the pitch factor k p1ν for the νth space harmonic is given by Equation 2.19. The number of turns Nc per coil of a double-layer stator winding is given by Equation 2.10 and the stator phase current I 1 = a p aw I

(2.37)

where I is the single conductor (aw = 1) coil current. Thus, the amplitude of the ν-th space harmonic of the MMF per phase √ 2 2 N1 kw1ν N1 kw1ν I1 ≈ 0.9 I1 . Fmν = (2.38) π νp νp The resultant winding factor kw1ν for the ν-th harmonic is determined by Equation 2.20. The MMF of a single-phase winding varying in space and time can be resolved into two waves rotating in opposite directions, i.e., F(x, t) =

π Fmν cos(ωt) cos ν x τ ν=1,3,5,... ∞

1 π 1 π Fmν cos ωt − ν x + Fmν cos ωt + ν x . (2.39) 2 ν τ 2 ν τ ∞

=

∞

Equation 2.39 for each harmonic describes two MMF waves with the same amplitudes 0.5Fmν rotating with the same angular speed ω/ p = 2π f / p in opposite directions. Harmonics 0.5Fmν cos(ωt − νπ x/τ ) rotate in the same direction as the fundamental stator magnetic field (and rotor). Those magnetic field waves are called forward-rotating waves. Harmonics 0.5Fmν cos(ωt + νπ x/τ ) rotate in the opposite direction and are called backward-rotating waves. The speed of the MMF waves can be found assuming that the waves are stationary for an observer moving with the angular speed ω, i.e., π cos ωt ∓ ν x = const. τ Thus,

π π x = const; ωdt ∓ ν d x = 0 τ τ and the linear synchronous speed of the νth harmonic wave is ωt ∓ ν

v sν =

1τ 1 f D1in 1 dx =± ω = ±2π = vs dt νπ ν p 2 ν


(2.40)

DK3193

noise

November 3, 2005

15:28

32


where ω = 2π f is the angular frequency for fundamental harmonic ν = 1, n s = f / p is the rotational speed of the fundamental frequency of the magnetic field called the synchronous speed (Equation 2.1), D1in = 2 pτ/π is the inner diameter of the stator, and v s = ωD1in /(2 p) = 2π( f / p)D1in /2 is the linear synchronous speed of the fundamental harmonic.

2.4.2

Three-phase stator winding

A three-phase symmetrical stator has three the same windings shifted in space by 2π/3 electrical degrees. The stator is fed with three sinusoidal currents i 1A , i 1B , and i 1C . For a balanced input current system, the amplitude of each phase current √ 2I1 is the same and phase shift is 2π/3 electrical degrees, i.e., √ i 1A = 2I1 cos(ωt) √ π (2.41) i 1B = 2I1 cos ωt − 2 3 √ π i 1C = 2I1 cos ωt − 4 . 3 Based on Equation 2.39 the space harmonics of the MMF produced by each phase winding are 1 π 1 π Fmν cos ωt − ν x + Fmν cos ωt + ν x 2 τ 2 τ

1 2π π = Fmν cos ωt − ν x + 0(ν − 1) 2 τ 3

1 2π π + Fmν cos ωt + ν x − 0(ν + 1) 2 τ 3

F1Aν =

2π π x− τ 3 1 2π π 2π + Fmν cos ωt − x− +ν ) 2 3 τ 3

1 2π π = Fmν cos ωt − ν x + 1(ν − 1) 2 τ 3

1 2π π + Fmν cos ωt + ν x − 1(ν + 1) 2 τ 3

F1Bν =

F1Cν =

1 Fmν cos 2

ωt −

2π 3

−ν

4π π x− τ 3 1 4π π 4π + Fmν cos ωt − x− +ν ) 2 3 τ 3 1 Fmν cos 2


ωt −

4π 3

−ν

(2.42)

DK3193

noise

November 3, 2005

15:28

Magnetic Fields and Radial Forces =

33

2π 1 π Fmν cos ωt − ν x + 2(ν − 1) 2 τ 3

1 2π π + Fmν cos ωt + ν x − 2(ν + 1) . 2 τ 3

According to Equation 2.42, harmonic waves of the MMF of each phase are shifted by the angle (ν ∓1)(2π )/3 one from each other. Harmonics ν = 3k = 3, 9, 15, . . . (in general, ν = m 1 k where k = 1, 3, 5, . . . and m 1 = 3) are shifted by (ν ∓ 1)(2π )/3 = (3k ∓ 1)(2π )/3 = 2π k ∓ 2π/3 or 120o . Thus, the sum of those harmonic waves is equal to zero, so that harmonics ν = 3k do not exist in three phase systems. For harmonic waves ν = 6k + 1 (in general, ν = 2m 1 k + 1) the shift angle is (ν − 1)(2π )/3 = 4kπ or 0o . The resultant wave is the arithmetic sum of waves in all three phases. For harmonic waves ν = 6k − 1 (in general, ν = 2m 1 k − 1) the shift angle is (ν − 1)(2π )/3 = 4kπ − 4π/3 or 240o . The resultant wave is 0. The resultant MMF of a three phase symmetrical winding fed with balanced three phase current system does not contain triple harmonic waves. Harmonics ν = 1, 7, 13, 19, . . . rotate forward and harmonics ν = 5, 11, 17, 23, . . . rotate backward. The linear speed of higher space harmonics is expressed by Equation 2.40. The magnitudes of harmonic MMFs according to Equations 2.38 and 2.43 are √ 3 3 2 N1 kw1ν N1 kw1ν Fmν = [Fmν ]m 1 =1 = I1 ≈ 1.35 I1 . (2.43) 2 π νp νp The total MMF of a three-phase symmetrical stator winding fed with a balanced current system ∞ π F1 (x, t) = (2.44) Fmν cos ωt ∓ ν x τ ν=6k±1 or F1 (α, t) =

∞

Fmν cos(ωt ∓ νpα) =

ν=6k±1

∞

Fmν cos(νpα ∓ ωt)

(2.45)

ν=6k±1

where the angle α is determined by Equation 2.30.

2.4.3

Polyphase stator winding

The time-space distribution of the MMF of the k-th phase winding of a symmetrical polyphase winding fed with a balanced current system can be expressed as

∞ 2π 1 π F1k (x, t) = [Fmν ]m 1 =1 {cos ωt − ν x + (k − 1)(ν − 1) 2 ν=1 τ m1

π 2π + cos ωt + ν x − (k − 1)(ν + 1) } (2.46) τ m1


DK3193

noise

November 3, 2005

15:28

34


where ω = 2π f is the angular frequency for the fundamental harmonic, τ is the pole pitch and ν = 2m 1 k ± 1,

k = 0, 1, 2, 3, 4, 5, . . .

(2.47)

The “−” sign before νπ x/τ is for the forward-rotating field and the “+” sign before νπ x/τ is for the backward-rotating field. The total MMF exicted by all m 1 phases is the sum of MMFs excited by each phase. Using the trigonometric identity cos(α ± β) = cos α cos β ∓ sin α sin β, the following equation can be written on the basis of Equation 2.46 F1ν (x, t) = F11 (x, t) + F12 (x, t) + · · · + F1m 1 (x, t) ∞ 1 π = Fmν cos ωt − ν x 2 ν=1 τ

1 2 × 1 + cos (ν − 1)2π + cos (ν − 1)2π + · · · m1 m1

∞ m1 − 1 1 π cos (ν − 1)2π − Fmν sin ωt − ν x m1 2 ν=1 τ

1 2 × 0 + sin (ν − 1)2π + sin (ν − 1)2π + · · · m1 m1

∞ m1 − 1 1 π sin (ν − 1)2π + Fmν cos ωt + ν x m1 2 ν=1 τ

1 2 × 1 + cos (ν + 1)2π + cos (ν + 1)2π + · · · m1 m1

∞ m1 − 1 1 π cos (ν + 1)2π + Fmν sin ωt + ν x m1 2 ν=1 τ

1 2 × 0 + sin (ν + 1)2π + sin (ν + 1)2π + · · · m1 m1

m1 − 1 sin (ν + 1)2π . (2.48) m1 Series in square brackets can be written as

1 2 m1 − 1 1+cos (ν ∓ 1)2π + cos (ν ∓ 1)2π + · · · + cos (ν ∓ 1)2π m1 m1 m1

sin(ν ∓ 1)π m1 − 1 sin(ν ∓ 1)π = cos (ν ∓ 1)2π = sin[(ν ∓ 1)π/m 1 ] m1 sin[(ν ∓ 1)π/m 1 ] (2.49)


DK3193

noise

November 3, 2005

15:28


35

Table 2.2 Band factor in Equation 2.51 for three-phase stator windings. ν

sin(ν−1)π sin[(ν−1)π/m 1 ]

sin(ν+1)π sin[(ν+1)π/m 1 ]

0 0 0 3 0 0 3 0 0 3 0 0 3 0

0 0 3 0 0 3 0 0 3 0 0 3 0 0

1 3 5 7 9 11 13 15 17 19 21 23 25 27

1 2 m1 − 1 0 + sin (ν ∓ 1)2π + sin (ν ∓ 1)2π + · · · + sin (ν ∓ 1)2π m1 m1 m1

sin(ν ∓ 1)π m1 − 1 = sin (ν ∓ 1)2π = 0. (2.50) sin[(ν ∓ 1)π/m 1 ] m1 Thus, the resultant MMF of a polyphase stator winding is π sin(ν − 1)π 1 cos ωt − ν x Fmν 2 ν=1 sin[(ν − 1)π/m 1 ] τ ∞

F1 (x, t) =

1 π sin(ν + 1)π cos ωt + ν x . Fmν 2 ν=1 sin[(ν + 1)π/m 1 ] τ ∞

+

(2.51)

The so-called band factor in Equation 2.51 is given in Table 2.2 by calculating sin(ν − 1)π = m1 sin[(ν − 1)π/m 1 ]

for

ν = 2km 1 + 1

k = 1, 2, 3, . . .

sin(ν + 1)π = m1 sin[(ν + 1)π/m 1 ]

for

ν = 2km 1 − 1

k = 1, 2, 3, . . . .


DK3193

noise

November 3, 2005

15:28

36


Otherwise, sin[(ν ∓ 1)π ]/ sin[(ν ∓ 1)π/m 1 ] = 0 including ν = 1. Finally, for higher space harmonics F1 (x, t) =

∞ 1 π [Fmν ]m 1 =1 m 1 cos ωt − ν x 2 ν=2km +1 τ 1

+

∞

1 2 ν=2km

1

π [Fmν ]m 1 =1 m 1 cos ωt + ν x . τ −1

The magnitude of the νth harmonic of the polyphase stator MMF is √ m1 m 1 2 N1 kw1ν Fmν = [Fmν ]m 1 =1 = I1 . 2 π νp

(2.52)

(2.53)

The winding factor kw1ν for the νth space harmonic is given by Equation 2.20. The peak value of the armature line current density or specific electric loading is defined √ as the number of conductors in all phases 2m 1 N1 times the peak stator current 2I1 divided by the armature stack length 2 pτ , i.e., √ m 1 2N1 I1 Am1 = . (2.54) pτ

2.5

Rotor magnetic field

The amplitude of the rotor MMF of an induction machine is √ m 2 2 N2 kw2 I2 Fm2 = π p

(2.55)

where N2 is the number of the rotor turns per phase, kw2 is the rotor winding factor for the fundamental space harmonic, and I2 is the rotor r ms current. The amplitude of the rotor MMF referred to the stator system √ m 1 2 N1 kw1 Fm2 = I2 (2.56) π p where the rotor current referred to the stator system is calculated as I2 =

m 2 N2 kw2 k s I2 . m 1 N1 kw1

(2.57)

The skewing factor for the fundamental harmonic is ks =

sin[π pbs /(t1 s1 )] sin(0.5π bs /τ ) = 0.5π bs /τ π pbs /(t1 s1 )

(2.58)

where bs is the slot skew. In practice bs = 2π τ/s1 = t1 or bs = 2π τ/s2 = t2 .


DK3193

noise

November 3, 2005

15:28


37

The peak value of the rotor line current density of an induction machine is expressed similar to that of the stator (Equation 2.54), i.e., √ m 2 2N2 I2 Am2 = . (2.59) pτ For a synchronous machine, as the rotor rotates at mechanical angular velocity m = , the excitation field and its space harmonics rotate all at the same mechanical angular velocity . The electrical angular frequencies of the rotor higher harmonics are all different and increase with the harmonic number, i.e., ωµ = µp

(2.60)

The magnetic flux density waveform (normal component) of the rotor of an induction machine is given by the equation b2 (α, t) =

∞

B2mµ cos µpα ∓ ωµ t + φµ .

(2.61)

µ=1

For the most important space harmonics the rotor angular frequency of an induction motor is ωµ = ω[(ks2 / p)(1 − s) + 1]. The magnitude of the rotor magnetic flux density can be found as [10] √ m 1 2 N1 kw1ν kw2µ µ0 B2mµ = I1 (2.62) π µp(1 + τ2ν )kw2ν kC g where kw1ν is the stator winding factor for the ν-th space harmonic according to Equation 2.20, kw2µ is the rotor winding factor for the rotor µ-th space harmonic, kw2ν is the rotor winding factor for the stator ν-th space harmonic, τ2ν = X 2ν / X mν is the leakage coefficient of the rotor winding for the ν-th stator harmonic, X 2ν is the rotor winding leakage reactance for the ν-th harmonic, and X mν is the mutual reactance for the ν-th harmonic.

2.6 2.6.1

Calculation of air gap magnetic field Effect of slots

The relationship between the air gap magnetic flux density b(α, t) and the MMF F(α, t) for uniform air gap g is b(α, t) = F(α, t)g (α)

(2.63)

where the relative permeance g (α) of the air gap is a function of the angle α given by Equation 2.30. Assuming a trapezoidal distribution of the magnetic flux density, the amplitude of the νth space harmonic will be equal to the flat-topped value of the flux density multiplied by the following coefficient of slot opening [87] kok =


sin[kρπb14 /(2t1 )] kρπb14 /(2t1 )

(2.64)

DK3193

noise

November 3, 2005

15:28

38


where t1 is the stator slot pitch, b14 is the stator slot opening and k is the number of harmonics which replace the air gap variation bounded by the active surfaces of ferromagnetic cores of the stator and rotor. The auxiliary function dependent on the relative slot opening is √ κ 2 1 + κ2 √ ρ= . (2.65) 5 + κ 1 + κ2 − 1 The relative slot opening can be estimated as follows • for induction machines

b14 g

(2.66)

b14 g + h M /µ0

(2.67)

κ= • for most PM brushless machines κ≈

where h M is the radial height of the PM per pole. The value of κ may affect the radial magnetic forces, but it is not necessarily a decisive factor in production of noise [220]. The relative permeance of the air gap in Equation 2.63 with slot opening effect being included is ∞ µ0 A0 + Ak cos(ks1 α) 2 gkC k=1,2,3 ∞ = g0 1 + Ak cos(ks1 α) = g0 λg1 (α)

g (α) =

(2.68)

k=1,2,3

where α = 2π x/(s1 t1 ) is according to Equation 2.30, s1 is the number of stator slots and t1 is the slot pitch. The permeances and Fourier coefficients in the above Equation 2.68 are • the constant component of the air gap relative permeance g0 =

A0 µ0 µ0 = = 2 kC g g

(2.69)

• the component of the relative specific permeance of the air gap varying with the angle α when only the stator core is slotted λg1 (α) = 1 +

∞

Ak cos(ks1 α)

(2.70)

k=1,2,3

• the relative value of harmonic permeances of the air gap dependent on the dimensions of the slot openings and air gap [40, 219], i.e., γ1 2 2 π Ak = g (α) cos(ks1 α)dα = −2g kok . (2.71) π 0 t1


DK3193

noise

November 3, 2005

15:28


39

It has been assumed that g (α) above the slot opening is approximated by a linear function. The second term in Equation 2.68 multiplied by g0 yields 2 g0 Ak = −2µ0 γ1 kok /t1 , i.e., as given in classical literature [40]. The angular displacement α according to Equation 2.30 can be expressed by the linear circumferential distance x. For induction machines the constant component (Equation 2.69) corresponds to the relative permeance of physical air gap g increased by Carter’s coefficient kC . For surface magnets g = gkC + h M /µrr ec and for buried magnets totally enclosed by laminations or mild steel g = gkC . The Carter’s coefficient kC =

t1 t1 − γ1 g

(2.72)

where

4

0.5κ arctan(0.5κ) − ln 1 + (0.5κ)2 . (2.73) π The relative permeance due to stator slot openings can also be calculated according to the method proposed by Weber [87, 236]. The coefficient of harmonic permeances in Equation 2.70 is

14 (k)2 Ak = −β(κ) sin(1.6π k) (2.74) 0.5 + kπ 0.78 − 2(k)2 γ1 =

where or

β(κ) = 0.1κ 0.5+1/κ

(2.75)

1 β(κ) = 0.5 1 − √ . 1 + κ2

(2.76)

The relative permeance according to Weber is g0 Ak = µ0 Ak /g where Ak is given by Equation 2.74. The parameter κ is according to Equation 2.66 or Equation 2.67 and b14 = . (2.77) t1 If both the stator and rotor core have slots, the resultant relative air gap permeance is g (α) = g0 λg1 (α)λg2 (α) (2.78) and the resultant Carter’s coefficient is kC = kC1 kC2

(2.79)

where λg1 and λg2 are relative specific permeances of the stator and rotor derived on the assumption that only one core has slots, kC1 is Carter’s coefficient for the stator, and kC2 is Carter’s coefficient for the rotor. The center axes of the stator and rotor teeth of an induction motor are aligned at the time instant t = 0 in the origin of coordinate system, i.e., α = 0 and the rotor angular frequency ω2 = 2π(1 − s) f = 2π(1 − s)n s p = 2π n m p = pm


DK3193

noise

November 3, 2005

15:28

40


where m is the mechanical angular speed of the rotor and n m is the rotational speed of the rotor in rev/s. Thus, the relative value of harmonic permance for the rotor is ∞ λg2 (α, t) = 1 + Al cos [ls2 (α − ω2 t)] . (2.80) l=1,2,3

Putting Equations 2.69, 2.70, and 2.80 into Equation 2.78 the resultant relative permeance of the air gap of slotted stator and rotor cores is g (α, t) = g0 λg (α, t) =

µ0 λg (α, t) kC g

(2.81)

where λg (α, t) = 1 + +

∞

Ak cos (ks1 α) +

k=1,2,3 ∞

∞

Al cos [ls2 (α − ω2 t)]

l=1,2,3

∞ 1 Ak Al {cos[(ls2 + ks1 )α − ls2 ω2 t] 2 k=1,2,3 l=1,2,3

× cos[(ls2 − ks1 )α − ls2 ω2 t]}.

(2.82)

The first term in Equation 2.82 describes the permeance of the equivalent uniform air gap g = kC g, second term describes harmonics of the stator permeance, third term describes harmonics of the rotor permeance, and the last term represents harmonics of permeance due to reciprocal effect of the stator and rotor. Harmonics of the magnetic flux density in the air gap are obtained by multiplying Equations 2.81 and 2.44 (distribution of the MMF) as given by Equation 2.63.

2.6.2

Effect of eccentricity

The variation of the air gap around the magnetic circuit periphery and with time is g(α, t) = R − r − e cos(α − ω t) = g[1 − cos(α − t)]

(2.83)

where g = R − r , R is the inner stator core radius and r is the outer rotor radius. The relative eccentricity is defined as =

e g

(2.84)

where e is the rotor (shaft) eccentricity and g is an ideal uniform air gap for e = 0. For the static eccentricity the angular frequency = 0. In the case of dynamic eccentricity: • for induction motors = (1 − s) =


f ω (1 − s) = 2π (1 − s) p p

(2.85)

DK3193

noise

November 3, 2005

15:28


41

• for synchronous motors f ω = 2π . (2.86) p p Equation 2.83 has been derived in Appendix B. The constant coefficient (Equation 2.69) in Fourier series (Equation 2.68) for a machine with eccentricity becomes [10]: = =

µ0 1 A0 √ = H/m2 (2.87) 2 gkC 1 − 2 The effect of the static and dynamic eccentricity on the higher harmonics of the air gap performance is difficult to calculate. Most methods are limited only to the influence of eccentricity on the fundamental harmonic of the air gap permeance [87], i.e., • the air gap relative permeance for k = 1 including the static eccentricity and variation of the permeance with slot openings γ1 2 ge,k=1 (α) = g0 Ak=1 λc1 cos(α) = −2µ0 kok λc1 cos(α) (2.88) t1 in which √ 1 − 1 − 2 λc1 = 2 √ (2.89) 1 − 2 • the air gap relative permeance for k = 1 including the dynamic eccentricity and variation of the permeance with slot openings g0 =

ged,k=1 (α, t) = g0 Ak=1 λc1d cos( t − α) γ1 2 = −2µ0 kok λc1d cos( t − α) (2.90) t1 where the angular speed equal to the rotor mechanical angular speed is according to Equation 2.85 or Equation 2.86. The angular displacement α in Equations 2.88 and 2.90 can be replaced by the linear displacement x according to Equation 2.30. To obtain λc1d , the relative static eccentricity in Equation 2.89 is replaced by the relative dynamic eccentricity d defined as ed (2.91) d = g in which ed denotes the dynamic eccentricity (linear radial dimension). To include the eccentricity in calculation of the relative air gap permeance, the permeances (Equation 2.88) and (Equation 2.90) should be added to Equation 2.68 or Equation 2.81. The effect of static and dynamic eccentricity can be taken into account according to Appendix B using the following coefficients √ 1 1 − 1 − 2 λge (α) ≈ √ +2 √ cos(α) (2.92) 1 + 2 1 − 2 √ 1 1 − 1 − 2 λged (α) ≈ + 2 cos ( t − α) . (2.93) 1 + d2 d 1 − d2


DK3193

noise

November 3, 2005

15:28

42


The resultant relative air gap permeance g (α, t) with slots and eccentricity effects taken into account is given in Appendix B by Equation B.5. Equation B.5 has been obtained by multiplying Equation 2.78 by the product λge (α)λged (α, t) which includes eccentricity. Since the relative eccentricity is small, i.e., ≤ 0.3, the permeance of the air gap due to eccentricity expressed in the form of Fourier series can be limited to the first two terms. For the dynamic eccentricity (neglecting the variation of the permeance with slot openings) ged (α, t) = g0 λged (α, t) 1 1 = µ0 [λoe + λc1 cos(α − t)] ≈ µ0 [1 + cos(α − t)]. gkC gkC (2.94) The wave of the magnetic flux density in the air gap pulsating with the eccentricity can be obtained by multiplying the first harmonic (second term) of Equation 2.94 by the fundamental stator MMF, i.e., cos(α − t)Fm1 cos( pα − ωt − φ) gkC = µ0 Fm1 cos[α(1 + p) − (ω + )t − φ)] + µ0 Fm1 2gkC 2gkC × cos[α(1 − p) + (ω − )t + φ)]

1 = µ0 + 1 − (ω + )t − φ) Fm1 cos pα 2gkC p

1 − 1 + (ω − )t + φ) + µ0 Fm1 cos pα 2gkC p = µ0 Fm1 cos[ν pα ∓ (ω ± )t ∓ φ)]. (2.95) 2gkC

b1 (α, t) ≈ µ0

The stator magnetic flux density higher harmonics due to eccentricity are ν = 1 ±

1 p

(2.96)

and the number of their pole pairs is p ± 1. For p = 1 the eccentricity harmonic numbers are ν = 2 or ν = 0. The second term in Equation 2.95 for ν = 0 represents an axial homopolar flux that can appear in a two-pole machine with eccentricity (Appendix B). The angular frequency for static eccentricity is ω = 2π f . For dynamic eccentricity and induction machines

1 ω ± = ω 1 ± (1 − s) . (2.97) p For synchronous machines s = 0 in Equation 2.97.


DK3193

noise

November 3, 2005

15:28


43

The stator magnetic flux density wave due to eccentricity generate additional magnetic flux density waves in the rotor [64, 200] b2 (α, t ) = B2m µ cos(µ p α − µ t − φµ ) (2.98) µ

where the rotor magnetic flux density higher harmonics due to eccentricity are s2 1 µ = k + 1 ± (2.99) p p and the angular frequency (see also Equation 2.134)

s2 µ = ων + ω k (1 − s ) p

(2.100)

where ων = 2π ν f . The rotor flux density waves originate from the stator flux density waves in the air gap with pole pairs p ± 1. For a synchronous machine, the slip in s = 0, Equation 2.100, and the number of rotor slots s2 = 2 p. The eccentricity fields contribute to the vibration and noise and produce the following effects: • cause unbalance magnetic pull that bends the shaft and magnifies the magnetic pull; • induce voltages in the parallel paths of stator windings and give rise to equalizing currents [123]; • reduce the mechanical stiffness of the shaft and the first critical speed of the rotor due to increased magnetic pull [41, 200].

2.6.3

Effect of magnetic saturation

The slot leakage and residual fields of the stator and rotor MMFs are large [87]. Consequently, the tooth tips become highly saturated (reduction of the relative magnetic permeability) which is equivalent to the increase in the slot opening from b14 to b14sat (Figure 2.6) according to Equation C.10. The maximum slot opening due to the tooth tip saturation corresponds to the region (along the pole pitch) of the maximum slot current. The fictitious slot opening b14sat is a periodic function with the period π/ p, i.e., equal to half the period of the working harmonic. Thus, the first two terms of Fourier series of the fictitious slot opening can be written as b14sat = b140 − b141 cos[2( pα − ωt)]

(2.101)

where b140 is the mean value of the slot opening along the pole pitch and b141 = b140 − b14 is the amplitude of the first harmonic of the slot opening distribution. The air gap permeance with respect to the magnetic saturation can be expressed by the following simplified equation [64, 179] (α, t) ≈ 0 + sat (α, t)


(2.102)

DK3193

noise

November 3, 2005

15:28

44

Noise of Polyphase Electric Motors b14 b14 sat Variation of Slot Opening

b14

t1 b14sat(a)

b14

a=

1 p x pt

t

t (a)

b140

(b)

Figure 2.6 Effect of magnetic saturation on the slot opening: (a) fictitious increase in the slot opening; (b) variation of a fictitious slot opening along the pole pitch due to saturation [87]. where sat (α, t) = −sat cos(2 pα − 2ω1 t − 2φs ).

(2.103)

Only the fundamental harmonic ν = 1 has been included.

2.6.4

Effect of rotor saliency

The effect of the PM rotor saliency can be included with the aid of the rotor magnetic flux density waveform normalized to B0 = 1 T. On the basis of Equation 2.61 the magnetic flux density waveform is b (α, t) = where

∞

b2 (α, t) = bµ cos(µpα ± ωµ t + φµ ) B0 µ=1

(2.104)

τ 4 1 cp sinh(α) − 6 cosh(α) c2p + (µπ/τ )2 µπ bt3 τ 2 2 +3 cosh(α) µπ τ − bp π π bt τ 2 1 τ 4 µπ/τ × sin µ sin µ +6 − + 2 τ τ c2p + (µπ/τ )2 µπ bt2 µπ π π bt × cosh(α) sin µ cos µ ; (2.105) 2 τ

4 bµ = τ

bt =

τ − bp ; 2


cp = 2

α ; bp

DK3193

noise

November 3, 2005

15:28


45

b p is the width of the pole shoe, and B0 bµ = B2mµ is given by Equation 2.61. The parameter α depends on the distribution of the normal component of the magnetic flux density in the range −0.5b p ≤ x ≤ 0.5b p . For α = 0 the magnetic flux density in the range −0.5b p ≤ x ≤ 0.5b p has constant value (flat curve). For embedded magnets with orifices in pole shoes, α = 0.5 . . . 1 and the magnetic flux density in the range −0.5b p ≤ x ≤ 0.5b p is a concave curve. Thus, the resultant air gap relative permeance with the rotor saliency effect taken into account is gsal (α, t = t0 ) = |b (α, t = t0 )|λg (α, t = t0 )

(2.106)

where b (α, t = t0 ) is according to Equation 2.104, λg (α, t = t0 ) is according to Equation 2.70 or Equation 2.82, and t0 is the arbitrarily chosen time instant. The resultant relative air gap permeance g according to Equation B.5 (Appendix B) with the effect of the rotor saliency taken into account plotted against the circumferential direction x = ατ/π at t = 0 is shown in Figure 2.7.

2.7

Radial forces

2.7.1

Production of radial magnetic forces

On the basis of Equation 2.45 the space and time distribution of the MMF of a polyphase electrical machine fed with sinusoidal and balanced current system can be expressed by the following equations:

3 ⋅ 10−4 2.5 ⋅ 10−4

Λ g (x)

2 ⋅ 10−4

1 ⋅ 10−4 5 ⋅ 10−5 0

0

0.05

0.1

0.15

0.2

0.25 x, m

0.3

0.35

0.4

0.45 0.5

Figure 2.7 Distribution of the air gap relative permeance according to Equation B.5 with the rotor saliency effect taken into account on the basis of Equation 2.106 for a PM brushless motor with s1 = 36, 2 p = 10, τ = 50 mm, b14 = 3 mm, g = 1.0 mm, e = 0.1 mm, ed = 0.1 mm.


DK3193

noise

November 3, 2005

15:28

46

Noise of Polyphase Electric Motors • for the stator F1 (α, t) =

∞

Fmν cos(νpα ∓ ωt)

(2.107)

Fmµ cos(µpα ∓ ωµ t + φ µ )

(2.108)

ν=1

• for the rotor F2 (α, t) =

∞ µ=1

where α is the angular distance from the origin of the coordinate system, p is the number of pole pairs, φµ is the angle between vectors of the stator and rotor harmonics of equal order, ν and µ are the numbers of space harmonics of the stator and rotor, respectively, ω = 2π f is the pulsation of the input current, and Fmν and Fmµ are the peak values of the νth and µth harmonics expressed by Equations 2.43 and 2.55, respectively. The product pα = π x/τ where τ is the pole pitch and x is the linear distance from a given axis results from Equation 2.30. For symmetrical polyphase stator windings and integral number of slots per pole per phase q1 , the stator space harmonics are expressed by Equation 2.47. For induction machines µ is expressed by Equation 2.23. For rotors of synchronous machines µ = 2k − 1

(2.109)

where k = 1, 2, 3, . . .. The instantaneous value of the normal component of the magnetic flux density in the air gap at a point determined by the angle α can be calculated with the aid of Equation 2.63 b(α, t) = [F1 (α, t) + F2 (α, t)]g (α, t) = b1 (α, t) + b2 (α, t)

T

(2.110)

where • for the stator b1 (α, t) = F1 (α, t)g (α, t) =

∞

Bmν cos(νpα ∓ ωt)

(2.111)

ν=1

• for the rotor b2 (α, t) = F2 (α, t)g (α, t) =

∞

Bmµ cos(µpα ∓ ωµ t + φ µ ). (2.112)

µ=1

According to the Maxwell stress tensor, the radial magnetic force per unit area or magnetic pressure waveform at any point of the air gap is pr (α, t) =

1 2 b (α, t) − bt2 (α, t) . 2µ0

(2.113)

Since the magnetic permeability of the ferromagnetic core is much higher than that of the air gap, the magnetic flux lines are practically perpendicular to the stator and


DK3193

noise

November 3, 2005

15:28


47

rotor cores. Thus, the tangential component bt (α, t ) of the magnetic flux density is much smaller than the normal component b(α, t ) and b2 (α, t ) 1 = [F1 (α, t ) + F2 (α, t )]2 2g (α, t ) 2µ0 2µ0 1 = [F 2 (α, t )2g (α, t ) + 2F1 (α, t )F2 (α, t )2g (α, t ) 2µ0 1 + F22 (α, t )2g (α, t )]

pr (α, t ) ≈

=

[b1 (α, t )]2 + 2b1 (α, t )b2 (α, t ) + [b2 (α, t )]2 . 2µ0

(2.114)

There are three groups of the infinite number of radial force waves: • the product [b1 (α, t )]2 of the stator harmonics of the same number ν, i.e., pr ν (α, t ) =

[Bm ν cos(νp α ∓ ωt )]2 B2 = m ν [1 +cos(2νp α∓2ωt )] (2.115) 2µ0 4µ0

• the mixed product 2b1 (α, t )b2 (α, t ) of the stator ν and rotor µ harmonics 2Bm ν cos(νp α ∓ ωt )Bm µ cos(µp α ∓ ωµ t + φµ ) 2µ0 1 = Bm ν Bm µ {cos[(νp α ∓ ωt ) − (µp α ∓ ωµ t + φµ )] 2µ0 + cos[(νp α ∓ ωt ) + (µp α ∓ ωµ t + φµ )]} 1 = Bm ν Bm µ {cos[ p α(ν − µ) ∓ (ω − ωµ )t − φµ ] 2µ0 + cos[ p α(ν + µ) ∓ (ω + ωµ )t + φµ ]} (2.116)

pr νµ (α, t ) =

• the product [b2 (α, t )]2 of the rotor harmonics of the same number µ, i.e., [Bm µ cos(µp α ∓ ωµ,n t + φµ,n )]2 2µ0 1 2 = B [1 + cos(2µp α ∓ 2ωµ t + 2φµ )] 4µ0 m µ

pr µ (α, t ) =

(2.117)

The following trigonometric identities cos2 α = 0.5(1 + cos 2α) and 2 cos α cos β = cos(α − β) + cos(α + β) have been used. The constant term equal to Bm2 ν /(4µ0 ) or Bm2 µ /(4µ0 ) in Equations 2.115 and 2.117 have no significance for noise generation and can be neglected because the static magnetic pressure is uniformly distributed along the air gap. In accordance with Equations 2.115, 2.116, and 2.117, the magnetic forces per unit area (Figure 2.8) can be expressed in the following general form pr (α, t) = Pmr cos(r α − ωr t)


(2.118)

DK3193

noise

November 3, 2005

15:28

48

Noise of Polyphase Electric Motors 3 ⋅ 105 2.5 ⋅ 105

Pr (x)

2 ⋅ 105 1.5 ⋅ 105 1 ⋅ 105 5 ⋅ 104 0

0

0.05

0.1

0.15

0.2

0.25 x, m

0.3

0.35

0.4

0.45

0.5

Figure 2.8 Distribution of the radial magnetic pressure along the air gap periphery at no load for s1 = 36, 2 p = 10, τ = 50 mm, g = 1.0 mm and eccentricity e = 0.1 mm. where Pmr is the amplitude of the magnetic pressure, ωr is the angular frequency and and r = 0, 1, 2, 3, . . . are corresponding orders of radial magnetic forces. The radial forces circulate around the stator bore with the angular speed ωr /r and frequency fr = ωr /(2π ). For small number of the stator pole pairs the radial forces may cause the stator to vibrate and produce acoustic noise. For PM brushless motors the full load armature reaction field is normally less than 20% of the open-circuit magnetic field [254]. Thus, the effect of the first and second term in Equation 2.114 on the acoustic noise at no load is usually lower than that of the third term, i.e., in most cases harmonics in the open-circuit magnetic field have a dominant role in the noise of electromagnetic origin [254].

2.7.2

Amplitude of magnetic pressure

The amplitude Pmr of the radial magnetic pressure of the r -th order in Equation 2.118 depends on the harmonics of the magnetic flux density participating in its production, i.e., • excited by the stator harmonics of the same number ν 2 Bmν N/m2 4µ0 • excited by the stator ν and rotor µ harmonics

Pmr =

Bmν Bmµ N/m2 2µ0 • excited by the rotor harmonics of the same number µ Pmr =

Pmr =


2 Bmµ

4µ0

N/m2 .

(2.119)

(2.120)

(2.121)

DK3193

noise

November 3, 2005

15:28


49

To obtain the amplitude of the radial force, the force pressure amplitude Pmr should be multiplied by π D1in L i where D1in is the stator core inner diameter and L i is the effective length of the stator core.

2.7.3

Deformation of the stator core

By only considering the pure circumferential vibration modes of the stator core, the deflection d of the stator core is an inverse function of the fourth power of the force order r [206] 1 d ∝ 4 (2.122) r Considering the product of the stator harmonics of the same number according to Equation 2.115, the lowest frequency of the magnetic force fr = 2 f varies at twice the line frequency and the order is r = 2 p. Since the force order r = 2 for a two-pole motor and r = 4 for a four-pole motor, the magnetic force in a four-pole motor will be 1/16 of that of the two-pole motor. The twice line frequency noise is generally significant only in two-pole motors because the stator deflection varies with the force order according to Equation 2.122. The largest deformation of the stator yoke ring occurs when the frequency fr is close to the natural mechanical frequency of the stator system. The most important from airborne noise point of view are low circumferential mode numbers, i.e., r = 0, 1, 2, 3 and 4. Vibration mode r = 0. For pulsating vibration mode or “breathing” mode r = 0 the radial magnetic force density p0 = Pmr =0 cos ω0 t (2.123) is distributed uniformly around the stator periphery and changes periodically with time. It causes a radial vibration of the stator core and can be compared to a cylindrical vessel with a variable internal overpressure [87]. Equation 2.123 describes an interference of two magnetic flux density waves of equal lengths (the same number of pole pairs) and different velocity (frequency). According to [154], the “breathing” mode r = 0 can cause acoustically harmful low wave number vibration, even with a high number of poles and sinusoidal stator current. This can happen when nearby low wave numbers are given a high admittance by structural discontinuities such as supports [154]. Vibration mode r = 1. For “beam bending” mode r = 1 [58] the radial pressure p1 = Pmr =1 cos(α − ω1 t)

(2.124)

produces a single-sided magnetic pull on the rotor. The angular velocity of the pull rotation is ω1 . Physically, Equation 2.124 describes an interference of two magnetic flux density waves for which the number of pole pairs differs by one.


DK3193

noise

November 3, 2005

15:28

50

Noise of Polyphase Electric Motors r=0

r=1

P

r=2

P

P

r=3

P

P r=4 P

P

P P P

P P

P P

P

P P

P

P

Figure 2.9 Deformation of the core caused by the space distribution of radial forces. Vibration modes r = 2, 3, 4. For “ovaling mode” r = 2 [58] and r = 3, 4, . . . wave-shaped deflections of the stator core will occur. Figure 2.9 shows space distribution of forces producing vibrations of the order of r = 0, 1, 2, 3, 4.

2.7.4

Frequencies and orders of magnetic pressure

Angular frequencies and orders of radial magnetic pressure result from Equations 2.115, 2.116, and 2.117, i.e.: • excited by the stator harmonics of the same number ν, i.e., ωr = 2ω;

fr = 2 f ;

r = 2νp

(2.125)

• excited by the stator ν and rotor µ harmonics ωr = ω ± ωµ ;

fr = f ± f µ ;

r = (ν ± µ) p (2.126)

• excited by the rotor harmonics of the same number µ, i.e., ωr = 2ωµ ;


fr = 2 f µ ;

r = 2µp

(2.127)

DK3193

noise

November 3, 2005

15:28


51

For the fundamental space harmonic ν = 1 the frequency of the radial magnetic pressure according to Equation 2.125 is fr = 2 f and the order (circumferential vibrational mode) r = 2 p. The rotating component with 2 p pole pairs and frequency 2 f can cause significant vibration, especially, in large machines, e.g., for Bm ν=1 = 0.8 T the magnitude of the radial magnetic pressure is 127.4 kN/m2 . For higher space harmonics ν > 1 the frequency of the radial force pressure is also fr = 2 f and the order r = 2νp. For salient pole synchronous machines the numbers of higher space harmonics for the stator and rotor are ν = 2km 1 ± 1 and µ = 1 ± 2k, respectively, where k = 0, 1, 2, 3, . . . and m 1 is the number of phases. The vibration order according to Equation 2.126 is r = p[2km 1 ±1 ±(1 ±2k )] or r = 2 p[k (m 1 +1)+1] and r = 2 p[k (m 1 +1)−1], e.g., for k = 0 the vibration order is r = 2 p, for k = 1 and m 1 = 3 the order r = 10 p and r = 6 p, for k = 2 and m 1 = 3 the order r = 18 p and r = 14 p, etc. The frequency of vibration is fr = f ± µf = f [1 ± 1 ± 2k] or fr = 2 f (1 +k ) and fr = 2 f k. The frequency of the rotor space harmonics is f (1 ±2k ). The rotor of a salient pole synchronous machine without any cage winding can be considered as a rotor with the number of slots s0 = 2 p. Thus, the angular frequency of the rotor space harmonics and rotor angular speed = 2 p f / p is f ωµ = ω ± s0 k = ω ± 2 pk 2π = 2π f (1 ± 2k ) (2.128) p The vibrational mode of the rotor harmonics of the same number µ according to Equation 2.127 is r = 2µp = 2(1 ± 2k ) p and frequency fr = 2 f µ . For a salient pole synchronous machine fr = 2 f (1 ± 2k ) and for induction machine fr = 2 f [1 ± k(s2 / p)(1 − s)] where µ = 1 ± 2k, k = 0, 1, 2, 3, . . . and s is the slip (Equation 2.2) for fundamental. Those harmonics of low orders produce significant radial forces, in PM synchronous motors also at zero stator current state. The frequencies of radial forces due to rotor space harmonics of the same number µ for PM synchronous motors are given in Table 2.3.

2.7.5

Radial forces in synchronous machines with slotted stator

Significant vibration and noise can be produced by interaction of the PM rotor field and slotted structure of the stator. The fictitious MMF wave excited by PMs is expressed by Equation 2.108. The PM rotor magnetic flux density waveform modulated by stator slots is a product of the MMF F2 (α, t) and relative air gap permeance λsl (α) due to slot opening varying with the angle α. According to Equations 2.68 and 2.71 sl (α) = −g0

∞

Ak cos(ks1 α)

(2.129)

k=1,2,3 2 where −g0 Ak = 2µ0 γ1 kok /t1 . By multiplying Equations 2.108 and 2.129 the magnetic flux density component produced as a result of the modulation of the


DK3193

noise

November 3, 2005

15:28

52


Table 2.3 Frequencies fr = 2 f (1 ± 2k ) and orders r = 2(1 ± 2k ) p of forces produced by the rotor magnetic field of synchronous motors. k 0 1 2 3 4 5 6 7 8 9 10

Frequency of radial force fr 2f −2 f and 6 f −6 f and 10 f −10 f and 14 f −14 f and 18 f −18 f and 22 f −22 f and 26 f −26 f and 30 f −30 f and 34 f −34 f and 38 f −38 f and 42 f

Order of radial force r 2p −2 p and 6 p −6 p and 10 p −10 p and 14 p −14 p and 18 p −18 p and 22 p −22 p and 26 p −26 p and 30 p −30 p and 34 p −34 p and 38 p −38 p and 42 p

rotor flux density by the stator slot openings is bλ (α, t ) = −g0

∞

Fm µ cos(µp α ∓ ωµ t + φµ )

µ=1,3,5

=

∞

Ak cos(ks1 α)

k =1,2,3

∞ ∞ 1 g0 Fm µ Ak cos[(µp ± ks1 )α ∓ ωµ t + φµ )]. (2.130) 2 µ=1,3,5 k =1,2,3

In the above Equation 2.130 the trigonometric identity cos α cos β = 0.5[cos(α + β) + cos(α − β)] has been used. The radial force pressure on the stator due to the rotor µth and stator permeance kth harmonic is 1 {0.5g0 Fm µ Ak cos[(µp ± ks1 )α ∓ ωµ t + φµ ]}2 2µ0 1 2 2 2 = F A {1 + cos[2(µp ± ks1 )α ∓ 2ωµ t + 2φµ ]}. (2.131) 8µ0 g0 m µ k

pr λ,µ,k (α, t ) =

Similar to the cogging torque, radial forces according to Equation 2.131 will be present both at zero stator current state and when the motor is fed with current. The order (circumferential mode) of the force pressure pr λ (α, t) is rλ = 2(µp ± ks1 ) and frequency fr λ = 2µf . The most important are low even orders rλ = 2, 4, 6, and 8 for the fundamental stator permeance harmonic k = 1, i.e., µ=

|0.5rλ ∓ s1 | = 1, 3, 5, . . . p

and

fr λ = 2µf.

(2.132)

Table 2.4 shows values of rλ , µ, and fr λ obtained on the basis of Equation 2.132 for selected numbers of pole pairs p and stator slots s1 . The most dangerous case


DK3193

noise

November 3, 2005

15:28


53

Table 2.4 Orders and frequencies of radial forces due to interaction of the rotor field and stator slotted core in three-phase synchronous and PM motors (Equation 2.132). p

rλ

2

4

2

8

3

6

4

8

5 5

2 8

s1

q1

µ

fr λ = 2µf

12 18 24 36 12 18 24 36 18 24 36 24 36 36 36

1 1.5 2 3 1 1.5 2 3 1 1.333 2 1 1.5 1.2 1.2

5 and 7 8 and 10 11 and 13 17 and 19 4 and 8 7 and 11 10 and 14 16 and 20 5 and 7 7 and 9 11 and 13 5 and 7 8 and 10 7 8

10 f and 14 f 16 f and 20 f 22 f and 26 f 34 f and 38 f 8 f and 16 f 14 f and 22 f 20 f and 28 f 32 f and 40 f 10 f and 14 f 14 f and 18 f 22 f and 26 f 10 f and 14 f 16 f and 20 f 14 f 16 f

is for p = 5 and s1 = 36 because the order (circumferential mode) is low, i.e. rλ = 2. This mode should be avoided in the design of PM brushless motors. For example, for 660 rpm, p = 5, s1 = 36 the input frequency is f = 55 Hz and fr λ = 14 × 55 = 770 Hz at rλ = 2. If the number of pole pairs is reduced to p = 4, the input frequency is f = 44 Hz and the frequency of radial forces are fr λ = 16 × 44 = 704 Hz and fr λ = 20 × 44 = 880 Hz at rλ = 8. Reduction of number of poles from 5 to 4 and keeping the same number of slots s1 = 36 increases the order of radial magnetic forces. The higher the order of radial magnetic forces, the lower the acoustic noise of magnetic origin. Similar considerations are given by Walker and Kerruish for large synchronous machines operating at no load [220]. According to [220] the frequency of predominant noise in synchronous machine is fr λ = 2µλ f where µλ is the integer nearest to s1 /(2 p). This frequency is produced by stator harmonic permeance k = 1 and rotor MMF harmonic µ = µλ . The analysis of Equation 2.131 shows that 2µp−2ks1 is least when µ−ks1 / p is least (not in agreement with [220]). For k = 1 the expression µ − s1 / p is least when µ is as close to s1 / p as possible, i.e., µ is the integer nearest to s1 / p. Considering a machine with 2 p = 10, s1 = 36, and f = 55 Hz, the ratio s1 / p = 36/5 = 7.2 and µ = µλ = 7. The frequency of the predominant noise is again fr λ = 2µλ f = 2 × 7 × 55 = 770 Hz and the order rλ = |2(µλ p − s1 )| = |2(7 × 5 − 36)| = 2.


DK3193

noise

November 3, 2005

15:28

54

2.7.6


Frequencies of vibration and noise

Product of stator space harmonics of the same number According to Equation 2.125, the frequency of the radial magnetic force is equal to the double the frequency of the line fundamental frequency. The order for the stator slot harmonics is r = 2νp = 2(ks1 ± p) (2.133) where ν is according to Equation 2.21. Interaction of stator and rotor harmonics For the fundamental time harmonic the field of each stator winding harmonic at a given point with the coordinate α pulsates with the angular frequency ω = 2π f , where f is the fundamental frequency. For an induction motor the angular frequency of the rotor slot harmonics is [87]

f s2 ωµ = ω ± ks2 m = ω ± ks2 2π (1 − s) = 2π f 1 ± k (1 − s) (2.134) p p where m = 2π f (1 − s)/ p is the mechanical angular speed of the rotor and k = 1, 2, 3, . . .. Thus, the angular frequencies of radial forces due to the stator and rotor slot harmonics result from Equation 2.126

f ωr = ω ± ωµ = ω ± ω ± ks2 2π (1 − s) (2.135) p The frequencies of radial forces s2 fr = f k (1 − s) p

and

s2 fr = f k (1 − s) ± 2 p

(2.136)

The orders are also determined by Equation 2.126 r = (ν ± µ) p = ks1 ± ks2 ± 2 p

(2.137)

where ν is according to Equation 2.21 and µ is according to Equation 2.23. For ν = µ = 1 the orders are r = 0 and r = 2 p. For ν = µ the orders are r = 0 and r = 2νp = 2µp. For a PM synchronous motor with salient poles or embedded magnets and without any cage winding, the higher harmonics of the rotor µ=k

so ± 1 = 2k ± 1 p

(2.138)

are called “pole harmonics” and their “winding coefficients” are equal to “winding coefficients” of the fundamental harmonic. The number of slots for a synchronous


DK3193

noise

November 3, 2005

15:28


55

motor without any cage winding so = 2 p and k = 1, 2, 3, . . .. Thus, the angular frequency of the rotor harmonics (m = = ω/ p) is ωµ = ω ± kso = ω ± 2 pk = ω ± 2 pk(2π

f ) = 2π f (1 ± 2k). p

(2.139)

The noise and vibration frequencies due to the stator winding and rotor harmonics are according to Equation 2.126, i.e., fr = f ± f (1 ± 2k) or

fr = 2(1 + k) f

and

fr = 2k f.

The corresponding orders of radial forces s1 r = (ν ± µ) p = k ± 1 ± 2k ± 1 p = ks1 ± 2 p(k + 1). p

(2.140)

(2.141)

Product of rotor space harmonics of the same number According to Equation 2.127 the angular frequency and frequency of the radial magnetic forces of an induction machine are

s2 ωr = 2ωµ = 2(ω ± ks2 m ) = 4π f 1 ± k (1 − s) (2.142) p

s2 fr = 2 f 1 ± k (1 − s) (2.143) p and the order r = 2µp = 2(ks2 ± p) (2.144) where µ is according to Equation 2.23. For synchronous motors s2 = s0 = 2 p and slip s = 0, so that Equations 2.143 and 2.144 become fr = 2(1 ± 2k) f ;

r = 2 p(1 ± 2k).

(2.145)

Radial force harmonics due to eccentricity The stator and rotor magnetic flux density waves due to eccentricity are expressed by Equations 2.95 and 2.98, harmonic numbers by Equations 2.96 and 2.99, and harmonic angular frequencies by Equations 2.85, 2.86, and 2.100. The angular frequency of the magnetic flux density wave for static eccentricity is ω = 2π f and according to Equation 2.95 for dynamic eccentricity this angular frequency is ω ± . The angular frequency of higher harmonic radial forces in an induction machine due to eccentricity for ν = 1 is calculated with the aid of Equations 2.95, 2.100, and 2.126, i.e.,

s2 r = ω ± ± ων=1 + ω k (1 − s) . p


DK3193

noise

November 3, 2005

15:28

56


The above equation for static eccentricity ( = 0) gives the following two frequencies

s2 s2 fr = f 2 + k (1 − s) and fr = f k (1 − s). (2.146) p p The basic orders of eccentricity harmonics are r =1

and

r =2

(2.147)

as there is a single-sided magnetic pull due to eccentricity (bending and ovaling modes). However, on the basis of Equations 2.96, 2.99, and 2.126, there are also higher orders of eccentricity force harmonics s2 1 1 r = (ν ± µ ) p = 1 ± ± k + 1 ± p p p p = p ± 1 ± (ks2 + p ± 1) or r = 2( p ± 1) ± ks2

and

r = ks2

(2.148)

that usually can be neglected. Vibration of zero order (r = 0) can also be produced for interference of space harmonics of the same order [10]. The angular frequency of higher harmonic radial forces in an induction machine due to dynamic eccentricity ( = 2π( f / p)(1 − s), according to Equations 2.85, 2.100, and 2.126 is

s2 1 r = ω 1 ± (1 − s) ± 1 ± k (1 − s) . p p The above equation gives the next two frequencies of radial forces due to eccentricity

s2 s2 1 1 fr = f 2 ± (1 − s) + k (1 − s) ; fr = f (1 − s) + k (1 − s) p p p p (2.149) The orders for frequencies (Equation 2.149) are determined by Equations 2.147 and 2.148. For synchronous motors the frequencies of radial forces due to eccentricity are

1 1 fr = 2 f (1 + k); fr = 2 f k; fr = f 2 (1 + k) ± ; fr = f 2k ± . p p (2.150)


DK3193

noise

November 3, 2005

15:28


57

Radial force harmonics due to stator winding parallel paths The effect of eccentricity can produce additional radial forces if the stator winding has parallel current paths. The maximum number of parallel paths is a p = 2 p. The magnitude of the magnetic flux density waveform excited by currents in parallel paths is usually not more than 10 to 15% of the magnitude of the fundamental waveform excited by the stator winding. Interaction of fundamental flux density wave of r = p and frequency f with backward rotating wave (r = p∓1, f = − f ) caused by internal short-circuit currents for > 0 gives radial magnetic force wave of r = 1 and fr = 2 f . The magnitude of this force increases with the load. To avoid noise, Heller and Hamata [87] recommend the following selection of the numbers of rotor slots of induction machines: p (2.151) − p(6c ± 1) = 1 if a p > 2 is odd s2 ± a s2 ± 2 p − p(6c ± 1) = (2.152) if a p > 2 is even 1 a |(s2 ± 2 p) − p(6c ± 1)| = 1

if

ap = 2

(2.153)

where c = 1, 2, 3, . . . is a small integer. Radial force harmonics due to magnetic saturation The saturation magnetic flux density equation is a product of the MMF for ν = 1 according to Equation 2.45 and the second term of the air gap permeance (Equation 2.102), (Equation 2.103), i.e., bsat (α, t) = F1,1 (α, t)sat (α, t) = −Fmν=1 cos( pα − ωt)sat cos(2 pα − 2ωt − 2φs ) 1 1 = − Fmν=1 sat cos(− pα + ωt + 2φs ) − Fmν=1 sat 2 2 × cos(3 pα − 3ωt − 2φs ) = −Bsat [cos( pα − ωt − 2φs ) + cos(3 pα − 3ωt − 2φs )]

(2.154)

where Bsat = 0.5Fmν=1 sat . The above Equation 2.154 describes two stator magnetic flux density saturation waves. One wave has frequency and number of poles the same as the fundamental and second wave has three times higher frequency and number of pole pairs than the fundamental. Equation 2.154 includes both the magnetic saturation of the stator teeth and core (yoke) [64, 179]. Assuming that only the harmonic ν = 3 in Equation 2.154 saturates the teeth, the flux density in the rotor due to saturation is: b2 (α, t) =

∞

Bsat,µ cos(µpα − ωµ − ψµ )

(2.155)

µ=1

where for induction motor µ=k


s2 + 3. p

(2.156)

DK3193

noise

November 3, 2005

15:28

58


According to Equation 2.134 the angular frequency of the rotor saturation harmonics

s2 ωµ = 3ω + ks2 = 3ω ± k 2π f (1 − s ) . p The angular frequency an order of radial magnetic forces due to saturation of the magnetic circuit can be found with the aid of Equation 2.126

s2 ωr sat = ω ± ωµ = ω ± 3ω ± k 2π (1 − s ) (2.157) p s1 s2 r = (ν ± µ) p = k ± 1 ± k + 3 p . p p The last two equations give the frequencies and orders of radial magnetic forces due to saturation in the following form

s2 s2 fr sat = k (1 − s ) + 4 f and fr sat = k (1 − s ) + 2 f (2.158) p p r = ks1 + ks2 + 4 p

and

r = ks1 + ks2 + 2 p .

(2.159)

Frequencies of radial magnetic forces produced by higher space harmonics in cage induction and PM synchronous motors are given in Tables 2.5 and 2.6.

2.8 2.8.1

Other sources of electromagnetic vibration and noise Unbalanced line voltage

The unbalanced line voltage causes vibration the frequency of which is equal to double the line frequency, i.e., fr unb = 2 f.

(2.160)

The effect of phase unbalance can be included by assigning different values of currents for each phase in Equation 2.42 and then calculating the resultant stator MMF similar to Equation 2.43.

2.8.2

Magnetostriction

Magnetostriction is the change of physical dimensions and crystals of a material in response to changing its magnetization. When a magnetostrictive material is subjected to an alternating magnetic field, it changes its shape and dimensions. Most ferromagnetic materials show some measurable magnetostriction. The


DK3193

noise

November 3, 2005

15:28


59

Table 2.5 Frequencies of radial magnetic forces produced by higher space harmonic ν = 1 (fundamental time harmonic n = 1) in induction motors. Source

Frequency Hz

Product of stator space haromics bν2 of fr = 2 f the same number ν Product of rotor space fr = 2 f [1 ± k(s2 / p)(1 − s)] harmonics bµ2 of the where s is slip for same number µ fundamental space harmonic Product of stator and rotor space harmonics bν bµ — general fr = f ± f µ equations Product of stator and rotor space harmonics bν bµ where fr = [k(s2 / p)(1 − s) ± 2] f ν = ks1 / p ± 1 and fr = [k(s2 / p)(1 − s)] f µ = ks2 / p ± 1 (the so called slot or step harmonics) Product of stator and rotor static eccentricity fr = [2 + k(s2 / p)(1 − s)] f space fr = [k(s2 / p)(1 − s)] f harmonics bν bµ Product of stator and rotor dynamic fr = [2 ± (1 − s)/ p eccentricity +k(s2 / p)(1 − s)] f space fr = [(1 − s)/ p harmonics +k(s2 / p)(1 − s)] f bν bµ Product of stator and rotor magnetic fr = [k(s2 / p)(1 − s) + 4] f saturation space fr = [k(s2 / p)(1 − s) + 2] f harmonics bν bµ


Order (circumferential mode) r = 2νp r = 2(ks1 ± p) k = 0, 1, 2, 3, . . . r = 2µp r = 2(ks2 ± p) r = (ν ± µ) p

r = ks1 ± ks2 ± 2 p

r =1 r =2

r =1 r =2

r = ks1 + ks2 + 4 p r = ks1 + ks2 + 2 p

DK3193

noise

November 3, 2005

15:28

60


Table 2.6 Frequencies of radial magnetic forces produced by higher space harmonic ν = 1 (fundamental time harmonic n = 1) in synchronous motors. Source

Frequency Hz

Order (circumferential mode) r = 2νp ν = 2km 1 ± 1 k = 0, 1, 2, 3, . . .

Product of stator space haromics bν2 of fr = 2 f the same number ν Interaction of the rotor fr = 2µλ f magnetic field and slotted where r = 2(µλ p ± s1 ) core of the stator µλ = integer [s1 / p] Product of rotor space haromics bµ2 of the fr = 2(1 ± 2k) f r = 2µp = 2 p(1 ± 2k) same number µ where µ = 1 ± 2k Product of stator and rotor space harmonics bν bµ — general fr = f ± f µ r = (ν ± µ) p equations Product of stator winding and rotor space harmonics fr = 2(1 + k) f bν bµ where fr = 2k f r = ks1 ± 2 p(1 + k) ν = ks1 / p ± 1 and µ = 2k ± 1 Product of stator and rotor static eccentricity fr = 2(1 + k) f r =1 space fr = 2k f r =2 harmonics bν bµ Product of stator and rotor dynamic eccentricity fr = [2(1 + k) ± 1/ p] f r =1 space fr = (2k ± 1/ p) f r =2 harmonics bν bµ Product of stator and rotor magnetic fr = 2(2 + k) f r = ks1 + 2 p(k + 2) saturation space fr = 2(1 + k) f r = ks1 + 2 p(k + 1) harmonics bν bµ


DK3193

noise

November 3, 2005

15:28


61

magnetostrictive coefficient is defined as the change in length l to the length l of a specimen when the magnetization increases from zero to its saturation value, i.e., l = (2.161) l The coefficient can be positive or negative. In the first case, the increase in the magnetic flux density causes expansion of the specimen. In the second case, the specimen shrinks as the magnetic flux density increases. For pure iron the saturation magnetostriction is = −2 × 10−5 to +2 × 10−5 , for laminated steels = −0.1 × 10−5 to +0.5 × 10−5 , for nickel = −5.2 × 10−5 to 0.8 × 10−5 and for polycrystal cobalt = −6 × 10−5 to −2.5 × 10−5 . Some alloys, e.g., iron and dysprosium DyFe2 or iron and terbium TbFe2 , display “giant” magnetostriction under relatively small fields. If the magnetostriction acts to contract a specimen, then this will act against any tensile stress on the material and leads to larger value of the Young’s modulus. The core of an electrical machine or transformer in alternating magnetic fields changes its dimensions cyclically. Figure 2.10 shows a sinusoidal variation of the magnetic flux density and corresponding variation of the magnetostrictive coefficient with time. The curve (t) shown in Figure 2.10 resolved into Fourier series has a constant component and a series of harmonic functions. Those harmonics approximately visualize vibrations relative to the mean position which is determined by the constant component. The fundamental wave of those vibration has a frequency twice the frequency of the flux density f (line frequency), i.e., f ms = 2 f

(a)

(2.162)

b(t)

t

(b)

Λ(t)

t

Figure 2.10 Variation of magnetostrictive coefficient for sinusoidal magnetic flux density: (a) magnetic flux density as a function of time; (b) magnetostrictive coefficient as a function of time.


DK3193

noise

November 3, 2005

15:28

62


while the order of the fundamental wave of the magnetostriction force in an a.c. rotating machine is rms = 2 p (2.163) For static electromechanical energy converters as transformers, inductors, fluorescent light ballast, etc., magnetostriction is the major cause of acoustic noise. For a.c. rotating machines magnetostriction can contribute considerably to the acoustic noise. Especially, in two pole machines ( p = 1) the magnetostriction force at the twice line frequency (Equation 2.162) can generate undesirable acoustic noise.

2.8.3

Thermal stress analogy

The effect of magnetostriction can be implemented in a similar way as thermal stresses are applied, i.e., the thermal expansion of the free body is calculated based upon the temperature distribution [45, 159]. The elastic strain in the x direction is el,x =

1 (σx − νσ y ) E

(2.164)

where σx is the external stress in the x direction, σ y is the external stress in the y direction, σz = 0, E is Young modulus and ν is Poisson ratio. Including thermal stresses when a material is heated x − αϑ ϑ =

1 (σx − νσ y ) E

(2.165)

where αϑ is the coefficient of thermal expansion, αϑ ϑ is the thermal strain and x = el,x + αϑ ϑ is the total strain (elastic and thermal). For a magnetostrictive material, the magnetostrictive strain λ(B) can also be added to the elastic strain [45], similar to the thermal expansion strain (Equation 2.165), i.e., 1 x − λ(B) = (σx − νσ y ) (2.166) E where the total strain (elastic and magnetostrictive) is x = el,x + λ(B). The magnetostrictive strain λ(B) = 10−6 B 2 (2.167) expressed as a function of the magnetic flux density in the direction of the vector B B is sometiems called the magnetostriction curve [45].

2.8.4

FEM model

The magnetic field is affected by the effect of magnetostriction. Thus, the coupled magnetic and mechanical system must be captured in one magnetomechanical matrix. The effect of magnetostriction can be built into the coupled system using a force distribution that is added to magnetic forces [45, 159, 210, 211].


DK3193

noise

November 3, 2005

15:28


63

The elastic energy of a mechanical finite-element is expressed in terms of the elastic displacement αel and mechanical stiffness matrix K U=

1 T α K αel . 2 el

(2.168)

In general cases the elastic displacement αel = α where α is the total displacement, because the material may show magnetostrictive or thermal expansion. Including the magnetostrictive displacement αms the elastic energy is U=

1 (α − αms )T K (α − αms ). 2

(2.169)

The strains λx , λ y in the x and y directions can be converted into nodal dise (e) placements αms = (αx,i , α ey,i ) where i = 1, 2, 3 means the element nodes with coordinates xi , yi . Considering the midpoint element (xm , ym ) as fixed

(e) αx,i (xi − xm )λx . (2.170) = (yi − ym )λ y α (e) y,i The nodal magnetostriction forces are obtained by multiplying the mechanical (e) stiffness matrix K (e) for one element by the magnetostrictive element αms of the nodes, i.e., (e) (e) Fms = K (e) αms . (2.171) The total magnetostriction force is a sum of magnetostriction forces of all elements, i.e., (e) Fms = Fms = K αms . (2.172) e

The magnetic forces (sum of reluctance forces and Lorentz forces) can be found using virtual works, i.e., A ∂M ∂W Fmag = − AT =− dA (2.173) ∂a ∂a 0 where M is the magnetic “stiffness” matrix, a is the linear displacement, and A is the magnetic vector potential component in the normal direction (z-direction). The magnetic energy 1 W = A T M A. (2.174) 2 The total energy of the magnetomechanical system is the sum of the elastic energy U according to Equation 2.168 and magnetic energy W according to Equation 2.174, i.e., E = U + W. (2.175) The magnetomechanical system is described by the matrix equations [44, 45]: • without magnetostriction

M 0 A T = (2.176) 0 K a R + Fmag


DK3193

noise

November 3, 2005

15:28

64

Noise of Polyphase Electric Motors • with feedback of magnetostriction

M 0 A T = 0 K a R + Fmag + Fms

(2.177)

where R = K a represents external forces and T = M A is the magnetic source term vector. Further procedures for deformation and vibration calculations are given in [44, 45].


DK3193

noise

October 24, 2005

19:9

3 Inverter-Fed Motors Most variable speed electric motors are fed from solid state converters (SSCs). The a.c.–a.c. SSC consists of a rectifier, intermediate circuit (filter), and inverter (Figure 3.1). The output voltage and output current of a SSC are nonsinusoidal, that is, contain higher time harmonics as a result of switching. The noise of electric machines with the stator winding nonsinusoidal current increases as compared with an equivalent machine with sinusoidal stator current. The increase in the noise of an induction motor fed from a pulse width modulation (PWM) inverter with switching frequency up to 7 kHz is from 7 to 15 dB(A) [80]. For higher switching frequencies between 7 and 16 kHz the increase in the noise is lower, usually from 2 to 7 dB(A) [80].

3.1

Generation of higher time harmonics

The stator current of an electrical machine fed from or loaded with a SSC is nonsinusoidal, that is, contains higher time harmonics n = 2m 1 k ±1 where k = 0, 1, 2, 3, . . .. The nonsinusoidal current can be expressed in terms of Fourier series as n=∞ √ n=∞ i 1 (t) = I1mn sin(ωn t − φin ) = 2 I1n sin(ωn t − φin ) (3.1) √

n=1

n=1

where I1mn = 2I1n is the peak value of the nth harmonic current. The angular frequency and frequency of higher time harmonic currents are, respectively ωn = nω = 2π n f

f n = n f.

(3.2)

Currents expressed by Equation 3.1 are shifted in each phase of the stator winding of a polyphase machine by the angle 2π/m 1 , where m 1 is the number of phases. For induction motors, the slip for the nth time harmonic is defined as sn =

nn s ∓ n s (1 − s) nn s 65


(3.3)

DK3193

noise

October 24, 2005

19:9

66

Noise of Polyphase Electric Motors Intermediate Circuit

Rectifier

Inverter VVVF

50 or

d.c.

60 Hz

IM a.c.

a.c.

Figure 3.1 Basic circuitry of a variable voltage variable frequency (VVVF) voltage source SSC for an induction motor. where nn s = n f / p is the synchronous speed for the nth time harmonic, n s (1 − s ) is the rotor speed, and s is the slip for the fundamental harmonic. The “−” sign is for the time harmonics n = 2m 1 k + 1 and the sign “+” is for the time harmonics n = 2m 1 k − 1 where k = 0, 1, 2, 3, . . .. Induction motors operate with small slip 0.02 ≤ s ≤ 0.05, so that 1 sn ≈ 1 ∓ . (3.4) n As the number of time harmonics increases, the slip (Equation 3.4) becomes close to unity. The slip sn , angular frequency ωn of rotor currents, and higher time harmonic slip ssn corresponding to the synchronous speed (Equation 2.1) are: • for forward-rotating magnetic field sn = 1 −

1 (1 − s ); n

ωn = n ωsn = ω[n − (1 − s )];

ssn = 1 − n (3.5)

• for backward-rotating magnetic field sn = 1 +

1 (1 − s ); n

ωn = n ωsn = ω[n + (1 − s )];

ssn = 1 + n (3.6)

where s is the slip for the fundamental harmonic according to Equation 2.2, ω = 2π f is the angular frequency for the fundamental harmonic, f is the input frequency of the fundamental harmonic, and n is the time harmonic number. Higher time harmonics in the stator current of permanent magnet (PM) brushless motors are discussed in Section 4.7.

3.2

Analysis of radial forces for nonsinusoidal currents

The frequencies and orders of most important radial forces produced by SSCs (inverters) are given in Table 3.1. A brief analysis of these forces follows.


DK3193

noise

October 24, 2005

19:9

Inverter-Fed Motors

67

Table 3.1 Frequencies and orders of most important radial magnetic forces produced by inverter harmonics. Frequency Hz

Source Product of stator 2 harmonics b1νn of the same number n Product of rotor harmonics of the same number Product of stator and rotor space harmonics Sums and differences of inverter fundamental and higher harmonics bn bn of different numbers Interaction of the switching frequency f sw and higher time harmonics f n = n f sw ± n f Interaction of the permeance and MMF harmonics associated with time harmonics Rectifier harmonics transmitted to the motor via intermediate circuit and inverter

3.2.1

Order (circumferential mode)

fr,n = 2n f , n = 2km 1 ± 1 where k = 0, 1, 2, 3, . . . fr,n = 2n f µ fr,n = f n ± f µ see also Equation 3.13 fr,n = f ± f n n = 2km 1 ± 1

r = 2νp r = 2 p for ν = 1 r = 2µp r = 2 p for µ = 1 r = (ν ± µ) p r =0 or r = 2p

fr,n = |(± f n ) − f | n = n

r =0

fr,n = |(± f n − f | fr,n = | f n ± µf |

r =0 r =2

fr = 2m 1 k f k = 1, 2, 3, . . .

r = 2p

Stator and rotor magnetic flux density

The space-time variation of the magnetic flux density for the n-th time harmonic can be expressed as a sum of space harmonics • for the stator b1n (α, t) =

∞

b1νn (α, t) =

ν=1


∞ ν=1

Bmνn cos(νpα − ωn t)

(3.7)

DK3193

noise

October 24, 2005

19:9

68

Noise of Polyphase Electric Motors where b1νn (α, t ) = Bm νn cos(νp α − ωn t ) • for the rotor b2n (α, t ) =

∞

b1µn (α, t ) =

µ=1

∞

Bm µn cos(µp α − ωµ,n t + φµ,n )

(3.8)

µ=1

where b1µn (α, t ) = Bm µn cos(µp α − ωµ,n + φµ,n ). In the above Equations 3.7 and 3.8, ν is the number of the stator space harmonics, µ is the number of the rotor space harmonics, n is the number of the stator time harmonics, Bm νn and Bm µn are the peak values of the magnetic flux density harmonics for the stator and rotor, respectively, α is the angular distance from the origin of the coordinate system, ωn = 2π n f is the angular frequency of the time harmonic current in stator windings, f is the fundamental frequency, ωµ,n is the angular frequency of the space harmonic µ in the rotor system for the given time harmonic n, φµ,n is the angle between vectors of the stator and rotor space harmonics for the given n, and µ0 = 0.4π × 10−6 H/m is the magnetic permeability of free space. Higher space harmonics result from the distribution of stator winding coils in slots, rotor winding slots, or geometry of rotor poles. Higher time harmonics in the stator current are generated by the power supply, that is, utility mains or SSC.

3.2.2

Stator harmonics of the same number

The radial force pressure Figure 2.8 is expressed with the aid of Maxwell stress tensor — Equation 2.114. The first term in Equation 2.114 is the product of the stator harmonics [b1νn (α, t)]2 of the same number ν or n, i.e., [b1νn (α, t)]2 [Bmνn cos(νpα ∓ ωn t]2 = 2µ0 2µ0 2 B = mνn [1 + cos(2νpα ∓ 2ωn t)]. 4µ0

pr,n (α, t) =

(3.9)

For higher time harmonics n > 1 the angular frequency of stator space harmonics is ωn = 2π f n , where f n = n f is the frequency of the n-th higher time harmonic. The constant term for each harmonic ν and n has no significance for noise generation as the magnitudes Bmνn are uniformly distributed along the air gap. For the fundamental harmonic ν = 1 and n = 1 the frequency of the radial force pressure is fr = 2 f and the order (circumferential vibrational mode) r = 2 p. For higher space harmonics ν > 1 and fundamental time harmonic n = 1 the frequency of the radial force pressure is also fr = 2 f and the mode r = 2νp. For ν = 1 and n > 1 the frequency of the radial force pressure is fr,n = 2n f and the order is r = 2 p. Of course, there will be many more harmonics of radial forces as mixed products of stator magnetic flux density harmonics of different numbers, for example, 1 and 7, also produce radial forces.


DK3193

noise

October 24, 2005

19:9

Inverter-Fed Motors

3.2.3

69

Interaction of stator and rotor harmonics

The second term in Equation 2.114 is the mixed product of the stator and rotor harmonics 2b1νn (α, t)b2µn (α, t), i.e., pr,n (α, t) = = =

2b1νn (α, t)b1µn (α, t) 2µ0 ∞ 2 ν=1 Bmνn cos(νpα ∓ ωn t) ∞ µ=1 Bmµn cos(µpα ∓ ωµ,n t + φµ,n ) 2µ0 1 2µ0

∞ ∞

Bmνn Bmµn {cos[(νpα ∓ ωn t) − (µpα ∓ ωµ,n t + φµ,n ]

ν=1 µ=1

+ cos[(νpα ∓ ωn t) + (µpα ∓ ωµ,n t + φµ,n ]} ∞ ∞ 1 = Bmνn Bmµn {cos[(ν − µ) pα ∓ (ωn − ωµ,n )t − φµ,n ] 2µ0 ν=1 µ=1 + cos[(ν + µ) pα ∓ (ωn + ωµ,n )t + φµ,n ]}.

(3.10)

The mixed products of the stator and rotor harmonics generate pairs of radial force pressures with pole pairs r = (ν ± µ) p and frequencies fr,n = f n ± f µ,n . For a synchronous machine the angular frequency of the rotor space harmonics for n > 1 is ωµ,n = 2π f n (1 ± k). For an induction machine with the number of rotor slots s2 , the angular frequency of the rotor winding harmonics for time harmonics n ≥ 1 and a “harmonic rotor” angular speed mn = 2π( f n / p)(1 − sn ) is fn ωµ,n = ωn + ks2 mn = ωn ± s2 k 2π (1 − sn ) p s2 = 2π f n 1 ± k (1 − sn ) . (3.11) p According to Equations 3.3 and 3.4, for n = 1 the slip sn=1 = s ≈ 0 and for n > 1 the slip sn ≈ 1. The frequencies of noise and vibration of induction machines due to the mixed products of b1νn (α, t)b2µn (α, t) for µ = ks2 / p ± 1 are s2 fr,n = f n ± f µ,n = f n ± f n 1 ± k (1 − sn ) . p The last equation can also be written as s2 fr,n = f n 1 ± 1 ± k (1 − sn ) p s2 s2 fr,n = f n k (1 − sn ) fr,n = f n k (1 − sn ) ± 2 . p p

(3.12)

(3.13)

The orders of the mixed product harmonics are expressed by Equation 2.137.


DK3193

noise

October 24, 2005

19:9

70


3.2.4

Rotor harmonics of the same number

The third term in Equation 2.114 is the product of the rotor harmonics [b2µn (α, t)2 ] of the same number [b2µn (α, t)]2 [Bmµn cos(µpα ∓ ωµ,n t + φµ,n )]2 = 2µ0 2µ0 1 = Bmµn [1 + cos(2µpα ∓ 2ωµ,n t + 2φµ,n )]. 4µ0

pr,n (α, t) =

3.2.5

(3.14)

Frequencies and orders of magnetic pressure for nonsinusoidal currents

Angular frequencies and orders of the radial magnetic pressure due to higher time and space harmonics result from Equations 3.9, 3.10, and 3.14, i.e.: • excited by the stator harmonics of the same number ν, i.e., ωr,n = 2nω;

fr,n = 2n f ;

r = 2νp

(3.15)

• excited by the stator ν and rotor µ harmonics ωr,n = (nω ±ωµ,n );

fr,n = (n f ± f µ,n );

• excited by the rotor harmonics of the same number µ, i.e., ωr,n = 2nωµ ;

fr,n = 2n f µ ;

r = (ν ±µ) p (3.16)

r = 2µp.

(3.17)

According to Equation 3.15, for the fundamental space harmonic ν = 1 and time harmonics n > 1, the frequency of the radial magnetic pressure is fr,n = 2n f and the order r = 2 p. According to Equation 3.16, for ν = 1 and time harmonics n > 1, the frequency of the radial magnetic pressure is fr,n = n( f ± f µ ) and the order r = (1 ± µ) p. For induction machines the frequencies of magnetic pressures as a function of slip sn are expressed by Equation 3.11. Putting s2 = so = 2 p and sn = 0 for a synchronous machine, the frequency becomes fr,n = f n (1±kso / p) = n f (1 ± 2k). According to Equation 3.17, for space harmonics µ > 1 and time harmonics n > 1, the frequency of the radial magnetic pressure is fr,n = 2n f µ and the order r = 2µp. For a salient pole synchronous machine fr,n = 2n f (1 ± 2k) and for induction machine fr,n = 2n f [1 ± k(s2 / p)(1 − sn )] since according to Equation 2.143 fr = 2 f [1 ± ks2 (1 − s)/ p].

3.2.6

Interaction of stator harmonics of different numbers

Inverters can supply a rich spectrum of higher time harmonics to the stator winding. Higher time harmonics of different numbers can produce significant radial forces, the frequency and order of which can be expressed by the following equations fr,n = (n ± n ) f ;


r =0

or

r = 2p

(3.18)

DK3193

noise

October 24, 2005

19:9

Inverter-Fed Motors

71

where n = n . Most important are magnetic forces due to sums and differences of the fundamental harmonic f with major higher time harmonics of the stator current [71]. The frequencies of those forces are fr,n = (1 ± n) f where n = 2km 1 ± 1.

3.2.7

Interaction of switching frequency and higher time harmonics

Large amplitude forces are also due to interaction of the switching frequency f sw and higher time harmonics f n = n f sw ± n f [138]. If n is an odd integer, n will be an even integer and vice versa, i.e., f n = f sw ± 2 f , f sw ± 4 f , f sw ± 6 f, . . . and f n = 2 f sw ± f , 2 f sw ± 3 f , 4 f sw ± 5 f, . . .. The most important case is the interaction between the fundamental field harmonic and higher time harmonics for which the frequencies and mode of vibration are [138]: fr,n = |(± f n ) − f |;

3.2.8

r =0

(3.19)

Interaction of permeance and magnetomotive force (MMF) harmonics

Interaction of the permeance field harmonics and MMF harmonics associated with higher time harmonics of the stator current can also result in important vibration, especially at full load and when force orders are low [138]. The frequencies and orders of radial forces are [138] s2 fr,n = |(± f n − f |; r = 0, 2. fr,n = | f n ± f 1 + k (1 − s) |; p (3.20)

3.2.9

Rectifier harmonics

Rectifier harmonics are transmitted via the intermediate circuit and inverter to the stator winding. They can produce radial forces with the following frequency and order fr,n = n f = 2m 1 k f ; r = 2p (3.21) where n = 2m 1 k, k = 1, 2, 3, . . ., and m 1 is the number of phases.

3.3 3.3.1

Higher time harmonic torques in induction machines Asynchronous torques

Asynchronous torques due to higher time harmonics are expressed by the following equations m1 (3.22) Tn = √ pN2 kw2 n I2n cos ψ2n = cT n I2n cos ψ2n 2


DK3193

noise

October 24, 2005

19:9

72


or

2 (I2n ) R2n (I )2 R2n Pw2n = m 1 p 2n =p (3.23) s n n sn ωn sn ωn √ where cT = m 1 pN2 kw2 / 2, N2 is the number of rotor turns per phase, kw2 is the rotor winding factor for fundamental, n is the stator magnetic flux produced by the nth time harmonic, ψ2n is the phase shift between the electromotive force (EMF) (excited by n ) and rotor harmonic current I2n , I2n is the rotor current due to the nth time harmonic referred to the stator system, R2n is the rotor resistance referred to the stator system, Pw2n are the rotor winding losses for the nth time harmonic, and 2π n f ωn n = = (3.24) p p

Tn = m 1

is the angular speed of the stator magnetic field due to the nth time harmonic. The direction of the torque (Equation 3.22) and (Equation 3.23) depends on the number of the higher time harmonic. The torque can be in the direction of rotation (propulsion torque) or in the opposite direction (braking torque). For a three-phase machine (m 1 = 3) higher time harmonics n = 1, 7, 13, . . . generate asynchronous torques in the direction of rotation and higher harmonics n = 5, 11, 17, . . . generate braking torques. For evaluation of torques, the electric losses in the stator winding can be assumed equal to the electric losses in the rotor winding, i.e. [21], Pw2n ≈ 0.5Pwn ≈ 0.5(SC R)2

Pw Pws=1 ≈ 0.5 3 n n3

(3.25)

where SC R = 3 . . . 7 is the starting current ratio (starting–to–rated current), Pwn are stator and rotor winding losses due tho the nth time harmonic, Pw are stator and rotor winding losses at rated load, Pws=1 are winding losses at locked rotor (s = 1), and Pw Pwn = (SC R)2 3 . (3.26) n Assuming that for higher time harmonics sn ≈ 1 and n = n, the asynchronous torque is inversely proportional to n 4 , i.e., Tn = p

Pws=1 Tst Pw2n Pws=1 ≈ ≈ 4 ≈p 3 4 s n ωn 2n × 1 × nω 2n n

(3.27)

where Tst ≈ 0.5Pws=1 / is the starting torque due to the fundamental harmonic. Since higher harmonic asynchronous torques are inversely proportional to n 4 , they are negligible and have little effect on the vibration and noise.

3.3.2

Pulsating torques

Pulsating torques are produced as interaction of currents and fluxes of different frequencies. Their frequencies are significantly higher than the fundamental frequency and their mean value is equal to zero. The number of these torques can


DK3193

noise

October 24, 2005

19:9

Inverter-Fed Motors

73

be extremely large as there is a very large number of interactions among all harmonics. The most important pulsating torques are those produced by the stator fundamental magnetic flux and rotor current higher harmonics expressed by the following equation: 2 pm 1 T p1,n = √ N2 kw2 I2n cos[(n ± 1)]ωt − ψ1,2n ] 2 = 2cT I2n cos[(n ± 1)]ωt − ψ1,2n ]

(3.28)

where is the magnetic flux due to the fundamental harmonic n = 1 and ψ1,2n is the phase shift between the EMF for n = 1 and rotor current I2n . The “−” sign is for harmonics n = 2m 1 k + 1 and the sign “+” is for the harmonics n = 2m 1 k−1. It means that in a three-phase machine the pulsating torques due to the 5th and 7th harmonics have the same frequency 6 f and rotate in opposite directions, i.e., f − (−5 f ) = 6 f and f − 7 f = −6 f . In three-phase induction motors the major torque ripple source is the 6th harmonic component. For the 11th and 13th harmonics the frequency of pulsating torques is 12 f , i.e., f − (−11 f ) = 12 f and f − 13 f = −12 f . The magnitudes of pulsating torques at SC R = 3 . . . 5 are 7 to 12% of the rated torque for the frequency 6 f (due to the 5th and 7th harmonics) and 0.8 to 1.5% for the frequency 12 f (due to the 11th and 13th harmonics [21]. Pulsating torques may have a signifucant influence on the vibration and noise.

3.4

Higher time harmonic torques in permanent magnet (PM) brushless machines

The interaction of the stator current and magnetic flux linkage harmonics produce: • constant torques if harmonics are of the same number; • pulsating torques if harmonics are of different numbers. For trapezoidal EMF and rectangular current the electromagnetic torque is constant (no pulsating torque) [117, 177]. All odd harmonics of the magnetic flux that interact with the stator current harmonics of the same number (except of triple harmonics) produce a constant torque, for example, fundamental harmonic of flux with the fundamental harmonic of current, 5th harmonic of flux and 5th harmonic of current, 7th harmonic of flux and 7th harmonic of current, and so on. However, the contribution of the odd harmonics n > 1 to the constant torque is negligible.

3.5

Influence of the switching frequency of an inverter

There is a significant influence of the switching frequency of a pulse width modulation (PWM) inverter on the overall motor noise. The switching frequency of the inverter is established by the frequency of a triangular waveform. In a PWM


DK3193

noise

October 24, 2005

19:9

74

Noise of Polyphase Electric Motors 80

75

Sound Pressure Level, dB(A)

70

65

60

55 Switching Frequency 1.8 kHz Switching Frequency 3.6 kHz Switching Frequency 7.2 kHz Switching Frequency 14.4 kHz

50

45

40 0

10

20

30

40 50 60 Frequency, Hz

70

80

90

100

Figure 3.2 Influence of the inverter on the sound pressure level of a 9.3 kW (12.5 hp), 60 Hz, 4-pole induction motor for four different switching frequencies [161]. inverter the sinusoidal control signal at the desired frequency is compared with a triangular waveform [156]. At speeds below the rated speed, the noise level of a small motor with a shaft driven fan decreases with the speed (Chapter 7, Equation 7.28). Below the rated speed the magnetic noise is predominant. When the speed exceeds the rated speed, the overall noise increases due to the increased noise of the fan. Figure 3.2 shows the overall noise of a 9.3 kW (12.5 hp), 4-pole induction motor fed from a PWM inverter for four different switching frequencies [161]. For 1.8-kHz switching frequency, the influence of the magnetic noise is considerable and the sound pressure level changes only ±4 dB(A) within the operation frequency range from 10 to 100 Hz. For 3.6, 7.2, and 144 kHz switching frequencies, the magnetic noise and consequently the overall noise is reduced significantly for operation frequencies below the rated frequency of 60 Hz [161]. Table 3.2 compares the sound pressure level radiated by small power induction motors fed with 60 Hz three-phase utility grid and PWM inverters [161]. It is clear that below the rated frequency 60 Hz the sound pressure level decreases as the switching frequency is increased. This is because the noise of magnetic origin


DK3193

noise

October 24, 2005

19:9

Inverter-Fed Motors

75

80

Noise Level (dB)

70

4 kHz Carrier

60 50 40

16 kHz Carrier

30 20 10 0

0

10

20

30

40

50

60

Motor Frequency (Hz)

Figure 3.3 Noise level vs. induction motor operation frequency for 4 and 16 kHz carrier frequencies of X2C series programmable inverters. (Courtesy of TB Wood’s, Chambersburg, PA, U.S.A. [244].) is predominant at low operation frequencies. The overall noise at low speeds is strongly affected by the switching frequency (compare the SPL at 30 and 60 Hz). At operation frequencies above the rated frequency 60 Hz, the influence of the cooling fan on the overall noise is very strong. The increased sound power level at 100 Hz is due to the fan and is practically independent of the switching frequency. The switching frequencies can cause pure tones that are very disturbing to the human ear [147]. The vibration frequncies due to fundamental waves and switching frequencies must not match the natural frequencies of the stator system. Otherwise, the audible noise increases significantly [147]. Modern inverters operate at higher switching frequencies ( f s ≥ 12 kHz), which results in lower harmonic content, especially at low speeds. The inverter losses increase with increasing the switching frequency and motor higher harmonic losses decrease with increasing the switching frequency [147]. Thus, at low speeds the influence of the switching frequency on the efficiency of the motor–inverter system is rather small [147]. Many inverters, for example, the X2C series TB Wood’s inverters have programmable carrier frequencies from 4 to 16 kHz for quiet drive operation [244]. Low carrier frequencies can generate audible motor noise, while at 16 kHz audible noise is virtually eliminated (see Figure 3.3).

3.6

Noise reduction of inverter-fed motors

Inverter-fed induction and PM synchronous motors are noisier than those fed with purely sinusoidal current. The radial magnetic force spectrum is richer and the chance of matching the exciting frequencies with natural frequencies of the stator


DK3193

noise

October 24, 2005

19:9

76


Table 3.2 Influence of inverter on sound pressure level of induction motors, dB(A) [161] Output Utility power kW

0.37 2p = 4 0.75 2p = 2 0.75 2p = 4 2.2 2p = 2 2.2 2p = 4 3.7 2p = 4 5.5 2p = 4 9.3 2p = 4 15.0 2p = 4

grid

Switching frequency of inverter 3.6 kHz

7.2 kHz

14.4 kHz

60 Hz Operation frequency Operation frequency Operation frequency Hz Hz Hz 30 60 100 30 60 100 30 60 100 46.1 52.1 50.5

58.9

42.2 49.3

58.9

37.1 46.8

58.8

61.2 61.2 68.2

73.1

45.9 60.5

72.7

44.7 60.4

72.8

44.7 60.3 52.9

58.8

47.0 57.1

59.5

34.3 46.8

58.5

67.5 60.6 68.5

80.8

54.9 68.4

80.8

53.3 68.2

81.1

53.1 68.7 60.6

63.8

56.6 65.4

64.7

46.6 52.9

63.4

54.4 56.3 58.7

66.8

53.7 65.6

67.0

44.3 54.8

66.5

56.3 61.8 59.6

68.1

52.9 63.6

68.2

43.0 56.5

68.0

61.7 65.1 71.1

73.7

57.2 68.1

73.4

46.7 61.9

73.4

69.2 68.7 73.4

80.9

54.8 69.8

80.8

52.5 69.3

80.7

system is greatly increased. The following strategies can be employed to reduce the acoustic noise of inverter-fed motors; • elimination of higher time harmonics; • operation with low commutation angle (the larger the commutation angle, the higher the magnitude of the torque pulsation); • dynamic variation of the switching frequency [147]; • high switching frequency (above 15 kHz) is a very efficient method but imposes high stress on solid switches and increased switching frequency losses; • random PWM [36, 82, 229]; • active noise control of the motor.


DK3193

noise

October 24, 2005

19:9

4 Torque Pulsations In addition to radial forces, torque pulsations and tangential forces also contribute to vibrations and magnetic noise. Normally, tangential forces have less effect on vibrations and noise than radial forces. Tangential forces can either slightly magnify or reduce the amplitude of the first order radial forces [87]. In this chapter, calculations of torque ripple and tangential forces with special attention given to cogging torque of permanent magnet (PM) synchronous motors will be discussed.

4.1

Analytical methods of instantaneous torque calculation

The electromagnetic torque of electrical machines is calculated analytically as T =−

∂W ∂

J

(4.1)

where W is the magnetic field energy in the air gap and is the mechanical angle measured around the air gap periphery. Equation 4.1 can be written as [87] T =−

gL i ∂ 2µ0 ∂

2π

[B(α, t)]2 dα.

(4.2)

0

Neglecting the magnetic saturation B(α, t) =

µ0 µ0 F(α, t) = [F1 (α, t) + F2 (α, t)] g g

(4.3)

where F1 (α, t) is the stator magnetomotive force (MMF) and F2 (α, t) is the rotor MMF. Thus, 2π µ0 L i ∂ T =− [F1 (α, t) + F2 (α, t)]2 dα. (4.4) 2g ∂ 0 77 Copyright © 2006 Taylor & Francis Group, LLC

DK3193

noise

October 24, 2005

19:9

78


For a stationary stator and coordinate system fixed to the stator, only the rotor MMF is dependent on the rotor position [87], i.e., µ0 L i T =− g

2π

[F1 (α, t) + F2 (α, t)]

0

∂F2 (α, t) dα. ∂

(4.5)

Harmonic torques are calculated on the basis of Equation 4.5 putting F1 and F2 in form of the Fourier series. A torque can be produced only by harmonics having an equal number of pole pairs [87].

4.2

Numerical methods of instantaneous torque calculation

The instantaneous torque can be numerically calculated by using • virtual work method

∂w(, θ ) ∂w (i, θ) T =− = ∂θ ∂θ =const i=const

(4.6)

where w, w , , i and θ are the magnetic energy, coenergy, flux linkage vector, current vector and mechanical angle, respectively. • Maxwell stress tensor 1 1 2 F = B( B · n) − B n d S; T = r × F (4.7) µ0 2µ0 T = Li

1 µ0

r Bn Bt dl

(4.8)

l

where n, L i , l, r , Bn and Bt are the normal vector to the surface S, stack length, integration contour, radius, radial (normal) component of the magnetic flux density and tangential component of the magnetic flux density, respectively. • Lorentz force theorem π/(2m 1 p) T = i j (t) 2 pN r L i B(θ, t)dθ j=A,B,C

=

−π/(2m 1 p)

1 [e f A (t)i 1A (t) + e f B (t)i 1B (t) + e f C (t)i 1C (t)] m

(4.9)

where p, θ, m and N are the number of pole pairs, mechanical degree, mechanical angular speed and conductor number in phase belt, respectively. The products e f A (t)i 1A (t), e f B (t)i 1B (t), and e f C (t)i 1C (t) represent electromagnetic power per phase.


DK3193

noise

October 24, 2005

19:9

Torque Pulsations

4.3

79

Electromagnetic torque components

The instantaneous torque of an electrical motor T (α) = T0 + Tr (α)

(4.10)

has two components (Figure 4.1), i.e.: • constant or average component T0 ; • periodic component Tr (α), which is a function of time or electrical angle α, superimposed on the constant component. The periodic component causes the torque pulsation also called the torque ripple. There are many definitions of the torque ripple. The torque ripple is caused by both the construction of the machine and power supply. The torque ripple can be defined in one of the following ways: Tmax − Tmin Tmax + Tmin Tmax − Tmin tr = Tav Tmax − Tmin tr = Tr ms [torque ripple]r ms Trr ms tr = = Tav Tav

tr =

(4.11) (4.12) (4.13) (4.14)

where the average torques in Equations 4.12 and 4.14 α+T p Tp 1 1 Tav = T (α)dα = T (α)dα Tp α Tp 0

T Tr T0

Time Tp

Figure 4.1 Constant and periodic components of the torque.


(4.15)

DK3193

noise

October 24, 2005

19:9

80


and the r ms or effective torque in Equation 4.13 Tp 1 Tr ms = T 2 (α)dα. Tp 0

(4.16)

The r ms torque ripple Trr ms in Equation 4.14 is calculated according to Equation 4.16 in which T (α) is replaced by Tr (α). The time T p in Equations 4.15 and 4.16 is the period of the torque waveform. For sinusoidal waveform the half-cycle √average value is (2/π )Tm , where Tm is the peak torque and the r ms value is Tm / 2. For the waveform containing higher harmonics, the r ms value of the torque ripple is

Trr ms = Trr2 ms1 + Trr2 ms2 + . . . + Trr2 msν . (4.17)

4.4

Sources of torque pulsations

There are three sources of the torque ripple coming from an electrical machine: (a) cogging effect (detent effect), that is, interaction between the rotor magnetic flux and variable permeance of the air gap due to the stator slot opening geometry; (b) distortion of sinusoidal or trapezoidal distribution of the magnetic flux density in the air gap; (c) the difference between permeances of the air gap in the d and q axis. The cogging effect produces the so-called cogging torque, higher harmonics of the magnetic flux density in the air gap produce the field harmonic electromagnetic torque, and the unequal permeance in the d and q axis produces the reluctance torque. The causes of torque pulsation coming from the supply are: (d) current ripple resulting, for example, from PWM; (e) phase current commutation.

4.5

Higher harmonic torques of induction motors

The theory of higher harmonic torques of induction machines due to stator and rotor slots have been developed by Görges (1896), Arnold (1909), Punga (1912), Dreyfus (1924), Krondl (1926), Lund (1932), Heller (1947), Alger (1954), Jordan (1962), Oberretel (1965), and others [87]. Harmonic torques of induction motors will not be considered in this book. Any induction motor may be imagined as a series of mechanically coupled asynchronous and synchronous motors having a different number of poles [87]. An asynchronous torque is created if a certain stator MMF harmonic of the order νp produces in the spectrum of the rotor MMFs a harmonic of the same order


DK3193

noise

October 24, 2005

19:9

Torque Pulsations

81

νp = µp [87]. To reduce asynchronous torques due to higher space harmonics, the recommended numbers of slots should be selected according to the following unequalities: s2 ≤ 1.25s1 (4.18) s2 < s1

(4.19)

|s2 − s1 ± p| ≥ 4

(4.20)

where s1 is the number of stator slots and s2 is the number of rotor slots. A synchronous torque is produced if the spectra of the stator and rotor MMF harmonics contain harmonics of the same order νp and if the rotor harmonic of this order is produced by a stator harmonic of another order νp = µp. To reduce synchronous torque the following number of slots should be avoided: • to reduce synchronous torques at standstill s2 = 2m 1 pk

k = 1, 2, 3, . . .

where

(4.21)

s2 = s1 − 2m 1 p

(4.22)

s2 = s1 − 4 p

(4.23)

• to reduce synchronous torques in the motoring region s < 1 s2 = 2m 1 pk + 2 p

(4.24)

• to reduce synchronous torques in the braking region s < 1 s2 = 2m 1 pk − 2 p.

4.6

(4.25)

Cogging torque in permanent magnet (PM) brushless motors

The frequency of the cogging torque is f c = s1 n s = s1

f p

(4.26)

where s1 is the number of stator slots, n s = f / p is the synchronous speed, p is the number of pole pairs and f is the input frequency. There are three sources of the torque ripple coming from the construction of the magnetic and electric circuit of a machine (Section 4.4). The most comprehensive literature review and comparison of methods of the torque ripple reduction is given in [26]. Analytical methods of cogging torque calculation usually neglect the magnetic flux in the stator slots and magnetic saturation of the stator teeth [1, 16, 18, 30, 46, 76, 83, 253, 256]. Cogging torque is derived from the magnetic flux density distribution either by calculating the rate of change of total energy stored in the air gap with respect to the rotor angular


DK3193

noise

October 24, 2005

19:9

82


position [1, 16, 18, 46, 76, 83] or by summing the lateral magnetic forces along the sides of the stator teeth [253]. Many cogging torque calculations are specific to a particular method of reducing the cogging effect [48]. The proposed anlytical method uses the energy of the magnetic field in the air gap and classical equations for the air gap magnetic flux density distribution in a.c. machines with slotted stators [40, 51, 87, 236].

4.6.1

Air gap magnetic flux density

The magnetic flux density distribution in the air gap of a PM machine can be analyzed in a similar way as that of an induction machine with slotted stator and smooth rotor [40, 87]. The normal component of the magnetic flux density in the air gap is a sum of the mean flux density bmean (x ) and periodical component with the period equal to the slot pitch t1 , i.e.: 2π b f (x ) = bmean (x ) + (4.27) Bk cos k t1 k where [40, 87] Bg . (4.28) kC Carter coefficient kC is expressed using the well known Equation 2.72 [87]. Considering the air gap magnetic field distribution only in the interval of one pole pitch τ , the flux density B0 = Bg where Bg is the flat-topped value of the magnetic flux density in the air gap. Fourier series coefficient Bk in Equation 4.27 depends on the approximation of the magnetic flux density distribution over the slot opening. For skewed slots Bk can take the following form (see Equation 2.71) bmean (x ) =

Bk = −2γ1

g 2 2 B0 kok ksk t1

(4.29)

where kok is the coefficient of the slot opening according to Equation 2.64 and ksk is the skew factor. Thus, Equation 4.27 can be brought to the form ∞ 1 2π g 2 2 b f (x ) = B0 − 2γ1 k k cos k x . (4.30) kC t1 k =1 ok sk t1 In order to include the influence of the finite width of PMs or pole shoes on the cogging effect, the flux density B0 must be expressed as a periodic function of the pole pitch τ , i.e., B0 = b P M (x), where b P M (x) is the distribution of the PM magnetic flux density normal component in a machine without slots along the x-axis. Thus, the normal component of the magnetic flux density distribution in the air gap (Figure 4.2) of a PM brushless motor with slotted stator core can simply be expressed as: b P M (x) b f (x) = + bsl (x). (4.31) kC


DK3193

noise

October 24, 2005

19:9

Torque Pulsations 1.0

bf (x)

83

1 0.8 0.6 0.4 0.2 0 −0.2 −0.4 −0.6

−1.0

−0.8 −1

0 0

0.1

0.2

0.3

0.4

x

0.5 2⋅p⋅τ

Figure 4.2 Distribution of the normal component b f (x) of the magnetic flux density in the air gap (Tesla vs. meter) of a 10-pole, 36-slot PM brushless machine according to Equations 4.31, 4.32, and 4.33. The x coordinate is in the direction of rotation. The magnetic flux density component excited by the rotor PMs is a sum of higher space harmonics µ, i.e.,

π b P M (x) = (4.32) Bg bµ ksµ cos µ x τ µ and the magnetic flux density component due to stator slots is approximated as

bsl (x) = −2γ1

∞

π 2π g 2 2 kok ksk cos k x Bg bµ ksµ cos µ x . t1 k=1 t1 τ µ

(4.33)

Equation 4.33 describes the distribution of the normal component of the magnetic flux density excited by PMs (slotless machine) for B0 = b P M (x). For periodic variation of the PM flux density ∞

π 1 g 2 2 2π b f (x) = Bg bµ ksµ cos µ x − 2γ k k cos k x . τ kC t1 k=1 ok sk t1 µ (4.34) The coefficient of Fourier series bµ for higher space harmonics µ = 2m 1 k ± 1 in Equation 4.34 depends on the shape of the normal component of the magnetic flux density distribution in the air gap. It can be expressed by Equation 2.105 [73]. The pole width b p –to–pole pitch τ ratio for PM brushless motors αi = b p /τ = 0.65 to 0.92. The parameter α in Equation 2.105 describes the shape of the magnetic flux density waveform under the PM or pole shoe. For a constant (flat-topped) value of the magnetic flux density in the range −0.5b p ≤ x ≤ 0.5b p as shown in


DK3193

noise

October 24, 2005

19:9

84


Figure 4.2 the parameter α = 0. For the magnetic flux density curve changing underneath the pole according to hyperbolic cosine law (concave curve) the parameter 0 < α ≤ 1. Equation 4.34 results from the classical theory of a.c. electrical machines, e.g., [40, 51, 87]. In the above Equations 4.31, 4.32, 4.33, and 4.34, kC is the Carter coefficient of the air gap taking into account slot openings, Bg is the flat-topped value of the periodical magnetic flux density waveform excited in the air gap by the rotor PMs, k = 1, 2, 3, . . ., µ = 2m 1 k ±1 are the rotor higher space harmonics, ksµ is the rotor PM skew factor, τ is the pole pitch, γ1 is the parameter depending on the slot opening b14 and the air gap g, t1 is the stator slot pitch, k0k is the stator slot opening factor and ksk is the stator slot skew factor. In general, both PMs and stator slots can be skewed. The skew factors are defined by the following equations: • the rotor PM skew factor ksµ

sin µπ b f s /(2τ ) = µπ b f s /(2τ )

(4.35)

• the stator skew factor for k = 1, 2, 3, . . . ksk =

sin [kπ pbs /(s1 t1 )] sin [kπ bs /(2τ )] = . kπ bs /(2τ ) kπ pbs /(s1 t1 )

(4.36)

The stator slot opening factor is defined by Equation 2.64 [87]. The skew of PMs is b f s and the skew of stator slots is bs . For most PM configurations the air gap g is to be replaced by an equivalent air gap g ≈ g + h M /µrr ec where g is the mechanical clearance, h M is the radial height of the PM (one pole), and µrr ec = 1.02 to 1.1 is the relative recoil permeability of the PM material [75].

4.6.2

Calculation of cogging torque

If the magnetic saturation and armature reaction are negligible, the cogging torque is independent of the stator current. The frequency of the fundamental component of the cogging torque is f c = s1 n s , where s1 is the number of the stator slots and n s is the rotor speed in rev/s. Assuming that the total energy W of the magnetic field is stored in the air gap, the cogging torque is expressed as Tc (x) = −

D2out d W dW =− d 2 dx

(4.37)

where D2out ≈ D1in is the rotor outer diameter, D1in is the stator inner diameter and the x axis is in the direction of rotation. The mechanical angle =


2τ . D2out

(4.38)

DK3193

noise

October 24, 2005

19:9

Torque Pulsations

85

The rate of change of the air gap coenergy is Li g W (x ) = 2µ0

b2f (x )d x

(4.39)

where L i is the effective length of the stator stack, µ0 is the magnetic permeability of free space, g is the air gap, and b f (x ) is the air gap magnetic flux density distribution according to Equation 4.31. Substituting Equations 4.31 and 4.39 into Equation 4.37 and assuming that the maximum energy change is in the interval X + a ≤ x ≤ X + b, the cogging torque equation becomes L i g D2out ∂ Tc (X ) = − 2µ0 2 ∂ x

X +b X +a

2

b P M (x ) + bsl (x ) kC

d x.

(4.40)

According to [46], the maximum energy change occurs for a = 0.5t1 and b = 0.5b14 + ct where ct = t1 − b14 . Taking into account only fundamental space harmonics µ = 1 and k = 1, the expression in the square bracket under the integral will have the following form

2

b P M (x ) + bsl (x ) kC

1 = B cos(βx ) − AB cos(αx ) cos(βx ) kC

2

where g 2 A = 2γ1 kok k2 ; t1 =1 sk=1 α=

2π ; t1

B = Bg bµ=1 ks µ=1

(4.41)

π . τ

(4.42)

β=

With a stationary stator, only the magnetic flux density excited by the rotor PMs depends on the rotor position with respect to the coordinate system fixed to the stator [87]. Magnetic flux density waveforms are visualized in Figure 4.3. It is easier to take the integral over b f (x)2 assuming that the rotor is stationary and the stator moves with synchronous speed, that is, only stator slots expressed by the term A cos αx change their position. Thus, L i g D2out X +b 1 Tc (X ) = − 2 B cos(βx) − AB cos(αx) cos(βx) . 2µ0 2 kC X +a ×(−ABα sin αx cos βx)d x (4.43) After performing integration with respect to x, the cogging torque equation takes the following form: Tc (X ) = −

L i g D2out [I1 (X ) + I2 (X )] 2µ0 2


(4.44)

DK3193

noise

October 24, 2005

19:9

86

Noise of Polyphase Electric Motors Stator

bf (x) N

bsl(x) x

S

ns

Rotor bPM(x)

bPM (x)

bf (x)

x

S

N

Rotor

ns

Figure 4.3 Magnetic flux density waveforms in the air gap: b P M (x) excited by the rotor PMs, bsl (x) due to stator slots and b f (x) resultant. Only b P M (x) moves with the rotor.

where 2α − AB 2 sin(αx) cos2 (βx) d x kC X +a 2 a+b AB a−b 2 sin α X + = sin α kC 2 2 α a+b + sin (α + 2β) X + α + 2β 2 a+b × sin (α + 2β) 2 α a+b + sin (α − 2β) X + α − 2β 2 a−b × sin (α − 2β) 2

I1 (X ) =

X +b


(4.45)

DK3193

noise

October 24, 2005

19:9

Torque Pulsations

87

2α A2 B 2 sin(αx ) cos(αx ) cos2 (bx ) d x X +a a+b 2 1 = AB sin 2α X + sin[α(a − b)] 2 2 α a+b + sin 2(α + β) X + × sin[(α + β)(a − b)] 4(α + β) 2 α a+b + sin 2(α − β) X + × sin[(α − β)(a − b)] . 4(α − β) 2 (4.46)

I2 (X ) =

X +b

Since I1 (X ) >> I2 (X ), the following shorter form of cogging torque equation can be used L i g D2out Tc (X ) ≈ − I1 (X ). (4.47) 2µ0 2 Figure 4.4 clearly shows that the fundamental frequency of the cogging torque (first integral) is f c = s1 n s , i.e., there are s1 pulses per one revolution. The fundamental waveform is modulated by another waveform with frequency f cp = 2 pn s , i.e., 2 p pulses per revolution. The second integral indicates that there is negligible frequency f c2 = 2s1 n s , i.e., 2s1 pulses per revolution.

4.6.3

Simplified cogging torque equation

Putting b P M (x) = Bg in Equation 4.40, the cogging torque will become a waveform with predominant frequency f c = s1 n s and constant amplitude. Such a simplified equation, which does not take into account the finite width of rotor poles, has been derived in earlier publications by the first author [75,76]. For b P M (x) = Bg , µ = 1 and k = 1 the simplified cogging torque equation has the following form 2 X +b Bg L i g D2out d − ABg cos(αx) d x 2µ0 2 d X X +a kC X +b 2 d Bg 2 2 2 2 2 − ABg cos(αx) + A Bg cos (αx) d x d X X +a kC kC

Tc (X ) = − =−

L i g D2out 2µ0 2

=−

L i g D2out [K 1 (X ) + K 2 (X )] 2µ0 2

(4.48)

where 4 a+b a−b 2 K 1 (X ) = − ABg sin α X + sin α kC 2 2 a+b K 2 (X ) = A2 Bg2 sin 2α X + sin[α(a − b)]. 2


(4.49)

(4.50)

DK3193

noise

October 24, 2005

19:9

88

Noise of Polyphase Electric Motors 0.2 0.16 0.12 0.08 I1(x) 0.04 0 I2(x) −0.04 −0.08 −0.12 −0.16 −0.2 −0.2 0.2

0 0

0.1

0.2

0.3

0.4

0.3

0.4

x

0.5 2⋅p⋅τ

(a) 0.005 0.005 0.004 0.003 0.002 0.001 0 I2(x) −0.001 −0.002 −0.003 −0.004 −0.005 −0.005

0 0

0.1

0.2 x

0.5 2⋅p⋅τ

(b)

Figure 4.4 Comparison of the first and second term in Equation 4.44: (a) integrals I1 (X ) and I2 (X ) according to Equation 4.45 and 4.46; (b) integral I2 (X ). Calculation results for a PM brushless motor with embedded magnets, s1 = 36 stator slots, 2 p = 10 poles, g = 1 mm, L i = 198 mm, t1 = 14 mm, τ = 50.3 mm, b14 = 3 mm, b f s = 0, bs = 14 mm, Bg = 0.76 T, n s = 11 rev/s. Since K 1 (X ) >> K 2 (X ) (Figure 4.5), Equation 4.48 can also be written in a shorter form: L i g D2out Tc (X ) ≈ − K 1 (X ) (4.51) 2µ0 2

4.6.4

Influence of eccentricity

The variation of the air gap for static eccentricity can be expressed analytically as [87] π g(x) = g 1 − cos x (4.52) pτ where g is the mean air gap, = e/g is the relative static eccentricity and e is the displacement between the stator and rotor axis. See also Equation 2.83 in which


DK3193

noise

October 24, 2005

19:9

Torque Pulsations 0.1 0.08 0.06 0.04 K 1(x) 0.02 0 K 2(x) −0.02 −0.04 −0.06 −0.08 −0.1 −0.1

89

0.1

0 0

0.1

0.2

0.3

0.4

0.5 2⋅p⋅τ

0.3

0.4

0.5 2⋅p⋅τ

x (a)

0.005 0.004 0.003 0.002 0.001 0 K 2 (x) −0.001 −0.002 −0.003 −0.004 −0.005 −0.005 0.005

0 0

0.1

0.2 x (b)

Figure 4.5 Comparison of the first and second term in Equation 4.48: (a) integrals K 1 (X ) and K 2 (X ) according to Equations 4.49 and 4.50; (b) integral K 2 (X ). Simulation has been done for a PM brushless motor with the same parameters as specified in Figure 4.4 caption. α is according to Equation 2.30. Since the air gap is now the function of the x coordinate, the magnetic energy change depends on the variation of the air gap with the x coordinate. It is difficult to analyze Equation 4.40 with the air gap g(x) in the square bracket (second term) dependent on the x coordinate. It is more convenient to use the simplified Equation 4.48 in which A = 2γ1

g(x) 2 k ksk=1 . t1 0k=1

(4.53)

Thus, Tc (X ) = −

L i g D2out [K 1 (x) + K 2 (X ) + K 3 (X ) + K 4 (X ) + K 5 (X )] 2µ0 2 ≈−

L i g D2out [K 1 (X ) + K 3 (X )] 2µ0 2


(4.54)

DK3193

noise

October 24, 2005

19:9

90


where K 1 (X ) and K 2 (X ) are according to Equations 4.49 and 4.50, respectively, and the remaining terms are K 3 (X ) = 2

1 a+b a+b sin ( + α) ABg2 sin ( + α) x + kC 2 2

a+b a−b + sin ( − α) x + sin ( − α) 2 2

(4.55)

a−b a+b K 4 (X ) = −A2 Bg2 2 sin x + sin 2 2 a+b a−b + sin ( + 2α) x + sin ( + 2α) 2 2

(4.56)

a+b a−b + sin ( − 2α) x + sin ( − 2α) 2 2 K 5 (X ) = −A2 Bg2 2

a+b 1 sin 2( + α) x + sin[( + α)(a − b)] 4 2

1 a+b + sin 2( − α) x + sin[( − α)(a − b)] 4 2 1 a+b + sin 2γ1 x + sin[γ1 (a − b)] 2 2

(4.57)

1 a+b + sin 2α x + sin[α(a − b)] . 2 2 Figure 4.6 shows that K 3 (X ) >> K 4 (X ) and K 3 (X ) >> K 5 (X ). Including the finite width of poles the cogging torque equation eventually be Tc (X ) = −

L i g D2out [I1 (x) + I2 (X ) + K 3 (X ) + K 4 (X ) + K 5 (X )] 2µ0 2 ≈−

L i g D2out [I1 (X ) + K 3 (X )]. 2µ0 2


(4.58)

DK3193

noise

October 24, 2005

19:9

Torque Pulsations

91

0.002 2 ⋅ 10−3 K 3 (x) 0.001 K 4 (x) K 5 (x)

0 −0.001

−2 ⋅ 10−3 −0.002

0 0

0.1

0.2

0.3

0.4

x

0.5 2⋅p⋅τ

(a) 1 ⋅10−4 1 ⋅10−4

5 ⋅10−5

K 4 (x)

0 −5 ⋅10−5

−1 ⋅10−4

−1 ⋅10−4

0 0

0.1

0.2

0.3

0.4

0.3

0.4

x

0.5 2⋅p⋅τ

(b) 1 ⋅10−6 1 ⋅10−6

K 5 (x)

5 ⋅10−7 0 −5 ⋅10−7

−1 ⋅10−6 −1 ⋅10−6

0 0

0.1

0.2 x

0.5 2⋅p⋅τ

(c)

Figure 4.6 Comparison of the third, fourth, and fifth terms in Equation 4.54: (a) integrals K 3 (X ), K 4 (X ) and K 5 (X ) according to Equations 4.55, 4.56, and 4.57; (b) integral K 4 (X ); (c) integral K 5 (X ). Simulation has been done for a PM brushless motor with the same parameters as specified in Figure 4.4 caption and relative eccentricity = 0.2.


DK3193

noise

October 24, 2005

19:9

92


4.6.5

Calculation and comparison with measurements

The results of analytical calculation of the cogging torque according to Equation 4.44 which includes the finite width of the rotor magnet are shown in Figure 4.7a while the results of calculations according to the simplified Equation 4.48 are shown in Figure 4.7b. The calculated and measured cogging torque produced by the same motor 2 p = 10, s1 = 36 and the same air gap magnetic flux density Bg = 0.76 T is shown in Figure 4.8. To capture the cogging torque pulsations, a torque transducer, oscilloscope and data acquisition system have been used. It can be seen that the measured cogging torque waveform is modulated by a sinusoid

1

1 0.8 0.6 0.4 0.2

T c (x)

0 −0.2 −0.4 −0.6 −0.8 −1 −1

0 0

0.1

0.2

0.3

0.4

0.5 2 ⋅p ⋅ τ

0.3

0.4

0.5 2 ⋅p ⋅ τ

x (a)

1.0

T c (x)

1 0.8 0.6 0.4 0.2 0 −0.2 −0.4

−0.6 −0.8 −1.0 −1

0

0.1

0

0.2 x (b)

Figure 4.7 Cogging torque calculation results (Nm vs. meter): (a) according to Equation 4.44; (b) according to Equation 4.48. Design data of the PM brushless motor are the same as specified in Figure 4.4 caption, relative eccentricity ≈ 0.2.


DK3193

noise

October 24, 2005

19:9

Torque Pulsations 1.2

93

Calculations Measurements

1 0.8 0.6

Torque, Nm

0.4 0.2 0 −0.2 −0.4 −0.6 −0.8 −1 −1.2

0

0.1

0.3

0.2

0.4

0.5

x. m

Figure 4.8 Comparison of calculated cogging torque according to Equation 4.58 and measured cogging torque for one full rotor revolution. Design data of the PM brushless motor are the same as specified in Figure 4.4 caption, relative eccentricity ≈ 0.2.

with the period equal to one revolution 2 pτ , ie.: π K (x) = −M sin x pτ

(4.59)

This effect is due to the misalignment of the rotor and transducer shafts. The small shaft misalignment is difficult to avoid. The constant M in Equation 4.59 is proportional to the degree of misalignment. In comparison with the finite element method (FEM), the presented analytical method gives immediate results and allows for fast analysis of different magnetic circuit geometries of PM brushless motors. It can easily be implemented in the design procedure of commercial PM brushless motors and new products. Equation 4.44 can capture the effect of the finite width of PMs while the simplified Equation 4.48 can only predict the fundamental frequency of the cogging torque and its approximate amplitude. Equation 4.48 can only be recommended for 2-pole machines with large pole shoe–to– pole pitch ratio αi ≥ 0.9. The effect of eccentricity is included by K 3 (x), K 4 (x) and K 5 (x) and can be predicted with the aid of Equation 4.54. For the detailed study of the cogging torque the best results can be obtained using Equation 4.58.


DK3193

noise

October 24, 2005

19:9

94


There is a certain discrepancy between the analytical prediction and measurements caused by the misalignment of the rotor and torque transducer shafts. On the other hand, cogging torque measurements are difficult and the accuracy of measurements is always questionable.

4.7

Torque ripple due to distortion of EMF and current waveforms in PM brushless motors

Torque ripple due to shape and distortion of the electromotive force (EMF) and current waveforms is called the commutation torque. Assuming no rotor currents (no damper, very high resistivity of magnets and pole faces) and the same stator phase resistances, the Kirchhoff’s voltage equation for a three-phase machine can be expressed in the following matrix form: 

    v 1A i 1A R1 0 0  v 1B  =  0 R1 0   i 1B  0 0 R1 v 1C i 1C      i L L L e d  A B A C A   1A   f A  L B A L B LC B i 1B + e f B . + dt L i 1C efC C A LC B LC

(4.60)

For inductances independent of the rotor angular position the self inductances L A = L B = L C = L and mutual inductances between phases L AB = L C A = L C B = M are equal. For no neutral wire i a A + i a B + i aC = 0 and Mi a A = −Mi a B − Mi aC . Hence,     i 1A v 1A R1 0 0  v 1B  =  0 R1 0   i 1B  v 1C i 1C 0 0 R1 



     L1 − M 0 0 i 1A ef A  d  i 1B  +  e f B  . 0 L1 − M 0 + efC 0 0 L 1 − M dt i 1C

(4.61)

The instantaneous electromagnetic torque is the same as that obtained from Lorentz Equation 4.9, i.e., Td =

1 [e f A i 1A + e f B i 1B + e f C i 1C ]. 2π n

(4.62)

) At any time instant, only two phases conduct. For example, if e f A = E (tr f , (sq) (sq) ) , i 1B = −I1 and i 1C = 0, the instantaneous e f B = −E (tr f , e f C = 0, i 1A = I1


DK3193

noise

October 24, 2005

19:9

Torque Pulsations

95

electromagnetic torque according to Equation 4.62 is [28, 177] (sq)

Td =

) 2E (tr f I1

(4.63)

2π n

(sq)

) where E (tr are flat-top values of trapezoidal EMF and square wave current f and I1 (Figure 4.9). For constant values of EMF and currents, the torque (Equation 4.63) does not contain any pulsation [177]. Since e f = ωψ f = (2π n/ p)ψ f where ψ f is the flux linkage per phase produced by the excitation system, the instantaneous torque (Equation 4.62) becomes

Td = p(ψ f A i 1A + ψ f B i 1B + ψ f C i 1C )

(4.64)

To obtain torque for periodic waves of the EMF and current, it is necessary to express those waveforms in forms of the Fourier series. The magnetic flux on the basis of Equation 4.32 wc f (t) = L i b P M (x) cos(µωt)d x 0

=

4 τ Li Bg 2 π S

∞

1 π sin(µS)ksµ [1 − cos(µ w c )] cos(µωt) 3 µ τ µ=1,3,5,...

(4.65)

where w c is the coil pitch, S is according to Figure 4.9a and ksµ is according to Equation 4.35. The EMF induced in the phase A is calculated on the basis of the

Ef (tr)

e fA S

el. degrees 0

60

120 180

240

300

360

(a) i aA S1

I1(sq) el. degrees

0 60 S2

120

180 240

300

360

(b)

Figure 4.9 Idealized (solid lines) and practical (dash lines) waveforms of: (a) phase EMF; (b) phase current.


DK3193

noise

October 24, 2005

19:9

96


electromagnetic induction law e f A (t ) = −N1

=

d [ f (t )] dt

∞

π 8 τ Li 1 Bg f N1 1 − cos µ w c sin(µωt ). (4.66) sin(µS )k s µ π S µ2 τ µ=1,3,5,...

The EMFs e f B (t ) and e f C (t ) induced in the remaining phase windings are shifted by 2π/3 and −2π/3, respectively. For phase A, the current rectangular waveform expressed as Fourier series is i 1A (t ) =

∞ 4 (sq ) 1 I1 cos(ν S1 ) sin(νωt ). π ν ν=1,3,5,...

(4.67)

Harmonics of the magnetic flux and current of the same order produce constant torque. In other words, the constant torque is produced by all harmonics 2m 1 k ± 1 where k = 0, 1, 2, 3, . . . of the magnetic flux and armature currents. Harmonics of the magnetic flux and current of different order produce pulsating torque. However, when the 1200 trapezoidal flux density waveform interacts with 1200 rectangular current, only a steady torque is produced with no torque pulsations [177, 178]. In practice, the stator current waveform is distorted and differs from the rectangular shape. It can be approximated by the following trapezoidal function i 1A (t ) =

∞ 1 4Ia(sq ) [sin(ν S1 ) − sin(ν S2 )] sin(νωt ) π(S1 − S2 ) ν=1,3,5,... ν 2

(4.68)

where S1 − S2 is the commutation angle in radians. The EMF wave also differs from 1200 trapezoidal functions. In practical motors, the conduction angle is from 1000 to 1500 , depending on the construction. Deviations of both current and EMF waveforms from ideal functions result in producing torque pulsations [28, 177]. Peak values of individual harmonics can be calculated on the basis of Equations 4.66, 4.67, and 4.68. Using the realistic shapes of waveforms of EMFs and currents as those in practical motors, Equation 4.68 takes into account all components of the torque ripple except the cogging (detent) component. Figure 4.10 shows trapezoidal EMF, trapezoidal current and electromagnetic torque waveforms, while Figure 4.11 shows the same waveforms in the case of sinusoidal EMFs and currents. The torque component due to phase commutation currents Tcom = Td − Tav where Tav is the average electromagnetic torque. The frequency of torque ripple due to phase commutation is f com = 2lm 1 f where l = 1, 2, 3, . . .. For sinusoidal EMF and current waveforms the electromagnetic torque is constant and does not contain any ripple. Torques shown in Figure 4.11 are not ideally smooth due to a slight current unbalance.


DK3193

noise

October 24, 2005

19:9

Torque Pulsations

97

300 200 efA(ωt) ef B(ωt)

0

efC(ωt) −200 −300

0 0

5

10

15

20

25

30 2⋅π⋅p

20.94

26.18

31.42 2⋅π ⋅p

4

5

6 2⋅π

ωt (a)

25 25 12.5 i1A(ωt) i1B(ωt)

0

i1C(ωt) −12.5 −25

−25

0 0

5.24

10.47

15.71 ωt (b)

150 150 100 Td(ωt) Tcom(ωt)

50 0 −50

−100 −100

0 0

1

2

3 ωt (c)

Figure 4.10 Electromagnetic torque produced by a medium power PM brushless motor (m 1 = 3, 2 p = 10, z 1 = 36): (a) trapezoidal EMF waveforms; (b) trapezoidal current waveforms; (c) resultant electromagnetic torque Td and its commutation component Tcom . The current commutation angle is 50 .


DK3193

noise

October 24, 2005

19:9

98

Noise of Polyphase Electric Motors 300 300 200 efA(ωt) 100 ef B(ωt)

0

efC(ωt)

−100 −200

−300

−300

0 0

5

10

15 ωt

20

25

30 2 ⋅ π ⋅p

(a) 25

i1A(ωt) i1B(ωt)

25 20 15 10 5 0

i1C(ωt) −5 −10 −15 −20 −25 −25

0 0

5

10

15 ωt

20

25

4

5

30 2⋅π⋅p

(b) 150 150 100

Telm(ωt) Tcom(ωt)

50 0 −50

−100 −100

0 0

1

2

3 ωt

6 2⋅π

(c)

Figure 4.11 Electromagnetic torque produced by a medium power PM brushless motor (m 1 = 3, 2 p = 10, z 1 = 36): (a) sinusoidal EMF waveforms; (b) sinusoidal current waveforms; (c) resultant electromagnetic torque Td and its commutation component Tcom .


DK3193

noise

October 24, 2005

19:9

Torque Pulsations

99

4.8 Tangential forces vs. radial forces For an accurate analysis of magnetic forces in electrical machines it is necessary to include both radial and tangential forces. The radial magnetic pressure obtained on the basis of Maxwell stress tensor, is expressed by Equation 2.114. The tangential component of the magnetic pressure can also be obtained with the aid of Maxwell stress tensor, i.e., pt (α, t ) =

1 b(α, t )bt (α, t ) = b(α, t )a (α, t ) µ0

N/m2

(4.69)

where b(α, t ) is the normal component of the magnetic flux density waveform according to Equation 2.110 or Equation B.6, bt (α, t ) is the tangential componet of the magnetic flux density distribution and a (α, t ) is the line current density. The radial and tangential components of the magnetic pressure for an 8-pole PM brushless motor are plotted in Figures 4.12 and 4.13. Every harmonic of the magnetic flux density in the air gap corresponds to a certain harmonic of the line current density. As a result of interaction of rotating harmonics of the magnetic flux density in the air gap and harmonics of the line current density, rotating tangential forces varying in the space and time are produced. Those forces can also be found on the basis of Biot-Savart law. The following boundary equalities for tangential components of the magnetic field can be written • at the stator core–air gap boundary bt1 (α, t ) = ±µ0 a1 (α, t )

(4.70)

Radial Magnetic Pressure, Pa

600000 500000 400000 300000 200000 100000 0 −100000

0

100

200 300 400 500 Circumference, mm

600

700

Figure 4.12 Distribution of the radial magnetic pressure along the air gap circumference for a PM brushless motor with 2 p = 8 and s1 = 36 as shown in Figure 2.1b as obtained from the 2D FEM simulation.


DK3193

noise

October 24, 2005

19:9

100


Tangential Magnetic Pressure, Pa

300000 200000 100000 0 −100000 −200000 −300000

0

100

200 300 400 500 Circumference, mm

600

700

Figure 4.13 Distribution of the tangential magnetic pressure along the air gap circumference for a PM brushless motor with 2 p = 8 and s1 = 36 as shown in Figure 2.1b as obtained from the 2D FEM simulation. • at the rotor core–air gap boundary bt2 (α, t) = ±µ0 a2 (α, t)

(4.71)

where the line current density waveforms can be obtained by differentiating the MMF waveforms • for the stator

or

√

m 1 2 N1 kw1ν dF1ν (x, t) π a1ν (x, t) = =− I1 sin ωt ∓ ν x dx π pτ τ

π = Am1ν sin ωt ∓ ν x (4.72) τ a1 (α, t) =

∞

Am1ν sin(νpα ± ωt)

(4.73)

ν=1

• for the rotor of an induction motor √

m 2 2 N2 kw2µ dF2µ (x, t) π a2µ (x, t) = =− I2 sin ωµ t ∓ µ x + φµ dx π pτ τ

π (4.74) = Am2µ sin ωt ∓ ν x τ or ∞ a2 (α, t) = Am2µ sin(µpα ∓ ωµ t + φµ ). (4.75) µ=1


DK3193

noise

October 24, 2005

19:9

Torque Pulsations

101

For the rotor of a PM synchronous motor it is recommended to use the tangential component bt2 (α, t) of the rotor magnetic flux density instead of the line current density a2 (α, t). Every harmonic of the tangential magnetic force, similar to the radial force, has a certain frequency fr and an order r which describes its distribution around the stator periphery. The tangential magnetic pressure due to the stator and rotor magnetic flux density components is obtained by putting Equations 4.70 and 4.71 into Equation 4.69. Three groups of the tangential pressure can be identified as a result of interaction: • harmonics of the stator magnetic flux density and stator line current density ptν (α, t) = b1ν (α, t)a1ν (α, t)

(4.76)

• harmonics of the stator line current density and rotor magnetic flux density ptνµ (α, t) = b2µ (α, t)a1ν (α, t)

(4.77)

• harmonics of the rotor magnetic flux density and rotor line current density ptµ (α, t) = b2µ (α, t)a2µ (α, t).

(4.78)

Frequencies and orders of tangential magnetic forces are expressed by the same Equations 2.125, 2.126, and 2.127 as for radial magnetic forces. The stator and rotor MMF waveforms are expressed by Equations 2.107 and 2.108 in which ∞ F1 (α, t) = ∞ F (α, t) and F (α, t) = F (α, t). The peak values of 1ν 2 2µ ν=1 µ=1 the stator and rotor line current densities Am1ν and Am2µ are expressed by Equations 2.54 and 2.59. Note that the peak values of line current densities obtained from the MMF waveforms contain stator and rotor winding factors kw1ν and kw2ν for higher space harmonics. Thus, the following relationships between the MMF and line current densities amplitudes for the νth space harmonics can be written • for the stator F1mν =

1τ Am1ν ; νπ

Am1ν =

1 τ Am2µ µπ

Am2µ =

νπF1mν τ

(4.79)

µπ F1mµ . τ

(4.80)

• for the rotor F2mµ =

Assuming the magnetic saturation of the stator and rotor core to be negligible, on the basis of Ampere’s circuital law the peak values of the magnetic flux densities as functions of MMFs are • for the stator B1mν = µ0


F1mν kC g

(4.81)

DK3193

noise

October 24, 2005

19:9

102

Noise of Polyphase Electric Motors • for the rotor

F2m µ (4.82) kC g The ratio of the radial–to–tangential magnetic pressures created by the stator and rotor magnetic flux density components are: B2m µ = µ0

• for the product of the stator harmonics of the same number pr ν (α, t ) b1ν (α, t ) τ = = cot(νp α ∓ ωt ) pt ν (α, t ) 2µ0 a1ν (α, t ) 2νπkC g

(4.83)

• for the product of the stator ν harmonics and rotor µ harmonics pr νµ (α, t ) 2b1ν (α, t )b2µ (α, t ) τ = = cot(νp α ∓ ωt ) pt νµ (α, t ) 2µ0 b2µ (α, t )a1ν (α, t ) νπkC g

(4.84)

• for the product of the rotor harmonics of the same number pr µ (α, t ) b2µ (α, t ) τ = = cot(µp α ∓ ωµ t + φµ ) pt µ (α, t ) 2µ0 a2µ (α, t ) 2µπ kC g

(4.85)

Since τ >> g, the magnitude of the radial magnetic pressure is much higher than that of the tangential magnetic pressure. Consequently, the tangential forces producing electromagnetic torques are much smaller than radial forces. Peaks in Figure 4.13 have been obtained due to inaccuracy of the finite element method (FEM) model. Equations 4.83, 4.84, and 4.85 can be used for calculation of the tangential magnetic pressure on the basis of radial magnetic pressure.

4.9

Minimization of torque ripple in PM brushless motors

The torque ripple can be minimized both by the proper motor design and motor control. Measures taken to minimize the torque ripple by motor design include elimination of slots, skewed slots, special shape slots and stator laminations, selection of the number of stator slots with respect to the number of poles, decentered magnets, skewed magnets, shifted magnet segments, selection of magnet width, direction-dependent magnetization of PMs. Control techniques use modulation of the stator current or EMF waveforms [20, 59].

4.9.1

Slotless windings

Since the cogging torque is produced by the PM field and stator teeth, a slotless winding can totally eliminate the cogging torque. A slotless winding requires increased air gap, which in turn reduces the PM excitation field. To keep the same air gap magnetic flux density, the height h M of PMs must be increased. Slotless PM brushless motors use more PM material than slotted motors.


DK3193

noise

October 24, 2005

19:9

Torque Pulsations

4.9.2

103

Skewing stator slots

Normally, the stator slot skew bs equal to one slot pitch t1 can reduce the cogging torque practically to zero value — Equations 4.35 and 4.36. Sometimes, the optimal slot skew is less than one slot pitch [136]. On the other hand, the stator slot skew reduces the EMF which results in deterioration of the motor performance. Skewed slots are less effective in the case of rotor eccentricity.

4.9.3

Shaping stator slots

Figure 4.14 shows methods of reducing the cogging torque by shaping the stator slots, i.e., (a) (b) (c) (d)

bifurcated slots (Figure 4.14a), empty (dummy) slots (Figure 4.14b), closed slots (Figure 4.14c), teeth with different width of the active surface (Figure 4.14d).

Bifurcated slots (Figure 4.14a) can be split into more than two segments. The cogging torque increases with the increase of the slot opening. When designing

(a)

(b)

(c)

(d)

Figure 4.14 Minimization of the cogging torque by shaping the stator slots: (a) bifurcated slots; (b) empty slots; (c) closed slots; (d) teeth with different width of the active surface.


DK3193

noise

October 24, 2005

19:9

104


closed slots (Figure 4.14c), the bridge between the neighboring teeth must be properly designed. Too thick a bridge (thickness measured in radial direction) increases the stator slot leakage to an unacceptable level. Too narrow a bridge can be ineffective due to high saturation. Because closed slots can only accept “sewed” coils, it is better to close the slots by inserting an internal sintered powder cylinder or to separate the stator yoke and tooth-slot section. In the second case, stator slots are open externally from the frame side.

4.9.4

Selection of the number of stator slots

The smallest common multiple Ncm between the slot number s1 and the pole number 2 p has a significant effect on the cogging torque. As this number increases the cogging torque decreases [83]. Similarly, the cogging torque increases as the largest common factor between the slot number and the pole number increases [30, 83].

4.9.5

Shaping PMs

PMs thinner at the edges than in the center (Figure 2.2b) can reduce both the cogging and commutation torque ripple [75]. Magnets may require a polygonal cross section of the rotor core. Decentered PMs together with bifurcated stator slots can suppress the cogging torque as effectively as skewed slots with much less reduction of the EMF [175].

4.9.6

Skewing PMs

The effect of skewed PMs on the cogging torque suppression is similar to that of skewing the stator slots. Fabrication of twisted magnets for small rotor diameters and small number of poles is rather difficult. Bread loaf shaped PMs with edges cut at a slant are equivalent to skewed PMs.

4.9.7

Shifting PM segments

Instead of designing one long magnet per pole, it is sometimes more convenient to divide the magnet axially into K s = 3 to 6 shorter segments. Those segments are then shifted one from each other by equal distances t1 /K s or unequal distances the sum of which is t1 . Fabrication of short, straight PM segments is much easier than long, twisted magnets.

4.9.8

Selection of PM width

Properly selected PM width with respect to the stator slot pitch t1 is also a good method to minimize the cogging torque. The magnet width (pole shoe width) is to be b p = (k + 0.14)t1 where k is an integer [136] or including the pole curvature b p = (k + 0.17)t1 [92]. Cogging torque reduction requires wider pole shoes than the multiple of slot pitch.


DK3193

noise

October 24, 2005

19:9

Stator System Vibration Analysis

4.9.9

105

Magnetization of PMs

There is a choice between parallel, radial, and direction-dependent magnetization. For example, if a ring-shaped magnet of a small motor is placed around the magnetizer poles without an external magnetic circuit, the magnetization vectors will be arranged similar to the Halbach array. This method also minimizes the torque ripple.

4.9.10

Creating magnetic circuit asymmetry

Magnetic circuit asymmetry can be created by shifting each pole by a fraction of the pole pitch with respect to the symmetrical position, or designing different sizes of North and South magnets of the same pole pair [18].


DK3193

noise

November 3, 2005

15:28

5 Stator System Vibration Analysis In previous chapters the magnitudes, frequencies, and orders of radial and tangential magnetic forces have been analyzed. Analytical and numerical methods of computations of these forces have been discussed. The noise and vibration of the motor structure are the direct responses of the excitation by these forces. For example, if the frequency of the radial magnetic force is close to one of the natural frequencies of the stator system and the force order r is the same as the circumferential vibrational mode m of the stator system, significant vibration and acoustic noise can be produced. In this chapter, the modal vibration behavior of the stator system is discussed. Several analytical approaches for the calculation of natural frequencies and vibrational modes of cylindrical stators with frames (enclosures) are given.

5.1

Forced vibration

The stator and frame of a motor structure are basically cylindrical shells. A complete analytical analysis of the vibration of the system may be difficult. However, if only the stator is the concern, or if the overall vibration behavior of the motor is attributed to the stator, a circular cylindrical shell can be employed as a simplified model of the stator for an analytical approach. The vibration of a continuous structure due to the force excitation is discussed in Appendix D. For a cylindrical shell, although the equation of motion is of the same form as Equation D.32, the differential operator is more complicated than that for the beam and the plate [190]. In fact, due to the curvature of the shell, the vibration in the three orthogonal directions, radial, axial, and tangential, are coupled to each other. A result of this coupling is that any excitation in one direction would cause vibrations in all three directions. Such effects are more prominent at low frequencies. This makes it difficult to solve the equations directly, and various 107 Copyright © 2006 Taylor & Francis Group, LLC

DK3193

noise

November 3, 2005

15:28

108


approximations/simplifications have to be made in the analytical solution [190]. A detailed and accurate analysis may have to reply on the numerical approach such as the finite element method (FEM). The stator vibration is primarily induced by the electromagnetic force acting on the inner surface of the stator core. As described in previous chapters, the electromagnetic force has radial and tangential components. Both components, strictly speaking, may excite the vibration in the radial direction which is the primary concern in acoustic radiation. But since the radial component is approximately an order of magnitude larger than the tangential component in general, it might be reasonable to neglect the contribution from the tangential component in the electromagnetic force, which means that the coupling in the vibration between the radial and tangential directions could be neglected. The forced vibration response of a cylindrical shell under the excitation of the electromagnetic force in the radial direction thus should be of the form as given in Equation D.37. As an example, a stator could be simplified as a circular cylindrical shell with both ends free of constraints. It is subject to the electromagnetic force as expressed in Equation 2.118. If only the circumferential modes of the stator are of interest, it can be shown that the force component of the order r only excites the mode of the same order for the stator, i.e. r = m. The amplitude of vibration displacements of mode m can be derived as Fm /M Am = 2 (ωm − ωr2 )2 + 4ζm2 ωr2 ωm2

(5.1)

where M is the mass (kg) of the cylindrical shell, ωm is the angular natural frequency of the mode m, ωr is the angular frequency of the force component of the order r , and ζm and is the modal damping ratio. The amplitude of force Fm = π D1in L i Pmr , in which D1in is the inner diameter of the stator core, L i is the effective length of the stator core, and Pmr is the magnitude of the magnetic pressure of the order r according to Equations 2.115, 2.116, or 2.117. It can be seen that when the frequency of the excitation force is close to the natural frequency of the stator; the vibration reaches a maximum and excessive noise may be generated. This result indicates that among all the electromagnetic force components, only those of which the frequencies are close to the natural frequencies of the stator modes are important to the stator vibration response. By rearranging Equation 5.1 a magnification factor can be defined as hm =

Am 1 = . 2 2 Fm /Mωm [1 − ( fr / f m ) ]2 + [2ζm ( fr / f m )]2

(5.2)

The h m value calculated by Equation 5.2 is plotted against ( fr / f m ) and different damping ratios ζm in Figure 5.1. The magnification factor increases as f = | fr − f m | decreases and the damping factor ζm decreases. Particularly at fr = f m , the vibration amplitude is controlled by the mechanical damping in the structure. Analytical determinations of the mechanical damping are generally difficult. Experimental approaches are usually adopted. The mechanical damping


DK3193

noise

November 3, 2005

15:28


109

Magnification factor hm

10 xm = 0.05

8 6

xm = 0.1 4 xm = 0.2

2 0

xm = 0.3 0 0

1

0.5

2

1.5

2.5

3 3

fr/fm

Figure 5.1 Magnification factor h m as a function of ( fr / f m ) for different damping factors ζm . of the stator is frequency dependent and very much affected by winding and the lamination of the stator core. An empirical expression for the small and medium sized electrical machines is suggested as [248] ζm =

1 (2.76 × 10−5 f m + 0.062). 2π

(5.3)

This relationship is plotted in Figure 5.2. For a more accurate result, the modal damping ratio ζm can be extracted directly from modal testing [223]. The conversion between various damping parameters, such as the damping ratio, logarithmic decrement, and the loss factor, and so forth, can be found in Table 10.1. 0.06 0.06 0.05 0.04 ξ m(f ) 0.03 0.02 0.01 0

0

0 0

2000

4000

6000 f

8000

1 ⋅ 104 10000

Figure 5.2 Variation of damping ratio ζ with frequency [248].


DK3193

noise

November 3, 2005

15:28

110


It should be noted that Equation 5.1 is derived for a single circumferential mode. The total response of the stator subject to the electromagnetic force excitation should be the superposition of the contributions from all the modes, as discussed in Appendix D. However, if the vibration modes are well separated over the frequency domain, Equation 5.1 can be employed around the natural frequency of each mode. Expressed as a function of the magnetic pressure magnitude Pmr , Equation 5.1 becomes, Fm π D1in L i Am = hm = Pmr h m . (5.4) M ωm2 M ωm2 The amplitude of the vibration velocity for mode m is, Vm = ωr Am = 2π fr

5.2

π D1in L i Pmr h m . M ωm2

(5.5)

Simplified calculation of natural frequencies of the stator system

Calcultation of natural frequencies of the stator system is essential in the vibration analysis of electric machines. In classical approaches [87, 119, 200, 248], the stator system, that is, the stator core, winding, and frame (enclosure) is considered as a single thick ring loaded with teeth and winding. Circumferential vibrational modes of a thin cylinder can be visualized in a similar way to force orders as shown in Figure 2.9. The natural frequency of the stator system of the mth circumferential vibrational mode can be expressed as: Km 1 fm = (5.6) 2π Mm where K m is the lumped stiffness (N/m) and Mm is the lumped mass (kg) of the stator system. For a stator core with the thickness h c , mass Mc , and the mean diameter Dc , the lumped stiffness and lumped mass in the breathing mode, that is, the circumferential vibrational mode m = 0 is [119, 248] K 0 = 4π

Ec hc L i ; Dc

M0 = Mc kmd = π Dc h c L i ρc ki kmd

(5.7)

where ρc is the mass density of the stator core, ki is the stacking factor, and kmd is the mass addition factor for displacement defined as [248] kmd = 1 +

M t + M w + Mi . Mc

(5.8)

In the above equation, Mt is the mass of all stator teeth, Mw is the mass of stator windings, Mi is the mass of insulation, and Mc is the mass of the stator core cylinder (yoke).


DK3193

noise

November 3, 2005

15:28


111

Equations 5.6 and 5.7 give the following formula for the natural frequency of m = 0 circumferential mode Ec 1 f0 = . (5.9) π Dc ρc ki kmd Typically, for laminations, the elasticity modulus E c = 200 × 109 Pa = 200 GPa, mass density ρc = 7700 kg/m3 , and stacking factor ki = 0.96. This will give the phase speed cc = 200 × 109 /7700 = 6096 m/s. For the circumferential vibrational mode m = 1 (bending mode) K 1 = 4π F1 =

Ec hc L i ; Dc

M1 =

2 ; 1 + κ 2 kmr ot /kmd

Mc kmd M0 = 2 F12 F1

(5.10)

hc κ=√ . 3Dc

(5.11)

In Equation 5.11, kmr ot is the mass addition factor for rotation [248] hc M w + Mi 1 s1 ct L i h 2t hc 2 kmr ot = 1 + + 1+ + π Dc I c Mt 3 2h t 2h t =1+

s1 ct L i h 2t π Dc I c

M w + Mi Mt

1+

(4h 2t + 6h c h t + 3h 2c )

(5.12)

where

h 3c L i (5.13) 12 is the area moment of inertia about the neutral axis parallel to the cylinder axis, s1 is the number of stator teeth (slots), ct is the tooth width, and h t is the tooth height. Ic =

Then, the natural frequency for the circumferential mode m = 1 is [119, 248] Ec 2 1 f1 = = f 0 F1 . (5.14) π Dc ρc ki kmd 1 + κ 2 kmr ot /kmd Generally, for circumferential vibrational modes m ≥ 2 [119, 248] K m = 16π Mm = Mc

kmd m 2 + 1 kmd m 2 + 1 = π D h L ρ k c c i c i Fm2 m 2 Fm2 m 2

Fm =

E c Ic 2 (m − 1)2 ka2 Dc3

κ 2 (m 2 − 1)[m 2 (4 + kmr ot /kmd ) + 3] 1+ m2 + 1


−1/2

(5.15)

DK3193

noise

November 3, 2005

15:28

112


where the coefficient ka > 1 accounts for end bells and frame support (foot or flange mounting). The natural frequency of the stator system for circumferential modes m ≥ 2 can be derived as Ec Ic m(m 2 − 1) 2 1 m(m 2 − 1) √ fm = ka Fm = f 0 κ √ ka Fm . 2 π Dc ρc ki kmd h c L i m2 + 1 m2 + 1 (5.16) Equation 5.16 is in fact a modified equation for circumferential vibration of a ring derived by Hoppe [90]. Most monographs on noise analysis of electrical machines, for example, [87, 119, 200, 248], recommend Equations 5.9, 5.14, and 5.16 for calculating natural frequencies of stator systems. However, Equations 5.9 to 5.16 do not guarantee a good accuracy of the calculation since the stator system is a complex structure which consists of the laminated stack with yoke and teeth, winding distributed in slots, encapsulation (potting), and frame (enclosure or casing). Equations 5.9 to 5.16 can only be used to estimate the natural frequency of the stator core alone, without any frame and end bells and with only partial influence of the winding and teeth.

5.3

Improved analytical method of calculation of natural frequencies

The analytical approach presented in this section is an attempt to consider the effects of winding, teeth, and particularly the outer frame on the natural frequencies of the stator system. In this method, the stator core, winding, and teeth, and the frame are modeled separately. Then a formula is suggested for the system.

5.3.1

Natural frequency of the stator core

If the stator length–to–mean diameter ratio is L i /Dc ≤ 1, the stator core can be considered as a ring and the following classical formula for m ≥ 2, first published by Hoppe in 1871 [90], can be used 1 m(m 2 − 1) √ fm = 2π m2 + 1

E c Ic 2 m(m 2 − 1) √ = ρl (0.5Dc )4 π Dc2 m 2 + 1

E c Ic ρc L i h c

(5.17)

where the area moment of inertia Ic about the neutral axis parallel to the ring axis is given by Equation 5.13, h c is the stator yoke thickness, and the mass per unit circumference ρl =

1 Mc = ρc (π Dc )L i h c = ρc L i h c . π Dc π DC


(5.18)

DK3193

noise

November 3, 2005

15:28


113

According to Equations 5.6 and 5.17 the lumped stiffness and mass of the finite length cylinder are, respectively, Km =

16π E c Ic m 2 (m 2 − 1)2 Dc3 m2 + 1

Mm = Mc = πρc Dc L i h c .

(5.19) (5.20)

However, Hoppe’s equation does not give good results if L i /Dc ≥ 1. Better results are obtained if the stator of an electrical machine is regarded as a cylindrical shell of infinite length. In most mechanical engineering textbooks, the natural frequency of an infinitely long shell for circumferential vibrational modes m ≥ 0 is expressed as [135] Ec

m fm = (5.21) π Dc ρc (1 − νc2 ) where νc is the Poisson ratio for the stator core. On the basis of Donnel-Mushtari theory [135], the parameter m , that is, the roots of the second order characteristic equation of motion are: • for circumferential mode m = 0

0 = 1 • for circumferential modes m ≥ 1 1

m = (1 + m 2 + κ 2 m 4 ) ± (1 + m 2 + κ 2 m 4 )2 − 4κ 2 m 6 ; 2

(5.22)

(5.23)

• where the nondimensional thickness parameter κ2 =

h 2c 3Dc2

(5.24)

To express Equation 5.21 in the form of Equation 5.6, the lumped stiffness can be written as Km =

4 2m π L i h c E c . Dc 1 − νc2

(5.25)

Meanwhile, Mm = πρc Dc L i h c , which is actually the stator core mass Mc . It is recommended to use Equation 5.21 even if L i /Dc < 1. The stator core (yoke) behind the teeth of 2-pole machines (2 p = 2) is thicker in the radial direction than that of machines with 2 p > 2. Therefore, the stiffness (Equation 5.25) of 2-pole machines is higher and deflection of the stator core is smaller than in machines with a larger number of poles. Two-pole machines can often be less noisy than multipole machines.


DK3193

noise

November 3, 2005

15:28

114


5.3.2

Natural frequency of a frame with end bells

A frame (enclosure) with end bells behaves as a circular cylindrical shell with both ends constrained mechanically. Mathematically, both ends can be approximately described as being either simply supported or clamped. With the circumferential modal patterns being illustrated in Figure 2.9, typical axial modal patterns for circular cylindrical shells with simply supported ends are shown in Figure 5.3. The axial vibrational modes are n = 1, 2, 3 . . .. For a closed circular cylindrical shell of finite length L f with simply supported condition at both ends, the characteristic equation of motion has the following form [135]

6mn − (C2 + κ C2 ) 4mn + (C1 + κ C1 ) 2mn − (C0 + κ C0 ) = 0.

(5.26)

This is actually a third order equation with respect to the nondimensional frequency parameter 2mn . Three sets of roots correspond to the vibrations in three orthogonal directions, in which the smallest real root is associated with the natural frequency of the flexural vibration of the frame. According to Donnell-Mushtari theory, the constants in Equation 5.26 are [135] 1 C2 = 1 + (3 − ν f )(m 2 + λ2 ) + κ 2 (m 2 + λ2 )2 (5.27) 2

1 2 2 2 2 23 − νf 2 2 2 2 C1 = (1−ν f ) (3 + 2ν f )λ + m + (m + λ ) κ (m + λ ) (5.28) 2 1 − νf C0 =

λ = nπ

Rf ; Lf

1 (1 − ν f ) 1 − ν 2f λ4 + κ 2 (m 2 + λ2 )4 2 C2 = C1 = C0 = 0 R f = 0.5(D f − h f );

κ2 =

(5.29) (5.30) h 2f

12R 2f

(5.31)

where ν f is the Poisson ratio, R f is the mean frame radius, D f is the external frame diameter, h f is the frame thickness, and L f is the frame length. When both ends of the frame are clamped, the characteristic Equation 5.26, is still valid, and the constants in the equation are the same as shown in Equations 5.27 n=1

n=2

n=3

Figure 5.3 Nodal patterns for a circular cylindrical shell supported at both ends by shear diaphragms.


DK3193

noise

November 3, 2005

15:28


115

to 5.30. However, Equation 5.26 should be solved with the parameter λ in Equation 5.31 replaced by λe , which according to Arnold and Warburton is [135] λe = nπ

Rf ; L f − L0

L0 = L f

0.3 n + 0.3

(5.32)

where L 0 is the correction to the length of the “clamped-clamped” shell, which depends on the axial modal number n. For either a “clamped-clamped” shell or a shell simply supported at both ends, the circumferential vibrational modes m = 0, 1, 2, 3, . . . should be calculated for each of the axial vibrational mode n = 1, 2, 3, . . .. The axial mode n = 0 does not exist. Thus, the natural frequency of the frame is Ef K mn 1 1 mn f mn = = (5.33) 2π R f 2π Mf ρ f (1 − ν 2f ) where E f is the elasticity modulus of the frame material, and ρ f is the material density. The lumped stiffness for the frame therefore is K mn =

2mn E f V f 2 2mn π L f h f E f = Rf R 2f 1 − ν 2f 1 − ν 2f

(5.34)

and M f is the mass of the frame, i.e., M f = ρ f V f = ρ f (2π R f )L f h f ;

5.3.3

V f = 2π R f L f h f .

(5.35)

Natural frequency of a stator core–frame system

Most electrical machines have stator cores pressed into frames (enclosures). A frame increases both the mass and stiffness of the stator. If the stator core is short, an external structure produces a pressure which is transmitted uniformly throughout the stator core. Since the resultant moment of inertia of coaxial cylinders is equal to the sum of moments of inertia of each cylinder, the natural frequency of the statorframe system is to be calculated under the assumption that the lumped stiffness of (f) the stator core K m(c) and frame K mn are in parallel, that is, the equivalent stiffness (c) (f) is K m + K mn . The same assumption has been made for equivalent masses of the stator core and frame, that is, the resultant mass of the stator-frame system is Mc + M f . On the basis of these assumptions, the natural frequency of the stator-frame system can be expressed as1 : (f) K m(c) + K mn 1 f mn ≈ (5.36) 2π Mc + M f (f) where K m(c) can be determined based on Equation 5.19 or 5.25, and K mn is according to Equation 5.34. Lumped masses Mc and M f are calculated from Equations 5.20 and 5.35.

1 See also S. Huang, M. Aydin, and T. A. Lipo, Electromagnetic vibration and noise assessment for surface mounted PM machines, IEEE PES Summer Meeting, Vol. 3, pp. 1417–1426, Vancouver, BC, Canada, 2001.


DK3193

noise

November 3, 2005

15:28

116


In large machines the deflection of the stator core which produces noise is about ±0.25 × 10−4 mm [220]. On the other hand, the manufacturing tolerances between the outer diameter of the core and inner diameter of the frame are from ±0.02 to ±0.1 mm. This means that the mechanical coupling between the core and frame is small for low amplitude deflections [220]. Small machines with stamped ring laminations are more likely to produce some mechanical coupling than segmental laminations of large machines. The mechanical coupling between core and frame is increased due to radial expansion after heating. In the calculation of noise and vibration for most constructions of electrical machines, the • frame may be neglected only in large machines; • frame and its moment of inertia cannot be neglected in small machines.

5.3.4

Effect of the stator winding and teeth

The winding is located in slots which are separated by steel teeth. The toothslot zone with the winding can be regarded as an additional ring internal to the stator core (yoke). Thus, the natural frequency of the stator-frame system with the winding will be f mn ≈

(f)

K m(c) + K mn + K m(w) Mc + M f + Mw

1 2π

(5.37)

where K m(w) is the lumped stiffness of the tooth-slot zone including the winding and Mw is the mass of the teeth, winding, and insulation. The lumped stiffness of the tooth-slot zone can be determined based on the method given in Section 5.3.1. • for short length winding (finite length cylinder) K m(w) =

m 2 (m 2 − 1) 2π E w Iw m2 + 1 Rw3

(5.38)

• for long length winding (infinitely long cylinder) K m(w) =

2m E w Vw Rw2 1 − νw2

(5.39)

The nondimensional frequency parameter m is according to Equations 5.22 and 5.23 for Rw = Rc − h sl , where h sl is the slot depth, Vw = 2π Rw h t (L i + 2h ov ) is the volume of the winding considered as a cylinder, L i is the effective length of the stator stack, h ov is the axial length of a one-sided end connection, E w is the equivalent elasticity modulus for winding and insulation, and νw is the Poisson ratio for the winding with insulation. The elasticity modulus for copper winding is 110 to 120 GPa and for polymer insulation 3 GPa. The equivalent elasticity modulus for copper and insulation assuming the slot filling factor 0.5 is approximately 9.4 GPa [241]. The Poisson ratio for copper is 0.33 to 0.36 and for polymer insulation exceeds 0.36. The mass of winding Mw = m 1 ρw N1ltur n aw a p sw


(5.40)

DK3193

noise

November 3, 2005

15:28


117

where m 1 is the number of phases, ρw is the mass density of conductors (8890 kg/m3 for copper, 2700 kg/m3 for aluminum), N1 is the number of turns per phase, ltur n is the average length per turn, a p is the number of parallel current paths, aw is the number of parallel wires, and sw is the conductor cross section area. Similar to the stator core, Equation 5.39 gives better results than Equation 5.38 even if (L i + 2h ov )/(2Rw ) < 1.

5.3.5

Analytical calculation of natural frequencies for a stator core-winding-frame system

A cylindrical stator system (core, winding, frame, end bells) of a 10 kW a.c. machine with s1 = 36 slots will be considered. The mass of the stator core (yoke) is 8.58 kg, mass of teeth is 10.27 kg, mass of the winding is 11.45 kg, mass of the insulation is 0.57 kg, and mass of the aluminum frame is 6.24 kg. The dimensions of the stator system are: • • • • • • • • •

stator core inner diameter D1in = 0.16 m stator core outer diameter D1out = 0.233 m effective length of the stator core L i = 0.1975 m stator tooth width ct = 5.87 mm stator slot opening b14 = 3 mm thickness of stator yoke (core) h c = 8.42 mm length of winding overhang h ov = 48 mm diameter of frame D f = 0.246 m length of frame L f = 0.359 m

Material parameters are specified below: • • • • • • • • •

specific mass density of laminations ρc = 7700 kg/m3 specific mass density of copper ρw = 8890 kg/m3 specific mass density of frame (aluminum) ρ f = 2700 kg/m3 modulus of elasticity of laminations E c = 200 × 109 Pa modulus of elasticity of winding with insulation E w = 9.4 × 109 Pa modulus of elasticity of frame (aluminum) E f = 71 × 109 Pa Poisson ratio of laminations νc = 0.3 Poisson ratio of copper νw = 0.35 Poisson ratio of frame (aluminum) ν f = 0.33

Natural frequencies of the stator core The mean radius of the stator core (yoke) is Rc = 0.5(D1out − h c ) = 0.5(0.233 − 0.00842) = 0.112 m and the parameter κ according to Equation 5.24 is κ 2 = h 2c / (12Rc2 ) = 0.00842/(12 × 0.1122 ) = 0.000469. Since the stator core is regarded as an infinite shell, the parameter m and natural frequencies f m of the core are only a function of circumferential modes m = 0, 1, 2, 3, . . .. The parameter m is calculated with the aid of Equations 5.22 and 5.23, e.g., m = 0.08 for m = 2. Natural frequencies of the stator core calculated on the basis of Equation 5.21 for circumferential modes 0 ≤ m ≤ 8 are given in Table 5.1.


DK3193

noise

November 3, 2005

15:28

118


Table 5.1 Natural frequencies of the stator core for 0 ≤ m ≤ 8 calculated on the basis of Equation 5.21. 0 7572

1 116

2 464

circumferential mode m 3 4 5 6 1313 2517 4070 5971

7 8218

8 10,811

Natural frequencies of the stator winding The stator winding distributed in slots is regarded as a homogenous cylinder with an equivalent modulus of elasticity E w = 9.4 × 109 Pa (copper, insulation, encapsulation) and specific mass density equal to that of copper, that is, ρw = 8890 kg/m3 . The mean radius of the stator winding is Rw = 0.5(D1in +h t ) = 0.5(0.16 + 0.028) = 0.094 m and the parameter κ according to Equation 5.24 is κ 2 = h 2c /(12Rw2 ) = 0.0028/(12 × 0.0942 ) = 0.0074. Since the stator winding is regarded as an infinite shell, the corresponding parameter m and natural frequencies of the winding are only a function of the circumferential modes m = 0, 1, 2, 3, . . .. The parameter m is calculated with the aid of Equations 5.22 and 5.23, e.g., m = 0.307 for m = 2. Natural frequencies of the stator winding calculated on the basis of Equation 5.21 with E c , ρc , νc , and Rc replaced by E w , ρw , νw , and Rw , respectively, for circumferential modes 0 ≤ m ≤ 8 are given in Table 5.2.

Natural frequencies of the frame The mean radius of the frame is R f = 0.25(D f + D1out ) = 0.25(0.246+0.233) = 0.12 m and the paremeter κ according to Equation 5.24 is κ 2 = h 2f /(12R 2f ) = 0.000246 where the thickness of the frame h f = 0.5(D f − D1out ) = 0.5(0.246 − 0.233) = 0.0065 m. A cylindrical frame with end bells can be regarded as a “clamped-clamped” shell for which the equivalent wavelength λe as a function of axial vibrational modes n = 1, 2, 3, . . . is expressed by Equation 5.32. Putting λe into Equations 5.27, 5.28, 5.29, and 5.30, the roots of the third order characteristic Equation 5.26

Table 5.2 Natural frequencies of the winding for 0 ≤ m ≤ 8 calculated on the basis of Equation 5.21 in which E c = E w , ρc = ρw , νc = νw , and Rc = Rw . 0 1857

1 113

2 570

circumferential mode m 3 4 5 6 1360 2471 3901 5647


7 7708

8 10,080

DK3193

noise

November 3, 2005

15:28


119

Table 5.3 Smallest polyroots 2mn of Equation 5.26 for 0 ≤ m ≤ 8 and n = 1, 2, 3. axial mode n 0 1 2 1 0.65 0.251 0.079 2 0.902 0.599 0.3 3 0.937 0.787 0.538

circumferential mode m 3 4 5 6 7 8 0.049 0.082 0.174 0.344 0.623 1.048 0.176 0.17 0.254 0.433 0.731 1.182 0.379 0.34 0.414 0.603 0.926 1.414

can be easily found with the aid of, for example, Mathcad technical calculation tool. For m = 0 and n = 1 the polyroots are 201 = 0.65, 0.837, and 2.02; for m = 2 and n = 1 the polyroots are 221 = 0.079, 2.318, and 6.516; for m = 4 and n = 1 the polyroots are 241 = 0.082, 6.372, and 18.729, and so on. Table 5.3 contains the smallest polyroots 2mn for 0 ≤ m ≤ 8 and 1 ≤ n ≤ 3. The natural frequencies f mn of the aluminum frame with end bells found on the basis of Equation 5.33 are given in Table 5.4.

Natural frequencies of the stator system Natural frequencies of the stator system (core with teeth, winding with insulation, and frame) have been calculated using simplified Equation 5.37. The stiffness of the core K m(c) is given by Equation 5.25, the stiffness of the winding K m(w ) is given (f) by Equation 5.39, and the stiffness of the frame K mn is given by Equation 5.34. The mass of the core with teeth is Mc = 8.58 + 10.27 = 18.85 kg, mass of the winding with insulation is Mw = 11.45 + 0.57 = 12.02 kg, and mass of the frame is M f = 6.24 kg. The natural frequencies of the stator system for 0 ≤ m ≤ 8 and 1 ≤ n ≤ 3 are given in Table 5.5. It is always recommended to verify the natural frequencies of the stator system obtained from analytical Equations 5.17, 5.21, 5.33, 5.36, and 5.37 with the 3D FEM structural modeling.

Table 5.4 Natural frequencies of the frame with end bells for 0 ≤ m ≤ 8 and 1 ≤ n ≤ 3 calculated on the basis of Equation 5.33. axial mode n 1 2 3

circumferential mode m 0 5819 6856 6990

1 3616 5589 6406

2 2033 3952 5294

3 1604 3027 4444


4 2072 2976 4210

5 3010 3638 4644

6 4233 4751 5606

7 5697 6173 6947

8 7392 7850 8584

DK3193

noise

November 3, 2005

15:28

120


Table 5.5 Natural frequencies of the stator system for 0 ≤ m ≤ 8 and 1 ≤ n ≤ 3 calculated on the basis of Equation 5.37. axial mode n 1 2 3

5.4 5.4.1

circumferential mode m 0 4722 4897 4921

1 1299 2002 2293

2 995 1568 2010

3 1720 1950 2270

4 3039 3133 3309

5 4775 4831 4940

6 6904 6947 7028

7 9419 9458 9526

8 12,316 12,352 12,414

Numerical verification FEM modeling

Numerical approach in engineering computations is typically used when a system is far too complex to analyze analytically. One of the most popular generalpurpose 3D FEM analysis software packages is Ansys. In the FEM, a complex system is divided into very small pieces called elements, the shape and size of which are decided by the user. The accuracy of the FEM results generally depends on the number of elements in the model. The forces or loads on each element are calculated one by one using differential equations. The results then can be presented in numerical or graphical form. The following generic steps are taken to solve a problem in Ansys or other FEM software packages: • Building the geometry. 2D or 3D representation of the object or system is constructed. • Defining material properties. A library of the necessary material properties that compose the object or system is created. • Mesh generation. Most software packages can generate the mesh automatically. However, it must be defined how the modeled system should be broken down into finite elements. • Application of loads. Constraints such as physical loadings or boundary conditions are assigned. • Solution. Solver module solves numerically differential equations. The FEM software needs to understand what kind of solution is expected, that is, steady state, transient, and so forth. • Postprocessing. The results can be presented in form of graphs, contour plots, or tables. Structural analysis of machine housings, machine parts, and machine tools is the most common application of the FEM. The term “structural” implies not only civil engineering structures but also naval, aeronautical, and mechanical structures. The following types of structural analyses can be performed with Ansys: • Static analysis. It is used to determine displacements, forces, and stresses under static loading conditions.


DK3193

noise

November 3, 2005

15:28


121

• Modal analysis. It is used to calculate the natural frequencies and mode shapes of a structure. • Harmonic analysis. It is used to determine the response of a structure to harmonically time-varying loads. • Transient dynamic analysis. It is used to determine the response of a structure to arbitrarily time-varying loads. • Spectrum analysis. This is an extension of the modal analysis to calculate stresses and strains due to a response spectrum or a power spectral density (PSD) input (random vibration). • Buckling analysis. It is used to calculate the buckling loads and determine the buckling mode shape. Nonlinearities in static, dynamic, and buckling analyses can include plasticity, stress stiffening, large deflection, large strain, hyper elasticity, contact surfaces, and creep. The modal analysis is necessary to determine the vibration characteristics, that is, natural frequencies and mode shapes of a structure or machine component. The natural frequencies and mode shapes are important parameters in the design of a structure for dynamic loading conditions. They are also necessary in a spectrum analysis, mode superposition harmonic analysis, and transient analysis. In vibrational and acoustic analysis of electrical rotating machinery, it is necessary to compare the frequencies and orders of radial magnetic forces with natural frequencies and vibrational modes of the stator system. All frequencies of magnetic forces that are close to natural frequencies of the stator system for the same force order as vibrational mode can produce dangerous vibration and excessive noise. Figure 5.4a shows the Ansys element plot of the stator laminated core with 36 skewed axial slots. The Ansys model consists of 3D solid elements and surface to surface contact elements taken from the Ansys element library. The laminated stator is modeled as a continuous solid. The material properties assigned to the stator laminations are orthotropic in nature. The orthotropic material properties enable a difference in stiffness between the radial and axial directions. The stator

(a)

(b)

Figure 5.4 Element plots: (a) stator laminated stack; (b) stator winding.


DK3193

noise

November 3, 2005

15:28

122

Noise of Polyphase Electric Motors n=1

n=2

n=3

Figure 5.5 Circumferential modes m = 2 for axial modes n = 1, 2, and 3 for an aluminum frame with end bells (outer diameter 0.246 m, inner diameter 0.233 m, and length 0.359 m) obtained from the 3D FEM analysis. stack is connected to the frame (enclosure) using a bonded contact pair. Since the stator is pressed into the frame, the contact pair is also used to calculate the prestress effect of the interference fit on the natural frequencies of the system. The volume of material representing the stator winding is modeled also using solid elements. The stator axial slots are considered fully filled with the winding material. The winding parameters (specific mass density, elasticity modulus, Poisson ratio) are an equivalent representation of copper, insulation, and epoxy resin. Winding elements are shown in Figure 5.4b. The Ansys model of the aluminum frame is shown in Figure 5.5. For the circumferential mode m = 2 three axial modes, i.e., for n = 1, 2, and 3 has been plotted. It has been assumed that the frame is furnished with end bells (“clampedclamped” shell).

5.4.2

Comparison of analytical calculations with the FEM

First, a simple thin circular square solid ring with its diameter 0.508 m, thickness h c = 0.00127 m, axial length L i = 0.00127 m, modulus of elasticity E c = 206.8 × 109 Pa, and specific mass density ρc = 7955 kg/m3 will be considered. The natural frequencies f 0 for m = 0 and f 2 for m = 2 of radial vibration will be determined. The frequency f 0 can be calculated with the aid of Equation 5.9 for ki = 1 (solid homogeneous ring) and kmd = 1 (no mass addition), i.e., 1 206.8 × 109 f0 = = 3194.6Hz. π 0.508 7955


DK3193

noise

November 3, 2005

15:28


123

Table 5.6 Comparison of natural frequencies for 0 ≤ m ≤ 4 and n = 1, 2, 3 obtained from the 3D FEM and analytical calculations for an aluminum frame with end bells. Values calculated analytically are in parentheses. axial mode

m=0

m=1

n=1

-

2927 2927 2919 2924 (3616) (3563) 5184 5095 5134 5093 5139 (5589) (5516) 6311 6031 6176 6045 6201 (6406) (6331)

n=2

n=3

(5819) (5693) 6175 6220 6157 6218 (6856) (6772) 6865 6612 6794 6637 6818 (6990) (6914)

circumferential mode m=2 m=3 m=4 1884 1841 1833 1827 1829 (2033) (1985) 3626 3519 3524 3506 3526 (3952) (3889) 5182 4918 4980 4906 4996 (5294) (5222)

1687 1533 1520 1513 1517 (1604) (1518) 2986 2772 2764 2743 2764 (3027) (2950) 4491 4129 4148 4087 4157 (4444) (4364)

2298 1966 1941 1924 1933 (2072) (1963) 3202 2782 2773 2727 2767 (2976) (2876) 4504 3923 3939 3843 3937 (4210) (4114)

method ANS N1 N2 N3 N4 A(5.33) AWL ANS N1 N2 N3 N4 A(5.33) AWL ANS N1 N2 N3 N4 A(5.33) AWL

The frequency for m = 2 can be calculated according to Hoppe’s formula (Equation 5.17), i.e., 2(22 − 1) 206.8 × 109 × 2.168 × 10−13 2 √ f2 = = 12.373Hz π × 0.5082 22 + 1 0.013 where ρc L i h c = 7955 × 0.00127 × 0.00127 = 0.013 kg/m and the moment of inertia according to Equation 5.13 is Ic = 0.001273 × 0.00127/12 = 2.168 × 10−13 m4 . The same frequencies obtained from Ansys are f 0 = 3226.4 Hz and f 2 = 12.496 Hz. The convergence is good. For thin square solid ring analytical equations give slightly smaller values of natural frequencies. In the next example natural frequencies of a frame of an electrical machine will be calculated. Tables 5.6 and 5.7 show the comparison of natural frequencies of an aluminum frame with end bells (outer diameter 0.246 m, inner diameter 0.233 m, length 0.359 m) obtained from the 3D FEM and analytical calculations. Two different FEM packages, namely Ansys and Nastran have


DK3193

noise

November 3, 2005

15:28

124


Table 5.7 Comparison of natural frequencies for 5 ≤ m ≤ 8 and n = 1, 2, 3 obtained from the 3D FEM and analytical calculations for an aluminum frame with end bells. Values calculated analytically are in parentheses. axial mode n=1

n=2

n=3

m=5

circumferential mode m=6 m=7 m=8

method

3404 2891 2836 2804 2819 (3010) (2897) 4080 3425 3410 3323 3392 (3638) (3526) 5177 4309 4343 4178 4326 (4644) (4535)

4826 4136 4023 3968 3990 (4233) (4118) 5395 4516 4470 4335 4434 (4751) (4636) 6351 5186 5231 4979 5193 (5606) (5491)

ANS N1 N2 N3 N4 A(5.33) AWL ANS N1 N2 N3 N4 A(5.33) AWL ANS N1 N2 N3 N4 A(5.33) AWL

6510 5656 5447 5355 5387 (5697) (5582) 7025 5939 5825 5631 5762 (6173) (6055) 7893 6446 6474 6123 6405 (6947) (6829)

8436 7446 7087 6937 6988 (7392) (7275) 8920 7654 7421 7149 7317 (7850) (7731) 9731 8027 7991 7527 7881 (8584) (8465)

been used. In order to show the convergence of the results with the number of elements, two mesh sizes have been examined in the Nastran analysis. Results obtained from different element types are also listed in Tables 5.6 and 5.7. The following abbreviations have been used: ANS = Ansys (cube elements), N1 = Nastran 2003 14 × 40 HEX8 elements, N2 = Nastran 2003 30 × 60 HEX8 elements, N3 = Nastran 2003 14 × 40 CQUARD4 elements, N4 = Nastran 2003 30 × 60 CQUARD4 elements, A(5.33) = analytical method according to Equation 5.33, AWL = analytical method according to Wang and Lai [226]. Both analytical methods give reasonable estimations on the natural frequencies except for the low order modes, such as m = 1, n = 1. Figure 5.6 shows circumferential modes m = 2 and 3 and axial mode n = 1 obtained from the 3D FEM Ansys for the stator with aluminum frame equipped with end bells (outer diameter 0.246 m, inner diameter 0.233 m, length 0.359 m), laminated core, and copper winding. The diameters are as follows: stator inner


DK3193

noise

November 4, 2005

14:37

Stator System Vibration Analysis m=2

125 m=3

Figure 5.6 (See color insert following page 236.) Circumferential modes m = 2 and 3 and axial mode n = 1 for the stator (aluminum frame with end bells, laminated core, and copper winding) obtained from the 3D FEM Ansys. (teeth) diameter 0.1601 m, stator core (yoke) inner diameter 0.1769 m, stator outer diameter 0.233 m, frame outer diameter 0.246 m. The stator core length L i = 0.1975 m, stator winding length with end connections 0.2935 m, stator frame length L f = 0.359 m. The natural frequencies obtained from analytical calculations (Equation 5.37) and Ansys 3D FEM software package for the stator system (aluminum frame with end bells, laminated core, copper winding) have been compared in Table 5.8. Although, the accuracy of Equation 5.37 is reasonable, it can be found that as the system becomes complicated, the accuracy of analytical solutions deteriorates. For a complete electrical machine system, it is always recommended to use the FEM analysis for accurate results. Table 5.8 Comparison of natural frequencies obtained from analytical calculations and Ansoft 3D FEM software package for the stator system (aluminum frame with end bells, laminated core, copper winding). Values calculated analytically are in parentheses. axial circumferential mode m mode n 0 1 2 3 4 5 6 7 8 1 4289 − 1009 2089 3332 4541 5785 − − (4722) (1299) (995) (1720) (3039) (4775) (6904) (9419) (12,316) 2 − − 1789 2388 3401 4553 − − − (4897) (2002) (1568) (1950) (3133) (4831) (6947) (9458) (12,352) 3 4934 − − − − − − − − (4921) (2293) (2010) (2270) (3309) (4940) (7028) (9526) (12,414)


DK3193

noise

November 4, 2005

14:37

6 Acoustic Calculations The sound radiated from structures is an impact of the structural vibration on the environment. From the vibro-acoustic point of view, it is a result of interactions between the structure and the ambient medium. Generally speaking, more vibrations may produce more sound. However, the relation between the two is not simple, especially for machinery structures. The sound radiation generally depends on the structure dimensions, boundary conditions, material properties, external excitations, and ambient medium conditions, and so on. In acoustics, in order to facilitate the discussions on the mechanism of sound generations from various structures, simple structural elements, such as plates, spheres, and cylinders, are treated as basic sound radiators because of their frequent occurrences in practical applications [160]. For more than a century, numerous papers and books have been published on almost every aspect of the topic. It is, therefore, not wise and impossible to cover all the materials in a document of limited size. The intention of this chapter is to summarize the outcome that may be related to the sound radiation from electric motor structures, and to present readers with the most recently updated results and techniques to benefit their relevant studies and practices. For this purpose, plane and cylindrical radiators are discussed directly without much emphasis on the basic acoustics theories. Readers might need to refer to Appendix A and [124, 176] for prerequisite knowledge for the contents of this chapter.

6.1

Sound radiation efficiency

As it is known, a vibrating structure produces sound waves in the ambient medium. The sound pressures at any specific location in the medium space, and the sound power radiated by the structure are the two parameters used to quantify the local and global acoustic effects in the field. They are generally functions of vibration levels on the structure surface, which are further determined by the boundary conditions and the external excitations. For a specific structure, higher vibration levels mean higher sound pressures in the field and higher radiated sound power. 127 Copyright © 2006 Taylor & Francis Group, LLC

DK3193

noise

November 4, 2005

14:37

128


In many cases, however, it is not the absolute value, but the transfer function from the structural vibration to the acoustic response that is of greatest interest. A properly defined vibro-acoustic transfer function could fundamentally reveal the energy transfer mechanisms and efficiently eliminate the duplications of the tedious derivation/calculation in accounting for the interactions between the structure and the ambient medium, thus simplifying the vibro-acoustic analysis. One of the most common definitions used in vibro-acoustic analysis is the sound radiation efficiency, as given by, σ =

ρ 0 c0 S v 2

(6.1)

where ρ0 , c0 are the density and the sound speed in the ambient medium (air) respectively, S is the area ofthe radiating surface, is the sound power radiated from the structure, and v 2 is the spatial averaged mean square velocity over the structure radiating surface. This expression relates the radiated sound power to the spatial averaged vibration level. Physically, it indicates the capability of a structure in sound radiation in comparison with a piston with the same surface area and same vibration levels under the condition that the piston circumference greatly exceeds the acoustic wavelength [176]. However, since the vibration distribution is generally affected by the excitation, the spatial averaged mean square velocity could be quite different for different excitation configurations. For example, a point force excitation might result in different sound radiation efficiencies from a distributed force or point moment excitations for the same structure. As a result, the modal radiation efficiency (σm ) associated with each vibration mode (m) is normally introduced in theoretical analyses [37], m σm = (6.2) ρ0 c0 S v m2 where m is the sound power radiated by the mth mode of the structure, and v m2 is the spatial averaged mean square velocity of the mth mode. The modal radiation efficiency is generally independent of the external excitation and, hence, reflects the inherent characteristics of the structure in the energy transfer. Since the total sound power is the sum of the power radiated from each vibration mode m , i.e., = m , the relationship between Equations 6.1 and 6.2 is [37], m

σm v m2 m σ = . v m2

(6.3)

m

As Equation 6.3 takes into account the contribution from each vibration mode, the result is also called the modal averaged radiation efficiency. In practice, the modal averaged radiation efficiency is what can be directly measured.


DK3193

noise

November 4, 2005

14:37

Acoustic Calculations

129

On the other hand, since the radiated sound energy is a loss to the structural vibrational energy, the radiation loss factor can be defined to describe the extent to which the vibration is damped due to the radiation of power from the structure to the ambient medium, i.e., ηrad = (6.4) ωE where ω is the angular frequency, E is the total vibrational energy of the structure. For a uniform structure with a spatially uniform vibration distribution over the structure surface E = M v 2 , where M is the mass of the structure [37]. The relationship between the radiation efficiency and the radiation loss factor is, σ =

M k0 ηrad ρ0 S

(6.5)

where k0 = ω/c0 is the acoustic wavenumber. Compared to the radiation efficiency, the radiation loss factor is a small number, especially for the radiation into the air. Since the radiation loss factor is also related to the spatial averaged mean square velocity over the structure surface, the modal averaged radiation loss factor (ηrad ) can be expected to have the same relationship to the corresponding modal averaged radiation efficiency (σ ) as that shown in Equation 6.5. Another parameter often used in the sound radiation analysis is the acoustic radiation impedance, defined on the radiating surface of the structure as, p( r )S ZR = (6.6) v r=r 0 where p( r ) is the sound pressure in the field, v is the vibration distribution over the structure, S is the radiating surface area, and the subscript r = r0 denotes the location on the surface of the structure. The real part of the impedance represents the energy radiated into the ambient medium, while the imaginary part of the impedance represents the energy stored in the near field of the source [124]. This expression is convenient to use when the vibration amplitude is uniform over the structure; examples are, a point source, a piston radiator, or a plate vibrating in one of its modes. The relation between the acoustic radiation impedance and the radiation efficiency can be found as, σ =

e[Z R ] ρ 0 c0 S

(6.7)

where e denotes the real part of a complex number.

6.2

Plane radiator

Plane structures are very common in industries. Substantial research has addressed the importance of the sound radiation from plane structures. For electric machine structures, plates are often found as accessory components that are necessary for the assembly and the installations; for example, the end-shields and the support


DK3193

noise

November 4, 2005

14:37

130


which in some cases are found important in the total sound power contributions. In this section, the sound radiation from an infinite plate is briefly reviewed to benefit the analysis of finite plates. Then the finite-sized plates are discussed in regard to their modal radiation and modal averaged radiation efficiencies.

6.2.1

Infinite plates

In order to have a better understanding of the sound radiation phenomena of practical plane radiators, it is helpful to have a simple review on the sound radiation from an infinite plate (Figure 6.1). For simplicity, it will be assumed an infinitely extended plate in the y = 0 plane is vibrating transversely with a velocity distribution given as (omitting the time component e jωt , without loss of generality), v(x) = v 0 e− jk p x

(6.8)

where k p is the wavenumber of the plate in bending motion. This equation actually represents a one-dimensional approaching wave propagating in the +x direction with the wavelength and the wave speed being λ p and c p , respectively. Since the vibration is uniform in the z direction (−∞, +∞), and so will be the acoustic field, the acoustic wave motion equation can be simplified as ∂ 2 p(x, y) ∂ 2 p(x, y) + + k02 p(x, y) = 0 ∂x2 ∂ y2

(6.9)

where p(x, y) is the radiated sound pressure in the field. The solution of Equation 6.9 may be generally written as p(x, y) = p0 e− jkx x e− jk y y

(6.10)

y

l0 ky

k0 kx

kp

q

x lp

Figure 6.1 Radiation from an infinite plate.


DK3193

noise

November 4, 2005

14:37


131

where k x and k y are the two components of the acoustic wavenumber k0 in the x and the y directions, respectively, and k02 = k x2 + k 2y . This expression must satisfy the acoustic boundary condition at y = 0, which is 1 ∂ p(x, y) v(x) = − jωρ0 ∂ y y=0 then,

kx = k p

p0 = v 0 ρ0 c0 k0 /k y = v 0 ρ0 c0 /

(6.11)

1 − k 2p /k02 .

The radiated sound pressure in the field is therefore given as √2 2 v 0 ρ0 c0 p(x, y) = e− jk p x e− j k0 −k p y 1 − k 2p /k02

(6.12)

From Equations 6.6 and 6.12, the radiation impedance per unit area of the plate can be obtained as p(x, y) ρ 0 c0 ρ 0 c0 ZR = = = (6.13) v(x) y=0 2 2 1 − k /k 1 − λ2 /λ2 p

0

0

p

Equation 6.13 indicates that for an infinite plate, the sound radiation has two fundamentally different behaviors. If the wavelength in the plate λ p is greater than the acoustic wavelength λ0 in the air (or equivalently, the wave speed in the plate c p is faster than the speed of sound in the air c0 , namely the supersonic effect), the radiation impedance of the plate is a real number, indicating that the energy is transmitted into the ambient medium. The sound radiation efficiency per unit area of the plate can be written as σ =

e(Z R ) 1 1 = = , or simply ρ0 c0 1 − k 2p /k02 1 − λ20 /λ2p

σ =

1 . (6.14) cos θ

If the structural wavelength is smaller than the wavelength of the sound wave in the air (or equivalently, the structural wave speed is slower than the speed of sound in the air, namely the subsonic effect), the radiation impedance is a pure imaginary number, which means that the sound pressure is 90◦ out of phase with the plate velocity. In this case, the energy will be exchanged between the plate and the acoustic medium with no sound energy radiated. The sound radiation efficiency of the plate is therefore zero, and the sound pressure simply decays along the y direction in the form as √2 2 jv 0 ρ0 c0 p(x, y) = (6.15) e− jk p x e− k p −k0 y . k 2p /k02 − 1 An important phenomenon that must be addressed here is, when the acoustic wavelength is matching the structural wavelength, i.e., λ p = λ0 , Equation 6.12


DK3193

noise

November 4, 2005

14:37

132


Radiation Efficiency, σ

100

10

1

0.1 1

0.1

10

f/fc

Figure 6.2 Radiation efficiency as a function of frequency for an infinite plate.

gives an infinity solution for the sound pressure in the field. This is because the radiated sound wave, under this condition, is propagating in the direction parallel to the structural wave in the plate, both waves having the same wavelengths and speeds. As a result, the sound wave will be enhanced continuously along the propagation path and the energy will be accumulated to infinity. In practice, due to the finite size and the damping in the plate, the radiated sound pressure always remains finite. However, large radiation still can be observed. Note that the wavelengths of both structural and acoustic waves are frequency dependent. A frequency, namely the critical frequency of the plate f c , at which the structural wavelength is equal to the acoustic wavelength (or equivalently, the structural wave speed is equal to the speed of sound in the air) exists, i.e., [55], cp = and therefore,

1/4 √ ω Eh 2 = ω· kp 12ρ(1 − ν 2 ) √

3c0 fc = πh

ρ(1 − ν 2 ) . E

(6.16)

In Equation 6.16 h is the thickness of the plate. E, ρ, and ν are the Young’s modulus, the density and the Poisson’s ratio of the plate material respectively. Figure 6.2 shows the variation of the sound radiation efficiency vs. the frequency. It is seen that for an infinite plate, sound radiation only occurs in the frequency range above the critical frequency f c . The radiation efficiency approaches to unity quickly as the frequency increases, indicating that the plate is an efficient radiator at high frequencies.


DK3193

noise

November 4, 2005

14:37


6.2.2

133

Finite plates in bending motion

When a plate is finite in size, its vibration will have modal behaviors. The modal parameters, such as natural frequencies, mode shapes, and so on, are inherent dynamic characteristics of a structure, depending on the geometry, material properties, and boundary conditions of the structure. Theoretically, a vibration of a structure can be described by a superposition of all the vibration modes with different participation amplitudes that are determined by the external excitations (see Appendix D). Consequently, the sound radiation due to the vibration of the structure can and will be related to the structural modal behavior. In order to discuss the effect of the finite size on the sound radiation from a plate, it is instructive to consider a plate strip of width l in the x direction embedded in an infinite baffle with simply supported boundary conditions at both edges, as shown in Figure 6.3. With the baffle, the sound radiated from each side of the plate does not interfere with each other, such that only the sound field on one side of the plate is considered here. Again, assume that the vibration in the z direction is uniform (−∞, +∞), the velocity distribution for a particular mode m over the y = 0 plane is,

v m (x) =

v 0 sin(mπ x/l) 0 < x < l 0 0 > x > l.

(6.17)

By transforming v m (x) to the wavenumber domain, Vm (k x ) =

l

v m (x)e jkx x d x = v 0 ·

0

mπ [(−1)m e− jkx l − 1] · l k x2 − (mπ/l)2

(6.18)

where Vm (k x ) and v m (x) satisfy, v m (x) =

1 2π

∞

−∞

Vm (k x )e− jkx x dk x .

(6.19)

Equation 6.19 indicates that the modal vibration distribution in the x direction can be decomposed into a series of waves propagating along an infinite plate in the x direction with different wavenumbers and magnitudes. The sound radiation from each of these waves thus can be treated as the radiation from an infinite plate with the velocity distribution as Vm (k x )e− jkx x . Based on Equation 6.12, the sound

y

lpm = 21/m x l

Figure 6.3 Modal behavior of a plate strip in a baffle.


DK3193

noise

November 4, 2005

14:37

134


pressure produced on the plate surface associated with each wave component is ρ 0 c0 k 0 Pm (k x )| y=0 = Vm (k x ) . k02 − k x2

(6.20)

Correspondingly, the total sound pressure due to all the wave components can be written as, ∞ 1 ρ0 c0 k0 − jkx x pm (x)| y=0 = Vm (k x ) e dk x . (6.21) 2π −∞ k02 − k x2 With the known vibration velocity distribution and its induced sound pressure on the plate surface, the time-averaged sound power radiated per unit length (in the z direction) of the plate is,   l ∞ 2 c k |V (k )| 1 ρ 0 0 0 m x m (ω) = e e  pm (x)| y=0 v m∗ (x)d x = dk x  2 4π 0 −∞ k2 − k2 0

x

(6.22) where “∗ ” denotes the complex conjugate. Note that the term within the integral is real only for −k0 < k x < k0 . By using Equation 6.18, Equation 6.22 can be rewritten as k0 m 2l 2 k x l − mπ 2 m (ω) = πρ0 c0 k0 v 0 dk x . (6.23) sin2 2 −k0 (k 2 l 2 − m 2 π 2 )2 k02 − k x2 x Although Equation 6.23 cannot be evaluated in a closed form, approximate solutions corresponding to the two distinguished regions discussed for the infinite plate can be obtained. • If k0 mπ/l, i.e., the acoustic wavelength is greater than the structural wavelength λ pm (or equivalently, in the subsonic region where the structural wave speed is slower than the speed of sound), Equation 6.23 becomes [37] 2 1 ρ0 c0 k0 v 02l 2 k0 ρ0 c0 k0 v 02 l m (ω) ≈ dk x = (6.24) 2m 2 π 3 2 mπ −k0 k02 − k x2 then the corresponding modal radiation efficiency is, 2 m 2k0 l 4m = σm = = . l mπ ρ0 c0lv 02 ρ0 c0l v m2

(6.25)

• If k0 mπ/l, i.e. the acoustic wavelength is shorter than the structural wavelength λ pm (or equivalently, in the supersonic region where the structural wave speed is faster than the speed of sound), one gets m (ω) ≈


DK3193

noise

November 4, 2005

14:37


135

ρ0 c0 k0lv 02 /[4 k02 − (m π/l )2 ] ≈ ρ0 c0lv 02 /4 . The modal radiation efficiency is, m k0 ≈ σm = ≈ 1. (6.26) 2 2 k0 − (m π/l )2 ρ0 c0 l v m Compared to the infinite plate, the most prominent feature of a finite plate in its sound radiation is that the modal radiation is not zero in its subsonic region. Given an explanation, a finite plate with simply supported edges in a particular mode is considered in Figure 6.4. For a higher order mode of which structural wavelengths λ px , λ py along the two directions are smaller than the acoustic wavelength λ0 in the air, as shown in Figure 6.4a, the adjacent subsections of which the vibration is 180◦ out of phase are separated by less than a wavelength in the surrounding air. The air displaced outward by one subsection moves to occupy the space left by the adjacent subsection, without being disturbed much in the far field, and thus very little power is radiated. In other words, the sound radiated from adjacent subsections cancel each other. At the corners, however, such “cancelation” is not complete with four monopole sources remaining and making primary contributions to the total sound power radiated from the plate. If only one of the structural wavelengths is smaller than the acoustic wavelength, for example λ px < λ0 as shown in Figure 6.4b, the adjacent subsections are separated by less than an acoustic wavelength in the l x direction, while the adjacent subsections in the l y direction are beyond an acoustic wavelength. In this case, the cancelations only happen along one direction as shown in the graph with two edges remaining as effective radiators. Along the l y direction, the two

Uncanceled Corners

+ ly l0 l py

Uncanceled Edges

− + +

−

+ Typical Region of Cancelations

+

−

+

l0

lpy

+

ly − +

l0

(a)

−

Typical Region of Cancelations − + −

lx

lx lpx

−+

−

lpx

l0 (b)

Figure 6.4 Sound radiation from: (a) uncanceled corners; (b) uncanceled edges of a finite plate.


DK3193

noise

November 4, 2005

14:37

136


subsections radiate sound independently. Thus, edges normally radiate more power than corners in the frequency range where they both occur. In the supersonic region, the structural wavelengths in both directions are greater than the acoustic wavelength, i.e. λ px , λ py > λ0 . All subareas radiate sound independently and make their contributions in the total sound power. If the plate is large compared to the acoustic wavelength, the boundary effect is negligible and the acoustic behavior of a finite plate approaches to that of an infinite plate. Figure 6.5 shows the modal radiation efficiencies of a series of modes associated with a baffled finite plate with simply-supported boundary conditions. As expected, all the modal radiation efficiencies generally increase as the acoustic wavenumber k0 in the subsonic region, and reach maximums around k0 = k p (k p = (nπ/l x )2 + (mπ/l y )2 for a simply-supported plate, where n, m are the mode numbers along the l x and l y directions, respectively). In the supersonic region, the radiation efficiencies are unity as that of the infinite plate. An interesting and important feature that may be concluded from Figure 6.5 is that, for each mode (n, m), the frequency at which the acoustic wavenumber k0 matches the structural wavenumber k p is different. In the frequency domain, lower order modes reach unity in their sound radiation efficiencies at lower frequencies, thus being more efficient in sound radiation than higher order modes. Apparently, an accurate calculation of the sound power radiated from a finite plate depends on a detailed analysis of the plate vibration from which the participation amplitudes of each mode in the total response can be determined.

10

Modal Radiation Efficiency

1 0.1 Mode (1, 1)

0.01

Mode (1, 3) Mode (3, 3)

0.001

Mode (1, 2) Mode (2, 3)

0.0001

Mode (2, 2) 0.00001 0.01

0.1

1

10

k0/kp(m, n)

Figure 6.5 Modal radiation efficiencies of a finite plate in a baffle [221].


DK3193

noise

November 4, 2005

14:37


137

However, if the plate is mechanically excited by a random noise, the vibration response within a frequency band f is primarily dominated by the modes of which the natural frequencies are within this band. Assume that the modal energies within this band are equally distributed, general expressions for the modal averaged radiation efficiency of the plate might be derived. Since each mode generally takes effect around their own natural frequencies because of a relatively large participation amplitude in the total vibration response, it has been shown that statistically a maximum that could exceed unity can still be observed in the modal averaged sound radiation efficiency around f c for a finite plate [55]. Two specific plate configurations that were considered in the analysis are that the plate is embedded in an infinite baffle and that the plate is unbaffled with the edges being acoustically free. Baffled plates Specifically, at low frequencies, the behavior of the plate is dominated by the first mode, normally the piston mode, of the plate. In this case, the radiation efficiency of the plate embedded in an infinite baffle can be calculated using the following equation [176, 212] σ =1−

2J1 (4k0 S/π ) 4S f 2 ∼ 2 8k0 S/π c0

for

λ0

√

2S

(6.27)

where J1 is the Bessel function of the first order. Ver and Holmer [212] suggested this equation generally applies below the first natural frequency of the plate, that is, f < [c02 /(2S f c )][P 2 /(8S) − 1] , where S is the area of one side of a plate, and P is the plate perimeter. When the baffled plate is in the multimodal region, a formula for the modal averaged radiation efficiency based on the simply supported boundary conditions is given in [146], i.e.,  σ cor ner + σ edge f < fc     √ σ = (6.28) l x /λc + l y /λc f = f c      √ 1/ 1 − f c / f f > fc where σ cor ner , and σ edge are the modal averaged radiation efficiencies associated with the radiation regions around corners and along the edges of the plate respectively. They were given as, σ cor ner

8 λ2 = 4· c × π S

σ edge =

√ (1 − 2α)/(α 1 − α 2 ) f < f c /2 0 f > f c /2

1 Pλ2c (1 − α 2 ) ln[(1 + α)/(1 − α)] + 2α · · 2 4π S (1 − α 2 )3


(6.29)

(6.30)

DK3193

noise

November 4, 2005

14:37

138


√ where α = f / f c . It was suggested that for the clamped boundary conditions, 3 dB should be added to the radiation efficiency in the range f < f c [212]. Figure 6.6 compares the modal averaged radiation efficiencies of two plates calculated analytically and by the boundary element method (BEM). In BEM calculations, the baffled plate was excited mechanically by a point force. To simulate the equally distributed modal energies, the BEM results were averaged over three excitation locations randomly chosen on the plate. It is seen that for small plates, the piston mode could dominate the acoustic behavior even higher than the critical frequency. As the plate becomes bigger, and thinner, the piston mode may only work at very low frequencies. However, Figure 6.6b shows that the piston mode region is wider than that suggested for simply supported and clamped boundary conditions. This is simply because the first mode of the plate (piston mode) is more efficient in sound radiation than higher order modes. Equation 6.28 works well in the multimodal region, which is above the frequency corresponding to the lower one of the natural frequencies f 21 and f 12 , i.e., f > [c02 /(2S f c )][P 2 /(8S) + 3l y /(2l x ) − 1] for a rectangular plate with l x and l y being the long and short edges of the plate, respectively. For baffled plates with completely free boundary conditions, it was shown that rigid motions, one piston and two rotation motions, may predominately contribute to the low frequency sound radiation. As a result, it seems difficult to derive a general expression for the modal averaged radiation efficiency because, depending on the excitations, the sound radiation efficiency may behave quite differently due to the contribution from one or some of these rigid motions [15]. However, studies have shown that if there is any localized constraints along which the two rotation motions are restricted, Equations 6.27 and 6.28, though developed based on simply supported boundary conditions, might give reasonable results in the sound radiation efficiency. Unbaffled plates As mentioned above, the baffle around the plate separates the space into halves, in which the sound radiation considered is only from one side of the plate. In practice, a typical example of this is a plate backed with a large cavity enclosed by rigid walls. The sound radiation from the two sides are isolated from each other or their interactions can be ignored. However, in many places, plane structures are found unbaffled. In this case, the sound radiation from both sides of the plate and their interactions should be considered. An instant effect of removing the baffle in the sound radiation discussions is that the radiating area in the definition of the sound radiation efficiency (Equation 6.1) is twice that of the baffled plate. For an unbaffled plate, the radiation from both sides may exchange the energy due to the oscillating inertial flow around the edge. Since the vibration velocities on both sides are 180◦ out of phase relative to their normal directions (into the ambient medium), the overall sound pressure field is of different characteristics from that of a baffled plate. Specifically, at low frequencies where the piston mode dominates the plate behavior, the unbaffled plate becomes a dipole source


DK3193

noise

November 4, 2005

14:37


139

10 BEM, Simply-Supported Plate


BEM, Clamped Plate 1

Equation 6.27

0.1

0.01

0.001 1000

100

10

10000

Frequency (Hz) (a) 10


BEM, Simply-Supported Plate BEM, Clamped Plate Equation 6.28

1

Equation 6.27

0.1

0.01

0.001 1000

100

10

10000

Frequency (Hz) (b)

Figure 6.6 Modal averaged radiation efficiencies of a baffled plate: (a) 0.17 m×0.19 m×0.01 m aluminum plate, 9 × 10 elements in the BEM model; (b) 0.6 m×0.7 m×0.005 m steel plate, 40 × 46 elements in the BEM model. acoustically. Asymptotic solutions for the radiation efficiency of an unbaffled circular plate with radius a were derived by Silbiger [188], i.e.,

σ =

8(k0 a)4 /(27π 2 ) k0 a 1 1 k0 a 1.


(6.31)

DK3193

noise

November 4, 2005

14:37

140


On the basis of Equation 6.31, a general expression for an arbitrarily shaped plate with the unbaffled acoustic boundary condition can be obtained as,

σ =

√ 4S 2 f 4 /c04 λ0 > √2S 1 λ0 ≤ 2S

(6.32)

where S is the area of one side of a plate. In the multimodal region, below the critical frequency, instead of pulsating sphere and cylinder sources radiating into the half space, corners and edges of an unbaffled plate will behave more like oscillating sphere and cylinder sources, which are less efficient in sound radiation [124]. As a result, the modal averaged radiation efficiency of an unbaffled plate with simply supported boundary conditions can be written as [172] σ = F plate (Fcor ner σ cor ner + Fedge σ edge )

for

f < fc

(6.33)

where F plate , Fcor ner , Fedge are the correction factors due to the unbaffled condition. Particularly, F plate is for the effect of the inertial flows that surround the plate when the acoustic wavelength is greater than the plate dimensions, while Fcor ner , Fedge account for the localized inertial flow effects associated with corner and edge radiations. Semiempirical expressions for these correction factors were given as, F plate =

53 f 4 S 2 /c04 ; 1 + 53 f 4 S 2 /c04

Fcor ner =

1 13α 2 ; 2 1 + 13α 2

Fedge =

1 49α 2 . 2 1 + 49α 2 (6.34)

It can be seen that these factors take effects in different frequency ranges corresponding to different multimodal radiation behavior of the plate. Above the critical frequency, since the acoustic wavelength is smaller than the structural wavelength, the radiation from each small subsection on the plate will not cancel each other and is making contributions independently in the total radiated sound energy. Although the corners and edges may still be less efficient due to inertial flow effects if the acoustic wavelength is greater than the thickness of the plate, the effect on the total radiation is limited and therefore may be ignored. As a result, the modal averaged radiation efficiency of an unbaffled plate is still unity above the critical frequency. Figure 6.7 compares the modal averaged radiation efficiencies of two plates calculated analytically and by the BEM. The BEM results were also averaged over three excitation locations. It can be seen that Equation 6.32 gives a good estimation for the sound radiation from the piston mode. In the multimodal region, there is about 3 dB difference in the sound radiation efficiency between simply supported and clamped boundary conditions below the critical frequency. Again, Equation 6.33 is valid in the range f > [c02 /(2S f c )][P 2 /(8S) + 3l y /(2l x ) − 1] for a rectangular plate.


DK3193

noise

November 4, 2005

14:37


141

10


1

0.1

0.01 BEM, Simply-Supported Plate BEM, Clamped Plate

0.001

Equation 6.32 0.0001 100

1000 Frequency (Hz)

10000

(a) 10 BEM, Simply-Supported Plate BEM, Clamped Plate


1

Equation 6.33 Equation 6.32

0.1 0.01 0.001 0.0001 0.00001 10

1000

100

10000

Frequency (Hz) (b)

Figure 6.7 Modal averaged radiation efficiencies of an unbaffled plate: (a) 0.17 m×0.19 m×0.01 m aluminum plate, 9 × 10 elements in the BEM model; (b) 0.6 m×0.7 m×0.005 m steel plate, 40 × 46 elements in the BEM model.

6.3

Infinitely long cylindrical radiator

Cylindrical shells are one of the most widely used structures in industries, for example, the casing of electric machines, and pipes used in various constructions. Due to the geometrical curvature, the vibration and the sound radiation is


DK3193

noise

November 4, 2005

14:37

142


of different behaviors from that of plane structures. Acoustically, an elementary cylindrical radiator is an infinitely long circular cylindrical shell. By employing the cylindrical coordinates (Figure 6.8), the acoustic wave equation can be written as (omitting the time component e jωt , without loss of generality), ∂ 2 p(r, θ, z) 1 ∂ p(r, θ, z) 1 ∂ 2 p(r, θ, z) ∂ 2 p(r, θ, z) + + + + k02 p(r, θ, z) = 0. ∂r 2 r ∂r r2 ∂θ 2 ∂z 2 (6.35) The general solution of this equation is [160] p(r, θ, z) = p0 (Ae jkz z + Be− jkz z ) cos mθ Hm(2) (kr r )

(6.36)

where p0 , A, and B are constants that are determined by the acoustic boundary conditions on the cylindrical shell surface, m is the circumferential modal number of the cylindrical shell, Hm(2) (·) is the second kind of Hankel function of the m th order, kr and k z are the two components of the acoustic wavenumber k0 in the r and the z directions respectively, i.e. k02 = kr2 + k z2 . Assume that the vibration velocity distribution on the surface of the infinite length shell (r = a) is taking the form as, v(θ, z) = v 0 e− jksz z cos mθ (6.37) where v 0 is the amplitude, ksz is the z component of the structural wavenumber ks , as shown in Figure 6.8. This equation represents a structural wave traveling along the +z direction of the cylindrical shell. With the acoustic boundary condition on the surface of the shell, i.e., 1 ∂ p(r, θ, z) v(θ, z) = − jωρ0 ∂r r =a

ksz

ks

ksq = n/a

rr z

aa

q

kz k0 kr

Figure 6.8 Cylindrical shell coordinates, structural and acoustic wavenumbers.


DK3193

noise

November 4, 2005

14:37


143

one obtains, A = 0;

B = 1;

k z = ksz ;

p0 = −

jv 0 ρ0 c0 k0 (2)

m (kr a) kr d Hd(k r a)

.

(6.38)

The circumferential distribution taking the same form in Equations 6.36 and 6.37 is also a result of complying with the acoustic boundary condition. Thus, Equation 6.36 can be rewritten as p(r, θ, z) = −

jv 0 ρ0 c0 k0 (2)

m (kr a) kr d Hd(k r a)

cos mθ Hm(2) (kr r )e− jkz z

(6.39)

where kr = k02 − k z2 . In the far field, i.e., kr r >> 1, an asymptotic solution of the radiated sound pressure can be obtained as jv 0 ρ0 c0 k0 2 j 2 (1+ 1 ) p(r, θ, z) = − e π 2 cos mθe− jkr r e− jkz z . (6.40) (2) d Hm (kr a) π k r r kr d(kr a)

The above Equation 6.40 indicates that for a particular axial structural wavenumber ksz (= k z ), the sound radiation from an infinite length cylindrical shell only happens under the condition k0 > k z . In the range k0 < k z , the pressure is 90◦ out of phase with shell surface velocity. No sound wave is propagating into the far field and the sound pressure near the shell surface decays along the radial direction. Based on Equation 6.40, the sound power radiated into the far field per unit axial length is 2π 2π l 1 v 02 ρ0 c0 k02 = | p(r, θ, z)|2r dθdz = cos mθdθ. (2) 2 m (kr a) 0 2ρ0 c0l 0 0 πlkr3 d Hd(k r a) (6.41) The mean square velocity on the shell surface is given by, 2π l v 02 2π 1 2 2 v = |v(θ, z)| adθdz = cos mθdθ. (6.42) 4πal 0 0 4π 0 The radiation efficiency per unit length thus has a simple and closed form as,  k0 ≤ k z 0 (2) 2 σ = (6.43) m (kr a) k0 > k z .  2k02 / πakr3 d Hd(k r a) Equation 6.43 clearly indicates that an infinite length cylindrical shell does not radiate sound below a certain frequency where the acoustic wavenumber matches the axial component of the structural wavenumber. In this frequency range, the acoustic wavelength is greater than structural wavelength in the axial direction, and the sound generated from the two adjacent subsections with 180◦ out of the phase movement will interfere with each other, resulting in a complete cancelation along the axial direction of the shell. Beyond this frequency range


DK3193

noise

November 4, 2005

14:37

144


when the acoustic wavelength is smaller than the structural wavelength in the axial direction, the sound generated from adjacent subsections is independent of each other and the total sound radiation from the shell becomes more efficient. As frequency increases, an asymptotic value of unity can be expected as for a plane radiator. For simplicity, Equation 6.43 is sometimes employed to calculate modal radiation efficiencies of finite length cylindrical shells by setting k z equal to n π/l, which are the modal wavenumbers in the axial direction of a cylindrical shell of length l and with simply-supported boundary conditions at both ends, where n is the axial mode number. According to Equation 6.43, for each vibration mode, there exists a cut-off frequency, at which the corresponding acoustic wavenumber k0 is equal to k z = n π/l. Below this cut-off frequency, the modal radiation efficiency is zero. Above this frequency, the modal radiation efficiency increases as frequency until it reaches unity where the acoustic wavenumber k0 is equal to the structural wavenumber (n π/l )2 + (m /a )2 . This result is not obvious because if one uses infinite flat plates as an analogy, the modal radiation efficiencies of infinite length cylindrical shells would be expected to be zero when k0 < (n π/l )2 + (m /a )2 due to the intercell cancelation. However, for cylindrical shells, the structural wave speeds in the axial and the circumferential directions have dispersion effects, the wave in the axial direction increasing faster with frequency than that in the circumferential direction due to curvature effects below the ring frequency. Therefore, the axial wave would catch up the acoustic wave much faster than the circumferential wave as the frequency increases. Hence, there is a frequency range within which the axial component of a vibration mode is supersonic while the circumferential component is still subsonic. The modal radiation efficiency would continue to increase with frequency until both components are supersonic. The frequency at which the axial component becomes supersonic is the cut-off frequency of this mode. Generally, the cut-off frequency increases with number n. Figure 6.9 shows the radiation efficiencies for a series mode associated with a simply supported shell calculated by Equation 6.43. Particularly, if the axial vibration distribution is uniform, i.e. k z = 0, Equation 6.43 gets a simpler form as [200] σ (m , k0 a ) =

=

2 2 (2) π k0 a d Hdm(k0(ka0)a )

(6.44)

(k0 a )2 [Ym (k0 a )Jm +1 (k0 a ) − Jm (k0 a )Ym +1 (k0 a )] [m Jm (k0 a ) − (k0 a )Jm +1 (k0 a )]2 + [mYm (k0 a ) − (k0 a )Ym +1 (k0 a )]2

where Jm (k0 a ) and Jm +1 (k0 a ) are the first kind Bessel functions of the m th , (m + 1)th order respectively, Ym (k0 a ) and Ym +1 (k0 a ) are the second kind of Bessel functions (Neumann functions) of the m th , (m + 1)th order, respectively. The sound radiation efficiencies of various mode number m vs. k0 a are plotted in Figure 6.10. Equation 6.44 allows for a simple calculation of the sound radiation efficiency of the pure circumferential modes m = 0, 1, 2, 3, . . ..


DK3193

noise

November 4, 2005

14:37


145


10

1

0.1 Mode (1, 2) 0.01

Mode (2, 2) Mode (2, 1)

0.001

Mode (3, 2) Mode (2, 3)

0.0001 0

0.5

1

1.5

2

k0 /ks(m, n)

Figure 6.9 Modal radiation efficiencies of a circular cylindrical shell calculated by the infinite length model l/a = 3.

6.4

Finite length cylindrical radiator

In practice most circular cylindrical structures are finite in length. As the length becomes shorter, it can be expected that the end effects would become more important, and the results obtained by using an infinite length model could be in severe error. By reviewing the literature, the techniques used in the analysis could be categorized into two groups, namely the statistical approach and the deterministic analysis. A statistical approach [150, 198] is employed when the resonant frequencies are densely packed in a frequency band, and the acoustic properties are lumped together so that individual frequency analysis becomes impractical. Obviously, this method is only valid at high frequencies or for relatively thin shells with high modal densities. The most successful work about a cylindrical source was carried out by Szechenyi [198] who developed a set of general, but approximate equations for predicting the radiation efficiencies of cylindrical shells having large length to thickness and radius to thickness ratios within a given frequency bandwidth. These results have been verified by experiments except for low frequencies [198]. In deterministic analysis, specific vibration velocity distributions and their acoustic response, which could be affected by the geometry and the boundary conditions of a cylindrical shell at each frequency, are considered. Since at low frequencies where the modal density is low, the behavior of each individual mode becomes important, the corresponding modal averaged radiation efficiency may have different characteristics from that at high frequencies. Various studies using deterministic analysis have been reported. The early contribution to the acoustic


DK3193

noise

November 4, 2005

14:37

146



2

1.5

1

0.5

m=1

m=2

m=4

m=5

m=3

0 0

1

2

3

4

5

6

7

8

9

10

k0a

Figure 6.10 Radiation efficiency factor curves according to Equation 6.44 for a cylindrical radiator and m = 0, 1, 2, . . . 5.

radiation from finite length cylindrical shells was made by Williams [238] who analyzed the radiation of finite length cylinders with a uniform radial vibration velocity profile and compared the sound field with that of the infinite length model. It was found that infinite length models may overestimate the acoustic radiation from finite length cylinders, and the error would increase as the length of the cylinder becomes shorter. Then, Schenck (1968) presented a surface Helmholtz integral formulation for obtaining approximate solutions of acoustic radiation problems for an arbitrary surface, which improved the accuracy of Williams, results in the axial direction. In order to take the reaction of the fluid inside the cylindrical shell into account, Sandman [182] examined the model of a baffled finite length cylindrical shell. It was found that the model is a reasonable approximation in determining the acoustic radiation from the surface of finite length cylindrical shells, and is applicable to arbitrary radial vibration distribution. Based on this model, Stepanishen [192] discussed the radiation impedance of different vibration modes of a finite length cylindrical shell, and Zhu [255] examined the “relative sound intensities” for different vibration modes. Also, Stepanishen [193] and Laulagnet et al. [134] investigated the effects of fluid on the acoustic radiation of finite length cylindrical shells. In this section, the sound radiation from acoustically thin shells is first introduced briefly. Then the baffled finite length cylindrical shell model is employed to calculate the modal radiation efficiencies and modal averaged radiation efficiencies for different boundary conditions. The limitation of the infinite length model and the effects of boundary conditions on the modal radiation efficiencies of finite length cylindrical shells are discussed last.


DK3193

noise

November 4, 2005

14:37


6.4.1

147

Acoustically thin shells

The acoustically thick and thin shells are distinguished in terms of shell curvature effects on the acoustic radiation behavior. For cylindrical shells, the ring frequency is an important parameter used to indicate the frequency range in which curvature effects are important. The ring frequency fring is defined for the condition when the wavelength of extensional waves in the shell is equal to the shell circumference (compare Equation 5.21), E 1 fring = (6.45) 2πa ρ(1 − ν 2 ) where a is the radius of the cylindrical shell, ρ is the density of the material, E is the Young’s modulus, and ν is the Poisson ratio. Above the ring frequency, the structural wavelength is so short that the structural wave propagation is very much controlled by the local stiffness. The curvature effects of the shell disappear, and the structure will behave like a plate. The sound radiation, as a result, will be governed by the compliances between the total structural wavenumber ks and the acoustic wavenumber k0 , as discussed in Section 6.2. In this frequency range, the critical frequency defined for the plate f c (Equation 6.14) is of physical significance because of no dispersion effects for the structural wavenumbers. Below the ring frequency, the curvature changes the stiffness in the axial direction, making the structural wave speed in this direction faster than that in the circumferential direction. Then the sound radiation is primarily determined by the behavior of each individual mode in the total response. The physical significance of the critical frequency defined for the plate f c becomes questionable, as will be seen in Section 6.4.2. Given these two different acoustical features, the cylindrical shells are consequently defined as acoustically thin for fring / f c < 1, and thick for fring / f c > 1 [89]. However, it should be emphasized that a shell can be geometrically thin (thickness h >>radius a) but still acoustically thick. For example, for an aluminum (Young’s modulus, E = 71 × 109 Pa, density ρ = 2700kg/m3 and the Poisson ratio, ν = 0.33) or steel (E = 210 × 109 Pa, ρ = 7850 kg/m3 and ν = 0.3) circular cylindrical shell, by using Equations 6.42 and 6.14, one obtains, fring h ≈ 67 . fc a

(6.46)

If a / h is smaller than 67, the shell would be considered to be acoustically thick but still geometrically thin. In practice, a typical example of acoustically thin shells is an airplane fuselage. For acoustically thin cylindrical shells, because of the high modal density, statistical analysis can be employed [150, 198]. By measuring the area ratio of the supersonic and subsonic regions that vibration modes possibly fall into in the wavenumber contour map, Szechenyi [198] generated a family of curves of the sound radiation efficiency for various cylindrical shell configurations, as shown in Figure 6.11. It is seen that for acoustically thin shells, that is, when


DK3193

noise

November 4, 2005

14:37

148

Noise of Polyphase Electric Motors 0 −2 Radiation Efficiency, 10 log10σ

−4

0.60

−6

0.55 0.50

−8

0.52

−10

fring /fc

2.0 5 1.1 .0 1 70 0. 0 0.5 0 0.3 0 0.2 .15 0 0 0.1 5 0.0

−12 −14 −16 −18 −20

0.20

0.15 0.10

0.51

2

0.0

−22 −24 0.02 0.04

0.30

0.1

0.4

1.0

2.0

4.0

10

f/fring

Figure 6.11 Modal averaged radiation efficiency of finite length acoustically thin cylindrical shells [198]. f c > fring , the radiation efficiency of cylindrical shells has three distinguished features corresponding to three frequency ranges. Below the ring frequency, the radiation efficiency increases at a rate of 3 to 6 dB per octave to a maximum at the ring frequency. Above the ring frequency, the curvature effects are no longer important and the cylindrical shells will vibrate like flat plates. Therefore, in the frequency range between f ring and f c , the radiation efficiency would first decrease, then increase, as the frequency approaches f c in a manner similar to flat-plate radiation. Above the critical frequency, the radiation efficiency then maintains a value of unity. In the analysis, Szechenyi [198] concluded that below the ring frequency, the modal averaged radiation efficiency is only a function of h/a, independent of the length and the boundary conditions. The influence of the length and the boundary conditions for thin shells is limited to the region f ring < f < f c being in a way similar to that in flat-plate radiation. Although analytical expressions could not be given for the full frequency range, there are specific cases where analytical calculations are possible.  √ f  0.346 1 − ν 2 ffc ≤ 0.48  fring   √   √   0.241 1 − ν 2 f fring 0.48 < f ≤ 0.83 fc fring σ = 2 1/2 1/2  (ha/l ) ( f / f ) f fc c  > 1 and fring > 10   1.86π(1−ν 2 )1/4 (1− f / f c )3/2 fring    f 1 >1 fc


(6.47)

DK3193

noise

November 4, 2005

14:37


6.4.2

149

Acoustically thick shells

According to Equation 6.46, most practical cylindrical shells encountered in industries are acoustically thick, for which Szechenyi’s results cannot be applied. For acoustically thick cylindrical shells, the physical significance of f c is complex because curvature effects play an important role in determining the flexural wave speed and the acoustic radiation behavior. By analyzing the sound transmission through the pipe walls, Holmer et al. [89] showed that the radiation efficiency of an acoustically thick cylindrical shell under acoustic excitation normally increases smoothly with frequency until it reaches a constant value of about unity at high frequencies, and the coupling between the acoustic modes and the structural modes has to be considered. However, as pointed out by Fahy [55], the acoustic behavior of thick shells is strongly influenced by the nature of the excitation. For example, mechanical excitation of pipes can produce quite different radiation efficiencies from those associated with the excitation by fluid flow or sound waves in the internal fluid. The acoustic radiation of a cylindrical shell subjected to a point force excitation was studied by Laulagnet and Guyader [134] theoretically. They found that for light fluid (in which the ratio of its specific impedance to the angular frequency is negligible compared to the shell mass per unit area), the coupling between the acoustic modes and the structural modes is negligible. As an example, Figure 6.12a shows the modal averaged radiation efficiencies of an acoustically thick steel cylindrical shell, 200 mm long, 1.6 mm thick and 63.5 mm in radius, with three different boundary conditions under mechanical point excitations, calculated by the BEM. The critical and ring frequencies of this cylindrical shell are 7392 Hz and 13,190 Hz, respectively. The number of nodes and elements of the BEM model is 2480 and 2400, respectively, so that the number of elements per acoustic wavelength below 8 kHz is greater than 6. A point force was applied to the surface of the shell in the radial direction at a point 6.6 mm away from one end. It is seen that depending on the boundary conditions the radiation efficiency reaches unity at a frequency much lower than the critical frequency f c . On the other hand, Szechenyi’s results [198] for acoustically thin shells are independent of the boundary conditions. Although Szechenyi’s results display a general trend similar to the BEM results, there are significant quantitative differences. Figure 6.12b shows the results for another two steel cylindrical shells, each 3 mm thick, and 19.5 mm in radius, but 20 mm and 60 mm in length respectively, with free-free boundary condition under a point excitation. The critical and the ring frequencies for these two shells are 4036 Hz and 41.9 kHz, respectively. The number of elements for the two models is 1000 and 3000, respectively. It can be clearly seen that for a thick shell, different lengths could result in significant differences in acoustic radiation. In contrast, Szechenyi’s results, which are only valid for acoustically thin shells, are independent of the length. In order to facilitate the discussions on sound radiation from each individual vibration mode in details, a steel acoustically thick circular cylindrical shell,


DK3193

noise

November 4, 2005

14:37

150


Modal Averaged Radiation Efficiency

100

10

1 fc

0.1 Simply-Supported

0.01

Clamped-Clamped Free-Free

0.001

Szechenyi’s result 0.0001

0

1000

2000

3000 4000 5000 Frequency (Hz)

6000

7000

8000

(a)

Modal Averaged Radiation Efficiency

10

1

0.1

0.01 BEM results, l = 20 mm 0.001

BEM results, l = 60 mm Szechenyi’s result

0.0001

0

1000

2000 3000 Frequency (Hz)

4000

5000

(b)

Figure 6.12 Modal averaged radiation efficiency of acoustically thick cylindrical shells calculated by BEM: (a) a = 63.5 mm, h = 1.6 mm, l = 200 mm; (b) a = 19.5 mm, h = 3 mm. ◦◦◦ Equation 6.70, l = 20 mm; Equation 6.70 l = 60 mm [227].


DK3193

noise

November 4, 2005

14:37


151

200 mm long, 1.6 mm thick, and 63.5 mm in radius, is chosen as an example. A simplified frequency-wavenumber characteristic equation for a cylindrical shell with neglecting in-plane deflections [50] is given by the following equation [169]: ω2 =

4 D K (1 − ν 2 ) ksz 2 · (ksz + ks2θ )2 + · 2 + k 2 )2 ρh ρha 2 (ksz sθ

(6.48)

where D = Eh 3 /[12(1 − ν 2 )], K = Eh/(1 − ν 2 ). It was shown [226] that this relationship is independent of the boundary conditions. For finite length cylindrical shells with different boundary conditions at both ends, the axial component of the structural wavenumber ksz takes different forms as discussed by Wang and Lai [226]. Based on Equation 6.48, a series of wavenumber curves associated with different frequencies obtained are presented in Figure 6.12. It can be clearly seen from Figure 6.13a that, unlike flat plates, the vibration modes of shells having

2 + k2 ; the same natural frequencies could have different wavenumbers ks = ksz sθ e.g., the wavenumbers at 13 kHz in the graph. Only above the ring frequency would the wavenumbers at the same frequency approach the same value. This phenomenon is due to the curvature effects which modify the wave speed in the axial direction below the ring frequency. To illustrate curvature effects on sound radiation, three structural wavenumber curves and three corresponding acoustic wavenumber curves are plotted in Figure 6.13b. From the acoustic point of view [55], the vibration modes could be categorized to be acoustically fast modes and acoustically slow modes, sometimes called supersonic and subsonic modes. Acoustically fast modes refer to those of which the structural wavenumbers are smaller than the corresponding acoustic wavenumbers, and the modal radiation efficiencies of these modes are unity. Acoustically slow modes are inefficient in acoustic radiation because structural wavenumbers are greater than the corresponding acoustic wavenumbers, which would lead to some cancelation in the radiation. For isotropic and flat plates [55, 37], the unique demarcation for these two cases is the critical frequency. However, for cylindrical shells, from Figure 6.13b, it can be seen that, below the critical frequency of 7.33 kHz, the structural wavenumber curve will always intersect the acoustic wavenumber curve, such as point A, at a given frequency. This point of intersection in the wavenumber domain changes as the frequency changes. When the frequency is the critical frequency, the curves are tangent to each other as shown by point B in Figure 6.13b. When the frequency is greater than the critical frequency, the two curves will no longer meet. This result indicates that at any frequencies below the critical frequency, acoustically fast modes and slow modes could exist simultaneously, and the demarcation depends on the frequency. Therefore, for cylindrical shells, it is impossible to define a unique “critical frequency” for describing the acoustic properties as for flat plates. Figure 6.13b shows that the critical frequency defined for plates indicates the condition for all possible vibration modes of cylindrical shells to be supersonic. Thus, the radiation efficiency of cylindrical shells should be unity above the critical frequency because the modal radiation efficiencies of all modes are unity.


DK3193

noise

November 4, 2005

14:37

152

Noise of Polyphase Electric Motors 250 f = 23 kHz ks

200

19 kHz 16 kHz 150 kzn

13 kHz (ring frequency)

100

ks

9 kHz ks

6 kHz

50

ks

3 kHz 0

0

50

100

200

150

250

kθm (a) 200 Acoustic Wavenumber f = 9 kHz 7.33 kHz (critical frequency)

150

kzn, k

Structural Wavenumber

100 9 kHz

5 kHz

f = 7.33 kHz

50

A 5 kHz

0

B 0

50

100 kqm, k

150

200

(b)

Figure 6.13 Wavenumber diagram for a cylindrical shell: kθm — circumferential structural wavenumber; k zn — axial structural wavenumber; k — acoustic wavenumber [227].


DK3193

noise

November 4, 2005

14:37


153

Below the critical frequency, since both supersonic and subsonic modes could exist simultaneously, the overall radiation efficiency could depend on the number and the types of modes dominating the vibration response. For example, if the mode is supersonic, the modal radiation efficiency is unity, and if it is subsonic, the modal radiation efficiency is less than one. Three points associated with the determination of the radiation efficiency below the critical frequency can be concluded here. First, the demarcation for each mode, i.e. whether the mode is supersonic or not, has to be determined. Second, the modal averaged radiation efficiency could be unity if supersonic modes dominate the response. Third, for a given cylindrical shell, the modal averaged radiation efficiency is dependent on the excitation and boundary conditions because the excitation determines the modal participation amplitudes and any changes of boundary conditions could make some supersonic modes subsonic or vice versa. By comparison, for flat plates, the boundary conditions are not likely to fundamentally affect the contribution of the vibration modes because statistically, the dominated modes occurring below the critical frequency are basically subsonic and the modes above the critical frequency are always supersonic [55, 37]. Therefore, it can be expected that the variation of the modal averaged radiation efficiency of cylindrical shells due to the change of boundary conditions could be substantial compared with flat plates. To explain the results of sound radiation efficiencies shown in Figure 6.12, the behavior of each individul mode of the cylindrical shell has to be examined. The natural frequencies of the cylindrical shell can be calculated based on the method described in [226]. Corresponding to each natural frequency, there must exist a structural wavenumber curve to which the vibration mode belongs and an acoustic wavenumber curve of the same frequency, as shown in Figure 6.13b. Note that at each point of the intersection of the two curves, the structural wavenumber is equal to the acoustic wavenumber. So one can compare the structural wavenumber of this mode ks with the acoustic wavenumber k0 to determine whether this mode is supersonic or not. For this purpose, an index is defined as follows, ωmn 2 L (m , n ) = k0 − ks = − ksz (n ) − (m /a )2 . (6.49) c0 Obviously, this index depends on the natural frequency ωmn and the structural wavenumbers of the vibration mode. If L = 0, the acoustic wavenumber is greater than the structural wavenumber, indicating that the mode is supersonic, and if L < 0, the mode is subsonic. In Figure 6.14, L of all the vibration modes associated with three different boundary conditions (simply supported, free, and clamped) is plotted against their own natural frequencies. It can be seen that the critical frequency ( f c ) of the equivalent flat plate is an indicator showing whether all the vibration modes of the cylindrical shell are supersonic or not. Below the critical frequency, corresponding to each supersonic mode in Figure 6.14, there is a peak in the modal averaged radiation efficiency in Figure 6.12a, for example, the peaks around 2300 Hz for simply supported and clamped conditions, 2800 Hz for free condition. According to Figure 6.14, although subsonic modes exist below the critical frequency, the number of supersonic modes increases as the frequency


DK3193

noise

November 4, 2005

14:37

154

Noise of Polyphase Electric Motors 100 80

DL (m−1)

60 40 20 0 −20

fc

−40 −60

0

1000 2000 3000 4000 5000 6000 7000 8000 fmn(Hz)

Figure 6.14 The index L of each vibration mode of the cylindrical shell: + + + simply supported; • • • free–free; ✷✷✷ clamped–clamped [227]. increases, so that the modal averaged radiation efficiency could reach unity at a frequency much lower than the critical frequency, as shown in Figure 6.12a.

6.4.3

Modal radiation efficiencies of acoustically thick shells

For acoustically thick shells, the modal radiation behavior of a particular mode could dominate the overall performance. The estimation of modal radiation efficiencies of cylindrical shells thus becomes necessary. Although the infinite length model could be employed for this purpose, errors will be introduced due to ignoring the end effects. To take the finite length into account, a model of an infinite length circular cylindrical shell with finite length vibration distribution, as shown in Figure 6.15, is adopted here. In the model, the sound field inside the cylindrical shell and the effects of internal sound field on the structural vibration and the outer x r q

a

z

y l

Figure 6.15 Finite length cylindrical shell model.


DK3193

noise

November 4, 2005

14:37


155

sound field are not considered. For cylindrical shell structures, this may be acceptable when the shell is excited mechanically and the fluid inside the cylindrical shell is light, such as air. As a result, the surface normal displacement of each vibration mode can be written as, (omitting the time component e jωt )

u mn (θ, z) =

u 0 γn (z) cos mθ 0 < z < l 0 l fring (the ring frequency). For a shell structure with a large radius such that the corresponding ring frequency fring is lower than the frequency range of interest, the structure can be treated directly as a flat plate in√SEA. Particularly, if the cylindrical shell is long enough so that l 2πa a/ h, the shell is basically in bending motion, regardless √ of the boundary conditions at both ends, in the frequency range below fl (= 3/5 fring h/a). In this case the


DK3193

noise

October 24, 2005

19:9

278


modal density of the cylindrical shell can be calculated by Equation 10.31 with the appropriate radius of gyration κ for the shell cross section [228]. For an orthotropic circular cylindrical shell, of which the stiffness along the axial and the circumferential directions are different, the modal density might be approximately calculated by Equation 10.36 with the ring frequency fring replaced with

(1) (2) (1) (2) fring fring , where fring = cl(1) /(2πa), and fring = cl(2) /(2πa)

with cl(1) and cl(2) being the longitudinal wave speeds in the two orthogonal directions respectively. Modal overlap Mi The modal overlap of a subsystem is defined as Mi = ωηi n i .

(10.37)

Physically, this parameter denotes the average number of modes lying within a typical modal half power bandwidth. A high modal overlap factor means that neighboring resonance peaks largely merge together, giving a high probability for a system to be characterized as resonant, which is essential to SEA applications. Although the modal overlap is a parameter that might not directly appear in the SEA equations, it is often used to justify the appropriateness of employing the wave approach for evaluating coupling loss factors within the system, and also might serve as one of the indicators showing whether the resonance occurring in a subsystem reaches a level for SEA to be able to give reasonable solutions. Generally, it is assumed that a system with its subsystem modal overlap factors less than unity is not amenable to an SEA approach. However, one should be clear that the actual number of modes being excited within a system depends on the initial condition of the excitations, not the modal overlap. The modal overlap only gives the maximum number of modes (in the average sense) that are possibly excited.

10.3.4

Limitations of SEA

As seen from the discussions above, several assumptions that are critical to the validity of SEA foundations (Equation 10.21) were made inevitably when extending the results for simple coupled oscillators to coupled multimodal systems. Due to the complexities and varieties of structure assemblies, these hypotheses were expressed quite generically in the modal space with no explicit math descriptions being given. As a result, the implications and relationships of these hypotheses to the apparent vibro-acoustic behavior of continuous structures are not very clear until now and are still today’s hot topics of interest to engineers and researchers. Currently, the most widely recognized condition, though not clearly defined, for a valid SEA analysis is that the coupling between subsystems should be weak. A criterion that was given in the early days of the SEA development is that the coupling would be weak if the ratio of the coupling loss factor to the internal loss factor is substantially less than unity, i.e., ηi j ηi , η j [189]. Later, the modal overlap was introduced as an indicator of the coupling strength between


DK3193

noise

October 24, 2005

19:9

Statistical Energy Analysis

279

subsystems. For a plate system, it was suggested that a modal overlap greater than unity and more than 5 modes in the frequency band are the conditions for the SEA theory to be appropriate to apply [56]. Recently, Wester and Mace [237] argued that the coupling strength between two plates (a, and b) may be described as weak if T02 (ka d)(kb d)(Ma Mb )/(4π 2 ) 1, where T0 is the transmission coefficient of the junction for a normally incident plane wave; d is the junction length; ka , kb are the bending wavenumbers; and Ma , Mb are the modal overlaps of the two plates respectively. However, various studies showed that all these criteria appeared to be the necessary, but not the sufficient conditions for a valid SEA analysis. Through the study on the spatial and frequency averaged Green functions of coupled systems, Langley [129] even argued that the weak coupling condition does not guarantee the validity of a classical SEA analysis, in which the coupling only exists between directly coupled subsystems, unless the wave field in each subsystem is also reverberant. Given as an example, a serially coupled five-plate structure, each plate coupled to adjacent ones with right angles by exhibiting a line junction is shown in Figure 10.10. The dimensions of the plates are, from left to right, 0.7 m × 0.6 m × 0.001m, 0.3 m × 0.6 m × 0.001m, 0.8 m × 0.6 m × 0.001m, 0.5 m × 0.6 m × 0.001m, and 0.4 m × 0.6 m × 0.001 m, respectively. For the steel material and with the internal damping loss factors assumed to be 0.05, the modal overlaps of all five plates were found to exceed unity above 400 Hz and the coupling strengths defined in [237] for all four junctions were less than unity above 500 Hz. Besides, above 500 Hz, the modal counts within each 1/3 octave bandwidth are greater than 5 for all the plates. All these indicate that the coupling between plates is weak in general at high frequencies. A classical SEA model which consists of five subsystems, each for one plate, with only direct coupling paths considered are, therefore, straightfoward and can be generated without much difficulty. In order to provide a benchmark result, the finite element method was employed in the analysis. The number of nodes for each plate is 2793, 1225, 3185, 2009, and 1617 correspondingly. In order to generate incoherent modal responses in the plates, a point force was applied separately in the normal direction at five positions chosen randomly over plate 1 and the spatial and frequency averaged vibration response of each plate was further averaged for the five different excitation positions. As a result, the upper limit of the valid frequency range

Plate 5

Plate 1

Plate 2

Plate 3

Plate 4

Figure 10.10 FEM model of the five-plate structure.


DK3193

noise

October 24, 2005

19:9

280


E2 /E1

Energy Ratio (dB)

−10

−20

−30

E5/E1 FEM

−40

- - - - SEA EFM

−50 100

1000 Frequency (Hz)

10000

Figure 10.11 Comparisons of energy ratios of plates calculated by SEA (- - -), FEM (—), and energy flow method (EFM) (•).

for this model was around 3000 Hz. The response of each plate, with only plate 1 being excited, was calculated by FEM and compared with the outcome from the classical SEA model in Figure 10.11. It is seen that although the classical SEA model predicts the response well for plate 2, which is adjacent to the excited plate 1, it severely underestimates the response of plate 5, which is further away from the excited plate even in the frequency range that the suggested weak coupling conditions were well satisfied [251]. The reason is that this particular structure construction acts as a spatial filter that favors a particular set of waves or modes in the energy transfer. A result of this is that the modal energies tend not to be evenly distributed among the modes within the subsystem as the energy transmits away from the excitation. In other words, the wave fields in those subsystems are not diffuse or reverberant enough to comply with the classical SEA requirements. In this case, the indirect power flows are expected to exist in the system [129]. At present the characteristics of the indirect power flow within the system is not well understood yet and the estimation of indirect coupling loss factors is difficult. The failure of the classical SEA in this example reveals that, for a particular structure, we still generally lack the knowledge and effective means to pre-justify which of the assumptions that are necessary for a classical SEA analysis are reasonable, which of them may be violated and if so, how much error it may result in the predictions. Nevertheless, by reviewing the research development, Langley [132] summarized the basic assumptions that appear to be critical to the applications of


DK3193

noise

October 24, 2005

19:9


281

the classical SEA theory and the conditions associated with the apparent vibroacoustic behaviors that may help to promote the assumptions to be real. The two basic assumptions were, • there must be an adequate number of modes within a subsystem in the frequency band. • there must be equipartition of energy among the modes within a subsystem, and the modes must be uncorrelated. This statement is equivalent to that the response in each subsystem must be a diffuse/reverberant wave field in which the waves are uncorrelated and carry almost the same amount of energy in each direction. The conditions have been described as, • high frequency excitations to generally ensure that more modes exist in the frequency band; • distributed excitation which is sometimes stated as “rain on the roof” excitation to ensure the equipartition of modal energy and the uncorrelated modal response; • excitations in multi-subsystems to promote diffuse wave fields in all subsystems; • an irregularly-connected pattern between subsystems so that each subsystem receives energy from different directions, promoting a diffuse wave field; • weak couplings between subsystems to promote uncorrelated behavior between coupled subsystems (in contrast to the “strong coupling” for which the modes in subsystems tend to be correlated, that is, the “global” mode behavior); • high modal overlaps to promote the multimodal behavior and the equipartition of modal energy. In practice, one may expect that one or more of the conditions may not be fulfilled completely in a particular application. However, it should be noted that it is the two assumptions rather than the conditions listed here that are essential to a valid classical SEA analysis.

10.4

Experimental SEA

For various complex structures, a thorough analytical SEA analysis on the power flow within the system and the vibro-acoustic responses due to specific excitations is not always feasible. This is not only because of the difficulties involved in determining the SEA parameters, such as the coupling loss factors and internal loss factors, by purely analytical approaches, but also due mainly to the lack of knowledge about the actual energy exchanging and sharing scheme within the system and their compliances with the classical SEA hypothesis or requirements. This gives little knowledge about confidence levels of analytical predictions, especially for a complex structure. The experimental SEA was first developed to provide a means to acquire the parameters necessary for an accurate SEA model in situ. The


DK3193

noise

October 24, 2005

19:9

282


measured parameters such as coupling loss factors and internal loss factors not only incorporate the uncertainties/variations that are neglected or impossible to consider in the analytical approaches, but also accommodate the effects associated with the particular substructuring scheme chosen in the test. As the mechanisms of indirect coupling loss factors and negative loss factors found in the measurements being revealed in the investigations, the technique has been extended beyond the original scope and playing an important role in studying the actual power flow behaviors within complex systems. A more general notion, namely the energy flow analysis which can be pursued either experimentally or analytically/numerically, has been developed based on, but not necessarily limited to the classical SEA framework. Nowadays, the experimental SEA or energy flow analysis has been recognized as an efficient tool to diagnose noise/vibration problems in regard to power flow paths within the system and to explore various engineering solution strategies by extrapolating the model parameters. In this section, the basic experimental SEA principle is discussed first, then recent developments within the experimental SEA framework are introduced briefly.

10.4.1

General theory

The basic idea of experimental SEA is, by measuring the vibrational energy levels of various subsystems for various power input configurations, to inversely yield the SEA parameter matrix. Since in SEA all the relevant parameters are frequency band and spatially averaged, the corresponding measurements are relatively robust but do not require detailed descriptions of the system. For a system with N subsystems, if only subsystem 1 is excited, a SEA power balance equation like Equation 10.22 can be written as,      E 11 η11 −η21 · · · −η N 1 P1  0   −η12 η22 · · · −η N 2   E 21       (10.38)  ..  = ω   .   .   · · · · · · . . . · · ·   ..  −η1N −η2N · · · η N N 0 E N1 where E i1 is the average total vibrational energy of subsystem i when subsystem 1 is excited. If each of the N subsystems is excited in turn, N sets of SEA power balance equations in the form of Equation 10.38 can be established. Assuming that all of the coupling and internal loss factors are independent of external excitations, that is, the coupling and internal loss factors keep unchanged when the excitations change from one subsystem to another. The combined matrix equation for coupling and internal loss factors can be written as,     P1 0 0 · · · 0 η11 −η21 −η31 · · · −η N 1  −η12 η22 −η32 · · · −η N 2   0 P2 0 · · · 0     1  −η13 η23 η33 · · · −η N 3      =  0 0 P3 · · · 0  ·  .. ..  ω  .. .. .. . . ..  .. . . ..  .  . . . . .  . .  . . −η1N −η2N −η3N · · · η N N


0 0 0 ··· PN

DK3193

noise

October 24, 2005

19:9


283



E 11 E 12  E 21 E 22   ·  E 31 E 32  .. ..  . . E N1 E N2

E 13 E 23 E 33 .. . E N3

−1 · · · E 1N · · · E 2N   · · · E 3N   ..  .. . .  ··· ENN

(10.39)

where E i j represents the average total vibrational energy of subsystem i when subsystem j is excited. It is seen that with all the subsystems being excited in turn, all the internal and the coupling loss factors can be solved from Equation 10.39. However, it was found that the solution of Equation 10.39 is very sensitive to a small variation (for example, less than 10%) in measured energy levels, making the energy matrix easily ill-conditioned. One of the consequences is that Equation 10.39 may yield negative coupling and internal loss factors which are not physically meaningful. To resolve this problem, Lalor [128] has proposed to split the calculation of coupling loss factors and internal loss factors into separate matrices. Thus, from Equation 10.39, N sets of matrix equations for the coupling loss factors ηri (relating to the ith subsystem) can be obtained as, −1  E 11  − EE 1i · · · EEr 1 − EEri · · · EEN 1 − EEN i   1 E i1 ii i1 ii η1i  i1 . ii  . .    . .. .. ..  ..    .. .       .. ..  .. ηri = P i  E ri E rr · .    . . −   ωE   . E ir E rr ii   ..     . .. .. . ..   .. . . .   ηN i 1 E 1N E 1i Er N E ri ENN E Ni · · · · · · − − − E E E E E E iN

ii

iN

ii

iN

ii

i = 1, 2, · · · , N

(10.40)

In this equation, the energy matrix has a smaller dimension, (N − 1) × (N − 1) for N > 3, than that in Equation 10.39, and therefore tends to be better conditioned. The internal loss factors ηi can be obtained by substituting Equation 10.40 into Equation 10.39 to give, E ji Pi − ηi j + η ji . ωE ii E ii j=1 j=1 N

ηi =

j= i

N

(10.41)

j= i

In fact, internal loss factors can be evaluated directly by considering the power balance of the whole system, that is, the input power to any one subsystem must be equal to the total power dissipated in all the subsystems. Thus, we have 

η1  η2   ..  .





E 11 E 21   E 12 E 22 1   =   ω  ··· ··· ηN E 1N E 2N


−1   · · · E N1 P1   · · · E N2    P2     . .. . · · ·   ..  ··· ENN PN

(10.42)

DK3193

noise

October 24, 2005

19:9

284


However, the use of Equation 10.40 still cannot guarantee that all the coupling loss factors obtained are positive. Note that in the energy matrix of Equation 10.40, the energy ratio E rr /E lr , l = r will be much greater than unity since the energy of a directly driven subsystem must always be greater than that of a subsystem either directly or indirectly connected to it if the weak coupling condition is satisfied (the weak coupling condition in here refers to ηi , η j ηi j ). Lalor [128] has simplified Equation 10.40 further by assuming that the diagonal terms in the energy matrix are much greater than the off-diagonal terms such that the off-diagonal terms could be neglected. The inverse of the energy matrix can be approximately written as   E E 1i E ri E N1 E Ni Er1 11 − − − · · · · · · E i1 E ii E i1 E ii  E i1 . E ii  .. ..   . .   . .   .. .   E ri E rr .  → . . −E   E ir rr   .. .. ..   . . .   E 1N − EE 1i · · · EEr N − EEri · · · EEN N − EEN i E iN

ii

iN

      →    

ii

iN

 0 ··· ··· 0  .  0 .. 0   .. ..  E rr . . .   E ir  .. . 0  0  ENN 0 ··· ··· 0 E

ii

E 11 E i1

(10.43)

iN

As a result, the coupling loss factor between any of the two subsystems has a simple form as 1 E ji P j . (10.44) ηi j ≈ ω E ii E j j Since all the terms in this equation must be positive, the coupling loss factor obtained from Equation 10.44 is always positive, thus eliminating the negative coupling loss factors completely. Another benefit of using Equation 10.44 is that it does not require inverting a large matrix, which makes the measurement even more simple because the calculation of a coupling loss factor is only related to the response of the two subsystems of interest, regardless of how many other subsystems exist within the system. However, great attention should be paid when using Equation 10.44 for a complex structure system because the assumption that the diagonal terms in the energy matrix are much greater than the off-diagonal terms may not be true for some subsystem pairs. For example, a diagonal term E rr /E lr , l = r could be smaller than an off-diagonal term E ir /E jr , i = r, j = r, i = j if subsystem j is further away from the driven subsystem r while subsystem i is close to the excitation. Though E ir , E jr and E lr all have smaller numbers compared to E rr , the


DK3193

noise

October 24, 2005

19:9


285

ratio E ir /E jr could be greater than E rr /E lr . Apparently, due to the complexities of real structures, it is rather difficult to give a full discussion on the validity of Equation 10.44. By taking a three subsystem SEA model as an example, Lalor [128] showed, however, that the assumption made here is quite reasonable and Equation 10.44 should apply with sufficient accuracy for subsystem pairs that are directly connected. In fact, it can be observed that Equation 10.44 implies that for any subsystem pairs considered, there is only one direct power flow path existing, other power flow paths related to these two subsysytems being neglected. Since Equation 10.44 is subject to the weak coupling assumption, it only applies to the subsystems that are directly connected, that is, the direct coupling loss factors. For indirect coupling loss factors, Equation 10.44 might be acceptable for the subsystem pairs separated by only one or two subsystems, depending on the coupling strength, and the damping values of the system.

10.4.2

Recent developments

Experimental SEA principles so far have inspired extensive studies and investigations that have reached the field beyond the traditional SEA territory. New findings and techniques that help to understand the fundamental pros and cons of classical SEA theory and enhance the modeling capabilities especially in the midfrequency range have been reported world widely in recent ten years. The midfrequency problems are referred to those in the frequency range beyond the upper limit of the FEM due to unacceptable computing resources required and the lower bound of the classical SEA because of its fundamental flaws as discussed above. With FEM and SEA having different concerns and strategies in extending their frequency limits into the midfrequency range, a better understanding of the characteristics of the energy flow in the mid to low frequency range is essential to impose new connotations and values onto SEA. Negative coupling loss factors Negative coupling loss factors found originally in many experimental SEA studies were believed to be related to the testing errors in energy ratios and the corresponding matrix inversion (Equation 10.39). Although using Equation 10.44 ensures positive coupling loss factors, the weak coupling assumption invoked in the approach introduces constraints and uncertainties in applying the method to strongly coupled systems and complex systems at low frequencies where the couplings tend to be strong in general. Woodhouse [243] proposed an iterative procedure to produce a symmetric matrix with positive loss factors. The approach was described later by other researchers as ‘massaging’ the system responses into the SEA framework, although fundamentally the system might not be of SEA behavior. Recent studies showed that although it is true that the derived coupling loss factors are sensitive to small variations in the responses due to matrix inversion, some aspects related to the uncertainties that appear to be fundamental to the structure or the approach itself should not be ignored. One of the arguments is that the


DK3193

noise

October 24, 2005

19:9

286


measurement, which is normally carried out on a specific structure, is only one sample of the assemblies, of which the results, from the statistical point of view, may have large deviations off the assembly mean. Given the dynamic behaviors of structures, such deviation is directly related to the modal overlap factors of the system. Greater modal overlap factors generally yield smooth, less varied solutions with relatively high confidence levels [62, 91]. At low frequencies where the modal overlap is normally low, large deviations can be expected and the negative loss factors might be observed. Moreover, based on a two-coupled-oscillators model, Sun et al. [195] investigated the relationship between the exact loss factors defined in the model and the derived loss factors in accordance with the experimental SEA principles. It was found that derived loss factors, including coupling loss factors and internal loss factors, could be different from the exact values defined in the model if there were energy dissipation mechanism in the path of power transfer, for example the damping at the interface of the junctions. In this case, the effect of the damping at the junction was distributed to subsystems as additional internal energy dissipations, thus making the derived internal loss factors as equivalent ones. This equivalency generally becomes obvious when the coupling is strong, which could trigger negative loss factors to appear at low frequencies in the derived solution. The implication of this finding for continuous structures is that the substructuring scheme chosen for a specific structure plays a critical role in deriving the negative loss factors. As discussed previously, for a practical coupled structure, how those oscillators are actually grouped and coupled, how the energy actually dissipates, and whether the substructuring scheme complies with the SEA requirements are not always clear. Although for relatively simple structures, subsystems may be reasonably determined in terms of the physical boundaries. It is very likely for a complex system that the substructuring scheme employed based on one’s prior experience does not describe the nature of the power flow topology in the structure adequately. Also in most cases, simplifications may have to be made in the SEA model by ignoring some structural details or components, which, however, may dissipate vibrational energies under specific conditions. By applying this inappropriate model, the behavior of the structure is forced to fit the corresponding energy balance equations. All the derived parameters are therefore equivalent ones which may no longer have actual physical significances. The negative loss factors, which strictly speaking should not be called loss factors in this case, are simply a result of such inappropriate substructuring scheme, not even mentioning the testing errors and the ensemble variations. Quasi-SEA or SEA-like analysis As discussed above, a classical SEA analysis which is subject to specific requirements is not always applicable to a specific structure. Although, the necessary conditions are favored toward a successful application of the approach as the frequency of interest goes higher and higher, care still should be taken as an inappropriate substructuring scheme may affect the results. Nevertheless, recent studies have shown that if a system can be generally divided into N subsystems


DK3193

noise

October 24, 2005

19:9


287

which are excited by random, steady, and distributed forces, under the assumptions that the system is linear and the excitations applied to different subsystems are uncorrelated, the average total energies of all subsystems are related to the average input power in a general form as, {P} = ω{C}{E}

(10.45)

where ω is the angular frequency, {P} is a (N × 1) vector of input powers, {E} is a (N × 1) vector of vibrational energies, and {C} is generally a (N × N ) matrix relating the power inputs and the vibration responses. Depending on the characteristics of the matrix {C}, three types of analyses, the classical SEA analysis, the quasi-SEA or SEA-like analysis, and the energy flow analysis, all of which fall in the scope of Equation 10.45 but have different valid conditions and governing territories, were summarized in [145], i.e.,: • The classical SEA only considers resonant transmissions between directly coupled subsystems. No indirect coupling paths exist in the model. All the power flow paths are reciprocal, and thus matrix {C} is symmetric with the elements being loss factors as defined in Equation 10.24. The coupling loss factors and internal loss factors are positive with their physical significances as defined. • Quasi-SEA or SEA-like analysis includes direct coupling paths and indirect coupling paths as well because the requirements for the classical SEA analysis are not being fully satisfied. The reciprocity relationship is still valid for all power flow paths, indicating that only resonant transmissions, either direct or indirect, are considered. Thus matrix {C} is still symmetric. However, negative loss factors might occur in the model due to the equivalencies in the parameters caused by the substructuring scheme that is not fully compliant with the physical nature of the power flow topology within the system. • The energy flow analysis basically includes all the power transmission paths, not only the resonant but also the no-resonant transmissions, in the model. Thus matrix {C} is generally full and not necessarily symmetrical, indicating that there is no reciprocity relationship in the system. The elements in the coupling matrix are generally energy flow coefficients and could be positive or negative. Energy flow method (EFM) Given as an example of the energy flow analysis, the energy flow method (EFM) developed recently is introduced briefly here [250, 144]. Assume that two coupled linear subsystems i and j are excited by uncorrelated stationary random forces. Then the averaged vibrational energies E i and E j in the two subsystems can be written as cii ci j Ei Si = (10.46) c ji c j j Ej Sj


DK3193

noise

October 24, 2005

19:9

288


where ci j are the energy influence coefficients (EIC), and Si is the auto spectral density of the excitations which are related to the power fed into the two subsystems in the form as d 0 Pi Si = ii (10.47) 0 djj Pj Sj where di j are the power influence coefficients (PIC). By assuming the rain-on-the-roof excitation, that is, the subsystem subjected to numerous uncorrelated point excitations proportional to modal mass matrix, the EIC’s and PIC’s can be derived as [250, 144] c ji =

k, p

djj =

kp

kp

mi m j

Bkp mk m p

(10.48)

m kk j Jk ωk π k

(10.49)

where m k , ωk are the modal mass and the natural frequency of the k th mode kp respectively in the coupled system, m i is a modal parameter that relates to the mass matrix and the modal vectors of each subsystem [251]. For a random noise with ω1 and ω2 being the lower and the upper bounds of a frequency band of interest, the cross-mode coupling terms Bkp and term Jk also have closed forms as √ √ √ √ −z k (ω2 − ω1 ) −z k −z p (ω2 − ω1 ) −z p 1 Bkp = e tan−1 + tan−1 π z p − zk ω1 ω2 − z k zk − z p ω1 ω 2 − z p ω2 − ω2 ω2 − ω2 1 Jk = (10.50) tan−1 2 2 k − tan−1 1 2 k 2 ηk ωk ηk ωk where z k = ωk2 (1 − jηk ), and ηk is the modal damping loss factor of the k th mode. Assume that the system responses are generally dominated by the resonant modes within the frequency band. Further simplifications to Equation 10.50 can be made. Bkp ≈

(ω + ω p )2 (ηk ωk + η p ω p ) 2k , 4 (ωk − ω2p )2 + (ηk ωk2 + η p ω2p )2

Jk ≈

π 2

(10.51)

The results obtained here are actually valid for any subsystem pairs in a multicoupled system. With the modal analysis done by the FEM, all the information needed for calculating the EICs and PICs is available in the FEM database. The spatial and frequency averaged response of each subsystem thus can be constructed by Equations 10.46 and 10.47. Since the actual modal information is considered in the calculations, the local as well as the global behaviors of the system are captured. Although the method still assumes the rain-on-the-roof excitations to facilitate the math derivations, no assumptions such as diffuse wave fields and high modal densities as necessary for a classical SEA analysis are required. As seen from Equations 10.46 and 10.47, the terms associated with the indirect coupling in the


DK3193

noise

October 24, 2005

19:9


289

system generally exist to represent the actual power flow in the system. The method is therefore reckoned to be effective in the mid-frequency range where the classical SEA is not valid. In Figure 10.10, the results calculated by the energy flow method for the five-plate system show good agreement with the standard FEM analysis. However, since the assumption of incoherent modal behavior is still necessary in this method, the predominant peak performance associated with individual modes that is usually seen at low frequencies should not be expected in the EFM results. This technique has already been implemented in MSC/NASTRAN v2004 as a postprocessing tool. Other methods that are currently under investigation by researchers include the fuzzy structure method [191], the energy finite element method [218], and the hybrid SEA/FEM method [131]. Readers may refer to these publications for details. Automatic substructuring It is worth noting that all the methods mentioned here still reply on the prior knowledge of the substructuring. As discussed above, inappropriate subsystem definitions may lead to matrix inversion problems, yielding nonphysical solutions for the model derived parameters. In trying to reduce the level of “human effect” involved in the model substructuring so as to improve the model quality, Gagliardini et al. [70] developed an automatic substructuring scheme based on, but not limited to, the finite element analysis. The basic idea was to identify the subsystems by performing an optimization searching and grouping scheme over the calculated or measured point to point frequency response functions (FRF) over the structure system, assuming that a subsystem is an energetic entity that must exhibit at the same time significant energy level difference with other subsystems, and minimized internal heterogeneity. The advantages of this approach are that it is independent of the user expertise, and it can be performed in each frequency band, which may lead to different substructuring schemes as necessary as required in SEA. This technique was shown quite effective and promising for complex structures, such as the vehicle floor. Variances in SEA responses It was argued that SEA results are more meaningful in the sense that the average is taken across a population or a ensemble of similar structures. In statistics, the variance of a data set, in addition to the mean value, is another important measure, from a different aspect, for a random process. For SEA, the significance of knowing the variance in the response is not limited to understanding the dispersion and the confidence level of the predictions. Information such as the probability that a given product will meet the specified requirement, the expected upper bounds of a vibro-acoustic environment, and the sensitivity of the mean and the variance to a particular structural component, all of which is valuable at the early stage of the product design, may become available. Recently, Langley and Cotoni [133]


DK3193

noise

October 24, 2005

19:9

290


presented a method to predict the variances in SEA responses. The method only considers the uncertainties caused in the dynamic properties of subsystems, for example, the local mode shapes and natural frequencies. Two basic assumptions were made to facilitate the analysis: the SEA model provides good estimations of the statistical mean which means that the SEA is built correctly and all standard SEA requirements are met; and in each subsystem, the wave fields are sufficiently random and the frequency of interest is above a transition frequency when the random variation in an individual natural frequency exceeds the mean spacing between two adjacent natural frequencies. It was further argued that the natural frequencies of a subsystem conform to the Gaussian Orthogonal Ensemble (GOE) statistics. A result of this was that the variance in the dynamic properties of a subsystem was shown not to depend on the detailed causes of uncertainty, which thus makes it possible to predict the variance without knowing the detailed differences from one product to the next. Two key factors that directly affect the variances in the SEA responses were found as the variance arising in the external input powers and that in the coupling powers between the subsystems. By introducing two new parameters associated with the loading and the coupling types, an analytical expression was developed for the variances in the SEA predicted total energy responses of each subsystem. It was concluded from the study that the variances for a built-up structure primarily depends on the modal overlap of each subsystem, the frequency band over which the response is averaged, the type of the excitation and the way in which the subsystems are coupled together. Generally, high modal overlaps and wide frequency average bands yield less variance in the SEA responses. R This method has been implemented in AutoSEA2version 2004 as the variance module.

10.5

Application to electrical motors

As seen in previous chapters, owing to the complexities involved, a thorough analytical vibro-acoustic analysis is almost impossible for a motor structure which normally consists of a stator, a rotor, and a casing. In the literature a lot of vibroacoustic analysis work was conducted only on the stator as it was believed that the stator would dominate the total vibro-acoustic response of the whole motor structure. This is only true, however, for large motor structures. For small and medium sized motors, as shown in previous chapters, all the structural details, such as the casing, end-shields and the support, should be taken into account in the analysis. Currently, numerical methods, such as the FEM and the BEM, are usually employed if the vibro-acoustic behavior of an electric motor is of interest at the assembled level. As described in Chapter 9, one may need an electromagnetic FEM model to calculate the force acting on the stator, and a structural FEM model to analyze the structural vibration behavior of the motor structure. Then an acoustic BEM model may be used to evaluate the acoustic response due to the electromagnetic force excitations. Although such numerical schemes seem to work


DK3193

noise

October 24, 2005

19:9


291

well, there are quite a number of limitations for it to be applied in practice. One of major concerns is that finite element/boundary element methods, by their nature, are limited to low frequencies. A large number of elements that are necessary for extending the upper frequency limit would require sufficient computing resources that may be expensive or even prohibitive. In this sense, SEA may hold promise as a tool for the vibration and acoustic analysis of complex motor structures, especially for high frequency responses. Compared to the applications to aerospace, automotive, and building structures, it is a new attempt to apply SEA to a motor structure and at the time of preparing this book, very few works have been reported. Delaere et al [43] have measured the internal loss factors of the stator and the cast iron frame, and the coupling loss factor between the stator and the frame by using the experimental SEA technique. Wang and Lai [229] developed a SEA model for a 2.2 kW induction motor to estimate the radiated sound power due to the electromagnetic origin. The work conducted at the Australian Defence Force Academy [229] is introduced as an example.

10.5.1

Subsystems of a motor structure

For a typical motor structure, there are four basic components, the stator, the rotor, the casing, and the end-shields. In addition, a motor may have other structural elements because of installation or specific requirements. Figure 10.12 shows the motor structure in the study. The stator was press fitted into the casing, and the stator/casing was supported by a plate which was mounted on four isolators. The modal testing results showed that the vibration of the casing behaves differently in three areas, the part attached to the stator and the other two parts on both sides of the stator [224]. Therefore, it is reasonable to divide the casing and the stator into three subsystems: the stator and the part of the casing attached to the stator; and the part of the casing on each side of the stator. The base plate is another subsystem. Since the end-shields only provide the boundary conditions to the casing, and their vibration and acoustic radiation are not so important, they are

Input Power

Stator Casing 1

Base Plate Casing 2

Figure 10.12 Motor structure for the SEA model [233].


DK3193

noise

October 24, 2005

19:9

292


neglected in our SEA model. Also, isolators and the rotor are neglected in the model because they are only important at low frequencies [224]. These simplifications may cause some error at low frequencies. Figure 10.11 shows the tear-down model of the motor structure. It can be seen that this model only involves the structural subsystems, implying that all the mechanical input power would be dissipated in the structure. Since the acoustic power is usually very small compared with the structural vibration power, this simplification is not expected to cause much error in the final results. Normally, the acoustic power radiated from the motor can be directly calculated by using the sound radiation efficiency and the estimated vibration energy of each substructure, as shown in Section 6.5.1. The SEA power balance equation therefore is, 

  E1 η11  E2   −η12 1  =   E 3  ω  −η13 −η14 E4

−η21 η22 −η23 −η24

−η31 −η32 η33 −η34

−1   0 −η41  P2  −η42    . −η43   0  η44 0

(10.52)

Here, only the stator (subsystem 2) is subject to the mechanical power input due to the electromagnetic force excitation.

10.5.2

Internal and coupling loss factors

The determination of internal and coupling loss factors is crucial for establishing the SEA model. Normally, internal loss factors of each subsystem have to be measured and the coupling loss factors could be either measured or predicted. In this study, the experimental approach developed by Lalor [128] has been employed to determine the internal and coupling loss factors. The experimental set up for the energy ratio and input power measurements is shown in Figure 10.13. The motor was excited by a B&K 4810 shaker driven

B&K2032 Analyzer

Power Amplifier Shaker Charge Amplifier Charge Amplifier

HP-300 Computer

Impedence Head

Charge Amplifier

Accelerometer Motor HP3569A Analyzer

Figure 10.13 Schematic diagram of the experimental set-up for measuring loss factors [233].


DK3193

noise

October 24, 2005

19:9


293

by a B&K 2706 power amplifier. The B&K 8001 impedance head was inserted between the shaker and the excitation point. The force and acceleration signals, via two B&K 2635 charge amplifiers, were input to a B&K 2032 analyzer for cross spectra processing. The imaginary part of the cross spectrum was then processed on a HP series 300 microcomputer to obtain the mechanical input power. In the experiment, each of the four subsystems was excited in turn. In order to obtain the uncorrelated modal behavior as required by SEA, two excitation positions, with the normal force applied in the horizontal and vertical directions respectively, were tested for subsystem 1 to 3 and corresponding responses were averaged. The vibration responses of each subsystem for every excitation set-up were measured by a B&K 4383 accelerometer at a total of 130 points (30 for subsystem one, 40 each for subsystems two and three, and 20 for subsystem four). A HP 3569A dual channel frequency analyzer was again used to store and process the data for these 130 points. By transferring the data to a PC, the spatial averaged vibration levels for each subsystem were obtained, and were processed into 1/3 octave bands from 100 Hz to 10 kHz. Figure 10.14 shows that some of the measured internal and coupling factors are high, even greater than unity. This is due mainly to the weak coupling assumptions made in Equation 10.44. Generally, the coupling strength is strong at low frequencies, and the energy tends to distribute uniformly over the whole structure. At high frequencies, since the couplings between subsystems are becoming weak as shown in Figure 10.14 the four subsystem SEA model appears to be reasonable.

10.5.3

Input power to the stator

In order to use Equation 10.52 to predict the vibration response of the motor structure, the mechanical power input to the stator has to be calculated from the electromagnetic force determined for the operating conditions. In vibro-acoustics, the mobility is a parameter that is often used to describe the ability of a structure in responding to a force excitation. The transmitted power can be easily derived if the mobility and the total force are known [37]. For an arbitrary structure, the mechanical mobility associated with a point excitation, namely the point mobility, is determined by the modal density and the mass of the structure, as shown in Equation 10.30. However, for a surface load, such as the electromagnetic force acting on the stator in induction motors, since the vibration response which strongly depends on the force distribution is quite different from that of a point force excitation, the mechanical mobility associated with a surface load, namely the surface mobility, has to be investigated. Generally the surface mobility may not have a unique definition because different force distributions on the surface may lead to different behavior of the surface mobility. However, for a particular case, such as induction motors in which the surface load has common characteristics, it might be meaningful to derive an equivalent surface mobility with a specified equivalent force. As a result, the input power P 2 is related to the real part of the


DK3193

noise

October 24, 2005

19:9

294


Internal Loss Factor

10 1 0.1 0.01 0.001

0.0001

100

160

h1

h2

h3

h4

250

400

630

1000 1600

2500 4000 6300 10000

1/3 Octave Center Frequency, Hz 10

Coupling Loss Factor

1

h 12

h 13

h 14

h 23

0.1 0.01 0.001 0.0001

100

160

250

400

630

1000 1600

2500 4000 6300 10000

1/3 Octave Center Frequency, Hz

Figure 10.14 Measured loss factors [233]. surface mobility M S of a structure subjected to an equivalent exciting force F by: P2 =

1 2 |F| e(M S ) 2

(10.53)

The electromagnetic noise of an electric motor is basically caused by the interaction of electromagnetic forces acting on the rotor and the stator. Since this force generally has a very large radial component (approximately an order of magnitude larger than the tangential component), the vibration of a motor structure is dominated by the flexural vibration. For the vibration response of the stator subject to electromagnetic force wave excitation, Yang [248] argued that axial vibration modes are not as important as circumferential modes, and, therefore, presented a formula for estimating the amplitude of the radial velocity of a stator core with the


DK3193

noise

October 24, 2005

19:9


295

circumferential vibration modes m ≥ 2, which is, 12ωFa a 3 vr |ω=ωm = 4 m ES h

(10.54)

where a is the radius of the stator, h is the thickness, S is the inner surface area of the stator, E is Young’s modulus, and F is the amplitude of the total force acting on the stator. Since the relationship between the circumferential mode number m and the corresponding natural frequencies can be expressed as [233]. ωm h h2 m 2 (m 2 − 2)2 ∼√ = · for m ≥ 2 (10.55) m2 2 ωring 12a m2 + 1 12a where ωring is the angular ring frequency of the stator core. At high frequencies, the modal overlap is high, that is, the natural frequencies of adjacent vibration modes are close to each other. As a result, the vibration response at an excitation frequency ω may be approximated by that of the nearby natural frequency ωm . By substituting Equation 10.55 into Equation 10.54, the real part of the surface mobility is approximately given by e(M S ) ≈

vr |ω=ωm 1 ≈ F Mω

(10.56)

where M is the total mass of the stator. Equation 10.56 indicates that under the excitation of the electromagnetic force wave, the real part of the equivalent surface mobility of a circular cylindrical shell, at high frequencies, is inversely proportional to the mass of the structure, and the frequency. This result supports the discussion made in [229] that for reducing the acoustic power radiated from the motor structure, the random modulation technique may not be effective when the switching frequencies are set high. This is because at high frequencies, the total mechanical power input to the motor is only related to the mass of the motor structure according to Equation 10.56. A change in the force spectrum content rather than the magnitude would not make any changes to the input power and the radiated acoustic power. It should be noted that although Equation 10.56 has been obtained from Equation 10.55 which is only applicable for circumferential modes m ≥ 2, the natural frequencies for a cylindrical shell with both ends free are zero for circumferential modes m = 0 and 1 [226]. In this study, the electromagnetic force for specific operating conditions was calculated by FEM. The corresponding total forces acting on the stator inner surface are shown in Figure 10.15.

10.5.4

Sound power radiated from the motor structure

As described above, the SEA model does not include the acoustic environment. The sound radiation cannot be obtained directly from Equation 10.52. However, with the vibration energy of each subsystem obtained from Equation 10.52, the


DK3193

noise

October 24, 2005

19:9

296

Noise of Polyphase Electric Motors 1000 ‘ideal’ sinusoid, no load ‘ideal’ sinusoid, full load PWM inverter, no load

Total Force, N

100

10 1

0.1 450 rpm 0.01 0

500

1000

1500 2000 Frequency, Hz

2500

3000

(a) 1000 ‘ideal’ sinusoid, no load PWM inverter, no load

Total Radial Force

100

10

1

0.1 1500 rpm 0.01 0

2000


8000

10000

(b)

Figure 10.15 Total radial force acting on the stator calculated by FEM: (a) frequency from 0 to 3000 Hz; (b) frequency from 0 to 10,000 Hz [233]. acoustic power radiated from each subsystem could be determined if the sound radiation efficiencies of each subsystem are known. In Section 6.5.1, the modal averaged radiation efficiencies for each of the four subsystems were estimated analytically. It can be seen from Figure 6.20 that at low frequencies, the stator and base plate are more efficient in acoustic radiation than the casing.


DK3193

noise

October 24, 2005

19:9


297

80

Sound Power Level, dB

70 60 50 40 30 20

Experiment

10 0

SEA 0

500

1000

1500 Frequency, Hz

2000

2500

3000

(a) 60 Experiment


50

SEA

40 30 20 10 0

0

500

1000


2500

3000

(b)

Figure 10.16 Comparison of the acoustic power calculated by SEA with experimental results: (a) PWM inverter, no load, 450 rpm; (b) Benchmark controller, no load, 450 rpm. (Continued on next page.) With the total force shown in Figure 10.15 as the input, by employing Equations 10.52, 10.53, and 10.56 and the sound radiation efficiencies of each subsystem shown in Figure 6.20, the acoustic power radiated from each subsystem, and hence, from the whole motor structure can be obtained. The results for the motor driven under various conditions are shown in Figure 10.16. It can be seen that the total sound power obtained from the statistical model agree fairly well with measured results in general. Note that at high motor speed, the aerodynamic noise dominates the total acoustic power, while SEA model here was only for the electromagnetic origin. At low speed, since the force calculation was


DK3193

noise

October 24, 2005

19:9

298

Noise of Polyphase Electric Motors 70 Experiment


60

SEA

50 40 30 20 10 0

0

500

1000

1500 Frequency, Hz

2000

2500

3000

(c)


70 60

Experiment

50

SEA

40 30 20 10 0

0

2000


8000

10000

(d)

Figure 10.16 (Continued): Comparison of the acoustic power calculated by SEA with experimental results: (c) Benchmark controller, full load, 450 rpm; (d) Benchmark controller, no load, 1500 rpm [233]. done only up to 3 kHz [3], the SEA calculations were also made up to 3 kHz. For the Benchmark inverter, the agreement at high frequencies is quite promising. At low frequencies, the variation in the SEA sound power prediction is directly related to that in the total force. Note that the FEM electromagnetic force model is two-dimensional [231]. A better correlation might be expected with improved electromagnetic force simulations. However, since the primary concern in SEA is the averaged response, specific peaks caused by prominent structural resonances,


DK3193

noise

October 24, 2005

19:9

Sound Power Level (dB)

Statistical Energy Analysis 60 50 40 30 20 10 0 −10 −20 −30

299

Experiment

SEA 0

500

1000 1500 2000 Frequency (Hz)

2500

3000

Sound Power Level (dB)

(a)

40 30 20 10 0 −10 −20 −30

Experiment

SEA 0

500

1000 1500 2000 Frequency (Hz)

2500

3000

(b)

Figure 10.17 Effects of the load and harmonics on the total acoustic power calculated by SEA: (a) increment of sound power due to PWM inverter compared to Benchmark controller at no load and 450 rpm; (b) increment of sound power with full load compared to no load at 450 rpm for Benchmark controller [233].

particularly at low frequencies, are not able be detected by this statistical approach, for example, the peak around 2.5 kHz in Figure 10.16. In Figure 10.17, the effects of the load and the harmonics in the current wavefront introduced by the PWM inverter on the radiated acoustic power are also shown. The agreement between the SEA prediction and the testing results are reasonably good.


DK3193

noise

November 3, 2005

16:24

11 Noise Control 11.1 Mounting 11.1.1

Foundation

A rigid foundation is essential for minimum vibration and proper alignment between the motor and load. Usually, a rigid and solid sole plate or common bed is used as a foundation (Figure 11.1). Reinforced concrete makes the best foundation for large motors and large driven loads (Figure 11.2). A heavy, poured concrete slab is the best foundation design. The concrete foundation should weigh at least 3 to 5 times the combined weight of the motor, coupling, and load machine. If the mass is sufficient, it provides rigid support that minimizes deflection and vibration of the whole system. The concrete foundation can be located on the soil, structural steel, or building floors. A fabricated steel base (sole plate) between motor feet and foundation is recommended. The foundation changes natural frequencies of the stator system of the machine. There will be four points of mounting for an electric motor installation, one at each corner of the mounting base. The best anchor bolts are L- or T-shaped, and should be set in pipe sleeves approximately 50 mm larger than the anchor bolt diameter. Electric motors with aluminum feet require suitable heavy flat steel washers (shim pack) between the foot and the mounting fasteners to spread the bolt-clamping load out over a suitable area. All mounting points of the motor and load machine must be on exactly the same plane to ensure the proper alignment between the motor and level. Grouting is the process of firmly securing equipment to the concrete base. The “grout” is a plastic filler which is poured between the motor sole plate and the foundation upon which it is to operate [204]. The base is a continuation of the main foundation designed to damp any machine vibration present and prevent the equipment from shaking loose during operation. A serviceable and solid foundation can be laid only by careful attention to proper grouting procedure [204].

301 Copyright © 2006 Taylor & Francis Group, LLC

DK3193

noise

November 3, 2005

16:24

302


2

3

4

Figure 11.1 Common bed as a foundation: (1) motor; (2) coupling; (3) load machine; (4) common bed.

A one piece polymer concrete shell can replace the base plate, foundation, anchor bolts, and grouting system (Figure 11.3). Small motors may incorporate a rigid mount, with the frame welded directly to a steel plate formed to match the shape of the frame and incorporating mounting holes. Special mounting brackets are also used.

4 3 2

1

Figure 11.2 Concrete foundation: (1) L-shaped concrete bolt; (2) pipe or sheet metal sleeve; (3) grout under entire base angle; (4) shim pack.


DK3193

noise

November 4, 2005

14:37

Noise Control

303

(a)

(b)

Figure 11.3 (See color insert following page 236.) Polymer concrete foundation Polyshield ® : (a) foundation; (b) electric motor on a polymer concrete foundation. (Photo courtesy of Structural Preservation Systems, Inc., Houston, TX, U.S.A.)

11.1.2

Principles of vibration and shock isolation

Vibration isolation is used to prevent or limit the amount of variable force transferred to the supporting structure by a motor in operation. Machine isolation can also be used to minimize the transmission of vibration to the motor. Variable forces are undesirable, as they may be distracting or destructive to a process. In an extreme case, vibration can be dangerous enough to destroy the supporting structure. Equipment is supported on vibration isolation devices such as springs and pads using different methods, depending on the equipment mounted and rigidity of the supporting structure.


DK3193

noise

November 3, 2005

16:24

304


(a)

(b)

(c)

Figure 11.4 Isolation bases: (a) rails and channels; (b) welded base; (c) concrete inertia base. (Courtesy of J.P. America, Hampshire, IL, U.S.A.) The vibration isolation rails are used for smaller motors Figure 11.4a. They add stiffness to the mounting platform of the motor [217]. The vibration isolation base is usually supplied to either support a piece of equipment with a frame which is not sufficiently stiff to be isolated independently, or to unitize the mounting of a piece of equipment and the motor [217]. It should be mentioned that bases are not sufficiently rigid to be mounted on isolation. The vibration base is constructed of structural members of sufficient depth to


DK3193

noise

November 3, 2005

16:24

Noise Control

305

(a)

(c)

(b)

(d)

Figure 11.5 Vibration isolators: (a) rubber-in-shear mount; (b) free-standing open spring mount; (c) housed spring mount; (d) restrained spring mount. (Courtesy of Greenheck, Schofield, WI, U.S.A.) provide rigidity for the isolation installation, typically channel, I-beam, or angle (Figure 11.4b). The concrete filled vibration isolation base also called the inertia base is stiffer and more stable than the steel vibration isolation base (Figure 11.4c). This type of base is preferable for load machines that are coupled directly to the motor, as the stiffness helps to maintain the coupling alignment [217]. There are several types of vibration isolators also called vibration mounts used in vibration isolation installations (Figure 11.5). The vibration isolator is a resilient member that acts both as a time delay and a source of temporary energy


DK3193

noise

November 3, 2005

16:24

306

Noise of Polyphase Electric Motors Table 11.1 Vibration limits. Type of electrical machine Motors 1000 to 2000 rpm Motors over 2000 rpm Generators

Balance condition peak-to-peak displacement mm

Overall peak velocity 10 to 1000 Hz mm/s

Overall peak acceleration g 0 to 5000 Hz

0.05

5

0.5

0.025 0.05

5 5

1.0 0.5

storage. The main purpose of isolator damping is to reduce or to attenuate the vibration as rapidly as possible. A good vibration mount slows the equipment response to the force or motion disturbance (time delay), absorbs temporarily the energy, and, to some extent, dissipates or damps the energy. When a well-designed vibration mount is installed, the natural frequency of the equipment with the mount is substantially lower than the frequency of the vibration source. The rubber-in-shear mount (Figure 11.5a) consists of two load plates of steel which are embedded in a rubber pad [217]. The equipment to be isolated is bolted to the top load plate, and the bottom load plate is attached to the supporting structure. The open spring (Figure 11.5b), is the simplest of the spring mounts. It must be laterally stable without using a housing. This requires a horizontal stiffness to vertical stiffness ratio from about 0.75 to 1.25. The spring usually provides approximately 50% overload capacity [217]. The housed spring (Figure 11.5c), is similar to the open spring, but is contained in some type of enclosed housing. Often, this housing is telescoping, with rubber pads of some sort between the housing halves, providing some snubbing action for horizontal loads [217]. The restrained spring (Figure 11.5d), is the same in design as the open spring, but a housing or frame is included to restrain the vertical and/or horizontal motion of the spring.

11.1.3

Vibration limits

Table 11.1 is a general guide for acceptable vibration on electrical machines. This table was compiled from industry standards, some published specifications, from manufacturers, and from field experience.

11.1.4

Shaft alignment

The motor shaft and the driven shaft should be aligned within the tolerance of ±0.03 mm in both angular and parallel alignment [171]. This is necessary to prevent mechanical vibration due to the shaft misalignment with frequency given by Equation 7.3, i.e., f sh = 2n m where n m is the shaft speed in rev/s. The angular misalignment is the amount by which the center lines of the motor and driven shafts are skewed (Figure 11.6a). The parallel misalignment is the amount by which the centerlines of the motor and driven shafts are out of parallel (Figure 11.6b).


DK3193

noise

November 3, 2005

16:24

Noise Control

307 1

2

3 (a) 1 2

3 (b)

Figure 11.6 Shaft alignment: (a) angular; (b) parallel. (1) indicator; (2) indicator base; (3) coupling hubs. (Courtesy of Teco Westinghouse Motor Company, Round Rock, TX, U.S.A.)

11.2

Standard methods of noise reduction

Rotor unbalance noise reduction The noise caused by the rotor unbalance is reduced when the dynamic rotor unbalance is within an appropriate limit. There are several rotor balance criteria in common use which are based on the practical experience of electrical machine manufacturers [248]. A “rigid” rotor is a rotor whose operating speed n rig is much less than the first critical speed n cr . A practical flexible rotor operates below its first critical speed, if [248] n rig 2 e f > erig 1 − (11.1) n cr where e f is the allowable mass eccentricity for the practical flexible rotor and erig is the allowable mass eccentricity for a “rigid” rotor.


DK3193

noise

November 3, 2005

16:24

308


Table 11.2 Design measures for minimization of bearing vibration and noise. Measure

Effect

Reduction of vibration due to uncontrolled Coil springs as an motion of loose rolling elements. Modification axial preload on the bearing of the elastic properties of the bearing. Elastic damping elements Reduction of the noise and vibration at in the housing support structure frequencies above several hundred Hz. Bearings with cages made from More quiet operation at high speeds. synthetic plastic-type materials or solid machined cages Bearings with shields or seals Prevent ingress of dirt and foreign matter. Selection of different natural Mismatching the natural frequencies of frequencies of the bearing outer those two parts with an important ring and the bearing support mechanical and electromagnetic exciting structure force frequencies. Avoiding the imperfection Minimization of the bearing of bearing housing positions angular misalignment. Taking into account the effect Suitable clearance group (grade) of machine temperature rise to minimize the bearing induced vibration. on clearance between the rolling elements Precision bearings For machines with special low noise and vibration requirements. Selection of suitable grease Minimization of grease noise. A factory balanced machine may exhibit unacceptable noise and vibration on site when it is coupled to other machines because of: • coupling misalignment or unbalanced coupling; • unbalance of the coupled equipment. It is recommended to carry out balancing jobs in situ, that is, to balance the complete rotating system under actual operating conditions. Motor bearing noise reduction In addition to the proper choice of the bearing type and size, it is recommended to apply the design measures listed in Table 11.2 [248]. Stator stack noise reduction Loose stator laminations and mechanical stresses as a result of stacking and stator assembly contribute to the motor noise. The vibration and noise frequency due


DK3193

noise

November 3, 2005

16:24

Noise Control

309

to loose laminations is expressed by Equation 7.4 and is equal to double the line frequency with frequency sidebands approximately ±1000 Hz, This mechanical noise can be reduced by • impregnation (encapsulation) of the stator stack; • good stator stack and frame assembly. Reduction of the air gap magnetic flux density According to Equation 2.114 the amplitude of the radial force pressure is proportional to the stator and rotor amplitudes of the magnetic flux density (MFD) waves. Reduction of the MFD reduces the radial force waves causing the acoustic noise. On the other hand, the electromagnetic torque developed by the motor decreases. The air gap MFD can be reduced in one of the following ways: • enlarging the air gap g because the magnetic flux density wave b(α, t) is inversely proportional to the air gap, i.e., b(α, t) =

F(α, t) g

(11.2)

• increasing the number of the stator turns N1 per phase or extending the length of the stator stack L i , i.e., k E V1 Bmg = √ 2 2 f N1 kw1 τ L i

(11.3)

where Bmg is the peak value of the normal component of the air gap magnetic flux density, k E is the phase electromotive force (EMF)–to–phase voltage V1 ratio, f is the input frequency, kw1 is the winding factor for the fundamental harmonic, τ is the pole pitch, and L i is the effective length of the stator stack. Reducing the pulsating noise Homopolar flux waves are responsible for producing pulsating noise and vibration in 2-pole machines [248]. Those flux waves are produced by static and dynamic rotor eccentricity and magnetic permeance variations in the stator and rotor cores (Appendix B). The following measures are needed to reduce pulsating noise from 2-pole machines [248]: • the use of a dynamically balanced rotor; • minimization of the backward-rotating field; • minimization of magnetic permeance variations in the stator and rotor cores. Skewing Skewing of the stator slots introduces phase angles between the radial magnetic forces at different axial positions [248]. As a consequence, the average radial force, vibration, and noise level is reduced by skewing.


DK3193

noise

November 3, 2005

16:24

310


For small values of the stator slot skew bs , both the amplitude of the radial force (2.120) and amplitude of the radial displacement (5.1) is proportional to the slot skew factor ksk , i.e., Pmsk = ksk Pm

Amsk = ksk Am

(11.4)

where the stator skew factor for the kth harmonic is given by Equation 4.36, i.e., ksk = sin(k πbs /(2τ )/[k π bs /(2τ )], bs ≈ t1 = π D1in /s1 . Skewing is a successful method of noise reduction for short core machines, i.e., L i ≤ 300 mm [220]. For larger machines with longer cores skewing is not satisfactory and cannot be regarded as an effective way of noise reduction [220]. Reduction of sound radiation efficiency An electric motor with a cylindrical frame will be considered. The relative sound radiation factor also called the relative sound radiation efficiency varies with [248]: • the magnetic force order (circumferential mode), r ; • the effective radius of surface, ka = ωa /c, see Equation 6.44; • the length–to–diameter ratio of the machine, L f /D f , whre L f is the frame length and D f is the frame diameter. Figure 6.10 shows the variation of the relative sound radiation efficiency with ka for force orders (circumferential mode numbers) r = 0 to 6 and infinitely long cylinder, neglecting the length–to–diameter ratio. Finite length of the cylinder and consequently end effects cause that the radiation efficiency of a practical machine is different than that shown in Figure 6.10. The relative sound radiation factor of a practical machine is a function of the frame length L f and frame diameter D f (see Equations 6.44 and 6.61). The radiation efficiency for the given mode r of vibration increases as the ratio of the frame length L f to its diameter D f increases [200, 248]. This effect can be utilized for reducing the radiated acoustic power (6.73), (A.12). The machine with shorter length L f and larger diameter D f (lower L f /D f factor) radiates less accoustic power and is less noisy than that with longer length and smaller diameter. Reduction of dynamic vibration of the machine surface The reduction of the dynamic deflection of the stator-frame system surface requires a very accurate analysis of the natural and exciting frequencies. The following methods can be used to reduce the dynamic vibration: • Mismatching the natural frequencies of the machine structure and frequencies of important exciting forces. It is necessary to determine with a very high accuracy the natural frequencies of the stator-frame structure and rotor-shaft


DK3193

noise

November 3, 2005

16:24

Noise Control

311

system and then to compare the mechanical natural frequencies with frequencies of the important forces. • Adjustment of the mechanical coupling between the stator core laminations and the frame. By adjusting the stiffness of the connection between the frame and the core, stiffness of the frame and mass of the frame, the outer frame vibration can be reduced to only a small fraction of the stator core. • Increasing the stiffness of stator core (yoke) by increasing its thickness. Approximately, if the core thickness increases by 50%, the sound power level is reduced by 10.5 dB. Owing to the cost increase, this method is only recommended for small machines. • Increasing the damping capacity of the machine structure by incorporating damping materials in the machine, for example, sealing the gap between the stator core and frame with varnish or epoxy resin or inserting shear damping rings in the motor yoke. • Increasing the exciting force order r (Figure 2.9). The stator vibration decreases significantly when the force order r increases.

11.3

Active noise and vibration control

11.3.1

Principles of active noise control

The active noise control (ANC) is a technique in which unwanted noise is canceled by introducing controllable secondary sources. The outputs of the secondary source are arranged to interfere destructively with the noise from the primary source. The performance of an ANC system is monitored by error sensors that measure the residual noise. In contrast to passive techniques, ANC systems are small, portable, adjustable to different environments, and less costly. Although the idea of ANC is not new, its practical application had to wait for the recent development of sufficiently fast electronic control technology. An ANC system can be effective across the entire noise spectrum, but it is particularly appropriate at low frequencies, typically from 20–500 Hz, where passive methods using absorptive materials or noise partitions become inefficient or expensive. Most efficient noise cancelation is usually achieved by the combination of active and passive methods. ANC is based on the principle of superposition of sound waves. In most cases, to cancel a sound wave traveling in space, a second sound wave, the socalled antinoise wave, having the same amplitude but opposite phase to the primary sound wave, is created. Fundamental issues in choosing the control methods are the physical control strategy and the control approach [52]. The integral part of the ANC system is


DK3193

noise

November 3, 2005

16:24

312


2

1 Reference Signal

Error Signal

Control Signal 3 (a)

2

Error Signal

Control Signal

3 (b)

Figure 11.7 Basic ANC systems: (a) feedforward system; (b) feedback system [84].

the signal processing unit which produces the output signal on the basis of the input signals [52]. Figure 11.7 shows basic ANC systems for electric motor driven fans installed in a confined space, for example, air duct of industrial air handling system or office building air conditioning system [84]. The most common is an adaptive feedforward ANC system (Figure 11.7a). This control system consists of a reference sensor, control source, error sensor, and electronic controller (including the control algorithm) [84]. A reference sensor (microphone) samples the incoming signal, which is then filtered by the electronic controller. The electronic controller produces output power signal for the control source (loudspeaker). The error sensor provides a signal for the control algorithm to adjust the controller output signal in such a way as to minimize the sound pressure detected by the error microphone. In Figure 11.7b the electronic controller is an adaptive filter and algorithm (for adaptive system) or consists of a fixed low pass filter and amplifier (nonadaptive system). Although still in the development stage, ANC is receiving considerable attention for various applications from active hearing protectors to active attenuation of aircraft fuselage noise. It is also used in industrial apparatus, dynamic systems, automobiles, and building appliances as HVAC (heating, ventilation, and air conditioning).


DK3193

noise

November 3, 2005

16:24

Noise Control

11.3.2

313

Induction motor acoustic noise reduction

The acoustic noise of magnetic origin can be reduced by generating in the air gap the antinoise MFD with the same amplitude as the primary MFD wave, but shifted in space by 180◦ . Let us assume that the space harmonic ν = 7 produces excessive noise. According to Equation 2.111, the 7th harmonic stator MFD wave can be expressed as b17 (α, t ) = Bm7 cos(7 p α − ωt ). (11.5) The wave Equation 11.5 gives the rise to the following radial magnetic pressure due to the 7th harmonic of the stator MFD pr 7 (α, t ) =

1 b17 (α, t )2 . 2µ0

(11.6)

If the following anitinoise stator MFD can be generated b17a (α, t ) = Bm7a cos(7 p α − ωt − π )

(11.7)

and magnitudes Bm7 = Bm7a are equal, the total radial magnetic pressure produced by the two waves (Equation 11.5) and (Equation 11.7) is equal to zero, i.e., pr 7t (α, t) =

1 [b17 (α, t) + b17a (α, t)]2 = 0. 2µ0

(11.8)

Figure 11.8 illustrates a partial cancelation of the magnetic pressure produced by the 7th space harmonics ν = 7 of the primary and antinoise MFD waves (t = 0) with amplitudes Bm7 = 0.18 T and Bm7a = 0.7Bm7 in a 2-pole (2 p = 2) induction motor. Even if the MFD waves differ in amplitudes (Figure 11.8a), the magnitude of the reduced radial magnetic pressure is more than ten times lower than that produced by the primary wave (Equation 11.5), i.e., 12,890 N/m2 vs. 1160 N/m2 . This principle can be applied to a.c. machines for low frequency ANC of magnetic origin. A special hardware and software is required. In [13] an ANC of induction motor by injection into the stator winding higher current harmonics with controllable amplitude, phase, and frequency has been proposed. The induction motor is fed with a voltage source PWM inverter at constant voltage–to–frequency ratio V1 / f = const. The phase current signal is transferred to the synchronization system which in turn provides a signal (proportional to the frequency) to the constrol system. The control system adjusts the voltage harmonic parameters for the inverter [13]. The maximum magnetic noise of and electrical machine is for frequencies of radial magnetic forces which are the same or very close to natural frequencies of the stator system. Avoiding the excitation of stator system resonance frequencies can be a key factor in ANC strategy of induction motors [86]. A VVVF inverter-fed induction motor operates at constant voltage–to–frequency ratio, i.e., V1 / f = const. The acoustic feedback can be used to alter the V1 / f ratio in order


DK3193

noise

November 3, 2005

16:24

314

Noise of Polyphase Electric Motors 0.2 0.2 0.1 b17 (α, t) b17a (α, t)

0 −0.1

−0.2 −0.2

0 0

1

2

3 α (a)

4

1

2

3 α

4

5

6 2⋅π

1.5 ⋅104 1.5 ⋅104 1 ⋅104 pr7 (α, t) pr7t (α, t)

5000

0 0

0 0

5

6 2⋅π

(b)

Figure 11.8 Cancelation of radial magnetic pressure due to the 7th space harmonic of the MFD (Equation 11.5) by generation of the antinoise wave (Equation 11.7) for Bm7 = 0.18 T, Bm7a = 0.7Bm7 , 2 p = 2, and t = 0: (a) MFD waves according to Equations 11.5 and 11.7; (b) radial magnetic pressures according to Equations 11.6 and 11.8. to achieve a given torque–speed operating condition using a range of different inverter frequencies [86]. In order to maintain the same torque–speed operating point, it is necessary to vary V1 / f which results in motor operation at higher and lower magnetic flux than the rated flux. The torque–speed characteristic can be effectively changed. The excitation frequencies range from the nominal frequency f n to 1.1 f n to avoid sharp resonances which may have bandwidths of only 2 to 3 Hz [86]. The motor can maintain the required loading conditions and, at the same time, keeps away from the excitation at resonant frequencies. In this way, large noise peaks in the noise spectrum are suppressed. The ANC of an induction motor is shown in Figure 11.9. The controller responds to the feedback through a minimization strategy while the accurate


DK3193

noise

November 3, 2005

16:24

Noise Control

315 Input Noise Signal

IM

Comparator

Current Voltage

Controller

Figure 11.9 Schematic diagram for ANC of an induction motor [86]. excitation frequency and corresponding voltage are adjusted to minimize the vibroacoustic response at the operating speed and load [86].

11.3.3

Active vibration isolation

The structure-borne noise is produced by vibrating machine parts as, for example, engine crank shaft, gear box, compressor, pump, electric motor, and so on. The basic vibro-isolation method is to interrupt the propagation of vibration on the route from the source to the object [125]. Therefore, active vibration mounts and active shock absorbers have been developed. Those active parts are stiff enough to carry the static load and are dynamically resilient to avoid transmission of vibration. The vibro-isolated object is fitted with an appropriately controlled actuator which exerts a force that counteracts the forces producing vibration. For active vibration isolation the following actuators have found applications: electropneumatic actuators, electrohydraulic actuators, piezoelectric actuators, and electromechanical actuators (voice coil motors, moving magnet motors, linear motors). As active shock absorbers magnetorheological dampers with piston, magnetic circuit, and magnetorheological fluid are used. The dynamic vibration absorber is designed to reduce the influence of a force, the excitation frequency of which is nearly coincidental with the natural frequency of the system. An ideal undamped dynamic vibration absorber consists of an auxiliary mass and a stiff spring, tuned to the excitation frequency necessary to cause the steady amplitude of the system to be near zero at that frequency [29]. The so-called reaction mass actuator (RMA) combines an electricallypowered actuator with a reaction mass in order to generate time-variable forces [78]. An RMA can be added to existing equipment and deployed either at the source of vibration or, alternatively, at critical locations where vibration must be small. Smart magnetic materials, for example, magnetostrictive materials (Figure 11.10a) or linear electric motors (Figure 11.10b) can be used as RMA. Magnetostrictive


DK3193

noise

November 3, 2005

16:24

316


(a)

(b)

Figure 11.10 Reaction mass actuators: (a) magnetostrictive actuator; (b) linear induction motor. (Photo courtesy of SatCon, Cambridge, MA, U.S.A.) materials offer extremely high force density ratio at frequencies above 400 Hz, for example, canceling helicopter gearbox vibration and noise [78]. Linear motors offer cancelation at low frequencies, limited only by the stroke capability, and can be used for example, as electromagnetic shakers for flight flutter testing [78]. As an example, a high speed elevator will be considered. Small deformations or misalignments in the elevator guide rails can make them vibrate from side-to-side


DK3193

noise

November 3, 2005

16:24

Noise Control Linear Motor

317 Linear Motor

Figure 11.11 ANC of passenger elevator car using a high performance permanent magnet (PM) linear motor. (Photo courtesy of Mitsubishi, Japan.) making passengers feel uncomfortable. Mitsubishi in Japan have come up with a new technique which uses active vibration control and PM linear motors to stabilize the passenger compartment as it moves (Figure 11.11). The system uses accelerometers to sense the minuscule lateral vibrations. This information is fed to a small electronics package, which controls a pair of active roller guides attached to the bottom of the lift. Linear motors consume less than 20 W of power. The system will allow elevators to travel at speeds more than 5 m/s, with a lateral vibration acceleration level less than half the usual level, that is, 10 cm/s2 .


DK3193

noise

October 24, 2005

19:9

Appendix A Basics of Acoustics A.1

Sound field variables and wave equations

In acoustics, sound waves are not specifically limited to the sensation of hearing, but are broadly defined as the small fluctuations in the pressure, the medium density, the temperature and the velocity in the wave-carrying medium, such as air, which is caused by a disturbance such as a solid vibrating surface, or jet flows, and so on. Such fluctuations in the medium, spatially, constitute the sound field. In acoustics, the sound field is normally characterized by the variable (perturbation) values of the pressure and velocity in the medium, namely the sound pressure p and the medium particle velocity v. An equation, namely the acoustic motion equation or the wave equation, which governs the wave propagation in a homogeneous medium is given as [124] ∇ 2 p(r , t) −

1 ∂ 2 p(r , t) =0 c02 ∂t 2

(A.1)

where c0 is the sound wave speed or the phase speed of the wave in the medium (for air, c0 = 344 m/s at 20◦ C). Given as an example, assume one-dimensional sound waves (plane waves) traveling along the x direction. The corresponding wave equation becomes ∂ 2 p(x, t) 1 ∂ 2 p(x, t) − 2 = 0. 2 ∂x ∂t 2 c0 The general solution of this equation is of the form x x p(x, t) = g1 t − + g2 t + c0 c0

(A.2)

(A.3)

where g1 (·) and g2 (·) are arbitrary functions, with g1 (t −x/c0 ) generally representing a wave propagating in the +x direction and g2 (t + x/c0 ) in the −x direction. For a plane wave varying sinusoidally with time and traveling in the +x direction, 319 Copyright © 2006 Taylor & Francis Group, LLC

DK3193

noise

October 24, 2005

19:9

320


the solution can be written as,

x p(x, t) = p A cos ω t − c0

(A.4)

where p A is the wave amplitude in Pa, ω = 2π f , and f is the frequency of the wave in Hz. Often, the complex-number representation is used for the convenience in theoretical studies [124], i.e., p(x, t) = p A e j (ωt−k0 x)

(A.5)

where k0 = ω/c0 is the acoustic wavenumber. The acoustic wavelength λ0 , which is inversely proportional to the acoustic wavenumber, is expressed as a function of c0 and f , c0 2π 2π c0 λ0 = = m. (A.6) = k0 ω f The particle velocity in the sound field for this plane wave can be derived from the sound pressure as [124] v(x, t) = −

p A j (ωt−k0 x) 1 ∂ p(x, t) = e jωρ0 ∂ x ρ0 c0

(A.7)

In acoustics, the ratio between the sound pressure and the particle velocity is defined as the specific impedance of the wave or the wave impedance, i.e. zs =

p v

Ns/m3

(A.8)

The specific impedance is generally a complex number, indicating the capability of energy propagating to the adjacent medium particles along the traveling direction. For a plane wave of which the propagation is thorough, we have z s = ρ0 c0 . Particularly, ρ0 c0 is called the characteristic impedance of the medium. Any discontinuities of the wave impedance or mismatches between the wave impedance and the characteristic impedance along the propagation path would cause sound reflections or even energy dissipations [124]. The energy conveyed by the sound waves is usually described by the sound intensity and the sound power. The sound intensity is defined as the energy transmitted through a unit area perpendicular to the wave propagation direction per unit time. Tp 1 I = pvdt W/m2 (A.9) Tp 0 where T p is the period of the signal. When the sound pressure and the particle velocity are described by the complex-number representation, the sound intensity for a plane wave can be expressed as, I =

1 p 2A e( pv ∗ ) = 2 2ρ0 c0

(A.10)

where e(·) and the superscript ∗ represents the real part and the conjugate of a complex number, respectively.


DK3193

noise

October 24, 2005

19:9

Appendix A

321

The sound power is the total energy through the area S p perpendicular to the wave propagation direction per unit time, which is = IdS W. (A.11) Sp

The radiated sound power from a source has a general expression that relates the surface vibration velocity to the characteristic impedance of the ambient medium. = ρ 0 c0 S v 2 σ (A.12) where S is the area of the radiating surface, v 2 is the spatial averaged mean square velocity over the structure radiating surface, and σ is the sound radiation efficiency of the radiator. The radiation efficiency of different radiators is discussed in Chapter 6. The sound energy density (SED) per unit volume of a medium is D=

p 2A p2 = A 2 γ P0 ρ0 c0

(A.13)

where P0 is the static pressure of the medium, and γ = 1.4 for diatomic gases, for example, air.

A.2

Sound radiation from a point source

Assume a spherically symmetric source, for example a pulsating sphere with radius a and the surface velocity u(t), for which the sound field is expected to have no angular dependence. By utilizing spherical coordinates in which the source is located at the origin point, the wave equation can be written as, ∂ 2 p(r, t) 2 ∂ p(r, t) 1 ∂ 2 p(r, t) − 2 + =0 2 ∂r r ∂r c0 ∂t 2

(A.14)

where the sound pressure in the field p(r, t) is only a function of the radial distance r and the time t. Assume that the sphere surface is pulsating sinusoidally with time as u(t) = u 0 e jωt . The sound pressure response in the field will also have sinusoidal variations with time as p(r, t) = p A (r )eωt . The wave equation can be rewritten as, d 2 p A (r ) 2 d p A (r ) + k02 p A (r ) = 0. + (A.15) dr 2 r dr The general solution of Equation A.15 is of the form, p A (r ) =

A − jk0 r B e + e jk0 r r r

(A.16)

where A, and B are two constants. The first term represents an outgoing wave from the source, and the second term represents a wave propagating inward to the


DK3193

noise

October 24, 2005

19:9

322


origin. In a free-field where the inward going wave does not exist due to the acoustic unbounded boundary condition at infinity, we have B = 0. At the sphere surface, the acoustic boundary condition requires that the surface normal velocity u 0 be equal to the corresponding component of the medium particle velocity, which is 1 d p A (r ) v A (r )|r =a = − = u0. (A.17) jωρ0 dr r =a Thus, constant A can be determined from Equation A.17, and the sound pressure in the field can be obtained as, p A (r ) =

jρ0 c0 k0 Q 1 − jk0 a − jk0 r e 4πr 1 + k02 a 2

(A.18)

where Q = 4πa 2 u 0 is called the source strength, which is the volume velocity at the sphere surface. Equation A.18 indicates that the amplitude of the sound pressure generated by this spherical source in a free-field is inversely proportional to the distance between the receiver and the source, and also depends on the source strength, frequency, source dimension, and medium characteristics. Particularly, at low frequencies where the source radius is much smaller than the acoustic wavelength, i.e. k0 a 1, there is | p A | = (ρ0 c0 au 0 )(k0 a)/r ; and at high frequencies where the source radius is much greater than the acoustic wavelength, i.e. k0 a 1, we have | p A | = (ρ0 c0 au 0 )/r . It is seen that for a given vibration amplitude u 0 , the sound pressure in the field has smaller amplitudes at lower frequencies, indicating a less efficient sound radiation. Normally, a large source (a ↑) or large source vibration amplitudes (u 0 ↑)are necessary in order to have the low frequency acoustic response be at the same level as high frequency responses. An example of this is the speaker system. In particular, the source that satisfies the condition k0 a 1 is called the point source in acoustics, for which the sound pressure in the field has a simple expression as, jρ0 c0 k0 Q − jk0 r p A (r ) = e . (A.19) 4πr The medium particle velocity can be obtained as pA 1 1 d p A (r ) v A (r ) = − = . (A.20) 1− jωρ0 dr ρ0 c0 jk0r The specific impedance of the wave there is p jk0r z s = = ρ 0 c0 · . v jk0r − 1

(A.21)

It is seen that in the so-called near field where the field point is very close to the source compared to the wavelength, i.e. k0r 1, the particle velocity and the sound pressure are 90◦ out of phase such that a large amount of energy exchanges between the source and the medium, as if a layer of medium is moving with the


DK3193

noise

October 24, 2005

19:9

Appendix A

323

source [124]. In the far field, i.e., k0r 1, the particle velocity and the sound pressure are oscillating in phase like a plane wave, such that the specific impedance is equal to the characteristic impedance of the medium. The energy is thus conveyed by the medium and propagating via medium particle movements to infinity. The sound intensity (in the +r direction) at any point in the field can be obtained as, 1 ρ0 c0 k02 Q 2 e( p A v ∗A ) = . (A.22) 2 32π 2r 2 Often, the total energy radiated from the source is of interest because it represents the impact of the source on the environment. Therefore, the total sound power, which is the total energy transmitted through a spherical surface enclosing the source per unit time, is, 2π π ρ0 c0 k02 Q 2 = . (A.23) I r 2 sinθdθdϕ = 8π 0 0 I =

As expected, the total radiated sound power is a constant irrespective of the radial distance.

A.3

Decibel levels and their calculations

In engineering, the sound pressure is measured by the root-mean-square (rms) value pe , i.e., Tp 1 pe = [ p(t)]2 dt N/m2 . (A.24) Tp 0 Also, the logarithmic scale is adopted for quantifying the variations in the sound pressure because the human hearing response varies from the threshold of hearing of 20µPa to the threshold of pain of 200 Pa (i.e., a range of 10 millians). The sound pressure level (SPL) in decibels (dB) is defined as pe = 20 log pe + 94 pr e f

L p = 20 log10

dB

(A.25)

where pr e f = 2 × 10−5 Pa is the reference sound pressure in the rms value. Correspondingly, the sound intensity level (SIL) is defined as, L I = 10 log10

I = 10 log I + 120 Ir e f

dB

(A.26)

where Ir e f = pr2e f /(ρ0 c0 ) = 10−12 W/m2 is the reference sound intensity. The sound power level (SWL) is defined as L W = 10 log10

= 10 log + 120 r e f

dB.

(A.27)

The reference sound power r e f = 10−12 W is corresponding to the reference sound intensity transmitted through a unit area. For a plane wave or a wave that


DK3193

noise

October 24, 2005

19:9

324


can be treated as a plane wave over an area S p , the relationship between the sound power level and the sound pressure level is, L W = L p + 10 log10 S p

dB.

(A.28)

Equation A.28 indicates that for a source in the free field, the dB increments in the sound pressure level and the sound power level are the same. For example, an increase in the source output such as the sound pressure in the sound field doubled in the amplitudes would lead to a 6 dB increment in both the sound pressure level and the sound power level. However, the increment ratios in the sound power and sound pressure are different. Table A.1 shows the dB conversions of various sound power and sound pressure ratios. The plus and minus dB values correspond to the ratio greater and less than unity, respectively. For a sound field with multisources, the total sound pressure in the sound field should be the linear superposition of the sound pressure (with phase information included) generated by each of the sources. Particularly, if the sources are incoherent, that is, the sources are either operating at different frequencies or having random phases between their generated waves, the total sound pressure is added on a linear energy basis, which is pt2 = pi2 (A.29) i

where pt is the total sound pressure, and pi is the sound pressure associated with each source. Correspondingly, the relationship between the total sound pressure level L pt and each individual sound pressure level L pi is of the form L pt = 10 log

10

L pi 10

.

(A.30)

i

Based on Equation A.30, two incoherent waves with the same amplitudes will yield 3 dB higher in the total sound pressure level than either one of the waves. Given as Table A.1 dB conversions. Sound Power Ratio 1 1 1.25 0.8 1.6 0.63 2 0.5 2.5 0.4 3.15 0.315 4 0.25 5 0.2 6.3 0.16 8 0.125


dB 0 ±1 ±2 ±3 ±4 ±5 ±6 ±7 ±8 ±9

Sound Pressure Ratio 1 1 1.12 0.89 1.25 0.8 1.414 0.708 1.59 0.63 1.78 0.56 2 0.5 2.24 0.45 2.5 0.4 2.82 0.36

DK3193

noise

October 24, 2005

19:9

Appendix A

325

an example, assume that there are three incoherent sources, each generating 90 dB, 87 dB, and 87 dB at a specific point in the sound field separately. When three sources are operating together, 87 dB plus 87 dB yields 90 dB. With another 90 dB, the total sound pressure level at this point is 93 dB.

A.4

Spectrum analysis

Frequency spectrum is the representation of a signal in the frequency domain. The signal is broken into multiple periodic signals, each with an amplitude and phase. A tunable narrowband filter has a constant bandwidth w = f u − fl where f u and fl are the upper and lower half-power frequencies. The center frequency of the band is defined as f c = f u fl . (A.31) Spectrum analysis uses filters whose bandwidth is a fractional ratio of the center frequency f c of the filter. For example, a 1/3 octave filter centered at 1000 Hz would have a bandwidth of 260 Hz. Bandwidth (relative to a normalized center frequency of 1) is computed as 2x − 1, where x = 1, 1/3, 1/12 and 1/24 for octave, 1/3 octave, 1/12 octave and 1/24 octave respectively, which are the typical bandwidths used primarily for acoustical and vibration analysis. The ratio of the upper to lower half-power frequencies is f u / fl = 2 for the octave-band filter, f u / fl = 21/3 for the 1/3 octave-band filter, and f u / fl = 21/10 for the 1/10 octave-band filter. The preferred center frequencies and the corresponding logarithmic bandwidths for modern octave-band and 1/3 octave-band filters are shown in Table A.2.

Table A.2 Center frequencies and bandwidths for the preferred octave and 1/3octave-band. ISO band numbers 11, 12, 13 14, 15, 16 17, 18, 19 20, 21, 22 23, 24, 25 26, 27, 28 29, 30, 31 32, 33, 34 35, 36, 37 38, 39, 40 41, 42, 43

Octave-band center frequency f c Hz 16 31.5 63 125 250 500 1000 2000 4000 8000 16,000


One-third octaveband center frequencies Hz 12.5, 16, 20 25, 31, 40 50, 63, 80 100, 125, 160 200, 250, 315 400, 500, 630 800, 1000, 1250 1600, 2000, 2500 3150, 4000, 5000 6300, 8000, 10,000 12,500, 16,000, 20,000

DK3193

noise

October 24, 2005

19:9

Appendix B Permeance of Nonuniform Air Gap B.1

Permeance calculation

A space containing a quasistationary magnetic field may be partitioned into flux tubes or geometrical figures in which all lines of flux are perpendicular to their bases and no lines of flux cut their sides [173]. The permeance of a flux tube is G = µ0 µr

S = S l

(B.1)

where µ0 = 0.4π 10−6 H/m is the magnetic permeability of free space, µr is the relative magnetic permeability, S is the cross-sectional area, l is the total flux tube length, and = µ0 µr /l is the relative permeance per unit area. The relative permeance of the smooth air gap of a machine with ferromagnetic stator and rotor cores and magnetic saturation being neglected is g0 = µ0

1 g

(B.2)

where g is the equivalent air gap thickness which includes • the effect of slots for induction machine g = gkC

(B.3)

• the effect of slots and permanent magnet (PM) equivalent thickness for PM brushless machines hM g = gkC + (B.4) µrr ec In the above equation g is the air gap thickness (mechanical clearance), h M is the PM thickness (height), kC is Carter’s coefficient according to Equation 2.72, and 327 Copyright © 2006 Taylor & Francis Group, LLC

DK3193

noise

October 24, 2005

19:9

328

Noise of Polyphase Electric Motors 3 ⋅ 10−4 2.5 ⋅ 10−4 2 ⋅ 10−4 Λg 1 ⋅ 10−4 5 ⋅ 10−5 0

0

0.1

0.2

0.3

0.4

0.5

x

Figure B.1 Distribution of the air gap relative permeance according to Equation B.5 for PM synchronous motor with s1 = 36, 2 p = 10, b14 = 3 mm, g = 0.95 mm, e = 0.1 mm, ed = 0.1 mm. µrr ec is the relative recoil permeability of the PM. To take into account the increase in the air gap due to the stator slot opening, the mechanical clearance g is to be replaced by an equivalent air gap g = gkC . The relative air gap permeance with stator slot openings and eccentricity effect taken into account is g (α, t) = g0 λg1 (α)λg2 (α)λge (α)λged (α, t)

(B.5)

where g0 is the permeance of smooth air gap according to Equation 2.69, λg1 (α) is the relative value of harmonic permeance according to Equation 2.70 that includes the stator slot openings, λg2 (α) is the relative value of harmonic permeance according to Equation 2.80 that includes rotor slot openings, λge (α) is the relative permeance given by Equation 2.92 that includes the static eccentricity and, λged (α, t) is the relative permeance given by Equation 2.93 that includes the dynamic eccentricity. The relative permeances λge (α) and λged (α, t) have been derived below. The resultant permeance g (α, t = 0) according to Equation B.5 is plotted against the circumferential direction x = ατ/π in Figure B.1.

B.2

Eccentricity effect

The magnetic flux density distribution in an a.c. machine with nonuniform air gap can be described as follows: 1 b(α, t) = µ0 Fm cos( pα − ωt − φ) = ge (α)Fm cos( pα − ωt − φ) g(α)kC (α) (B.6) where g(α) is the air gap varying with the angle α, kC (α) is Carter’s coefficient depending on the angle α and φ is the angle between the field axis and axial plane from


DK3193

noise

October 24, 2005

19:9

Appendix B

329

R

α

r

O′

O

e

Figure B.2 Magnetic circuit of an electrical rotating machine with eccentricity. which the angle α is measured. According to Figure B.2 for small eccentricities e < 0.5g the air gap can be expressed as a function of the angular displacement α, i.e., g(α) = R − r − e cos α. For uniform air gap the eccentricity e = 0 and the air gap g(α) = g = R − r . Defining the static relative eccentricity according to Equation 2.84, the variation of the air gap with the angle α can be expressed as g(α) = g − e cos α = g[1 − cos α].

(B.7)

As the air gap decreases, Carter’s coefficient kC in Equation B.6 increases (Figure B.3). Assuming kC (α) = kC = constant (as for a uniform air gap), Equation B.6 takes the form b(α, t) = µ0

1 cos( pα − ωt − φ) cos( pα − ωt − φ) = Bm Fm gkC 1 − cos α 1 − cos α

= ge (α)Fm cos( pα − ωt − φ) = Bm cos( pα − ωt − φ)λge (α) where Bm = µ0

Fm gkC

(B.8) (B.9)

and 1 1 1 = µ0 λge (α). (B.10) gkC 1 − cos α gkC Equations B.8 and B.10 do not include the variation of the air gap permeance with slot openings. The relative permeance due to static eccentricity λge in Equation B.10 can be resolved into Fourier series as follows ge (α) = µ0


DK3193

noise

October 24, 2005

19:9

330

Noise of Polyphase Electric Motors 1.5 1.4 b14 = 5 mm 1.3 4 mm

kc

3 mm 1.2 2 mm

1 mm

1.1 1

0.002

0.001

0

0.003

0.004

0.005

g

Figure B.3 Carter’s coefficient kC according to Equation 2.72 as a function of the air gap for the stator slot openings b14 = 0.0001 m = 1 mm, 2 mm, 3 mm, 4 mm, and 5 mm. λge (α) =

1 = λ0e + [λck cos(k α) + λsk sin(k α)] 1 − cos α k

where • the constant coefficient λ0e

1 = 2π

π

−π

1 1 dα = √ 1 − cos α 1 − 2

(B.11)

• the magnitudes of the first harmonics √ 1 − 1 − 2 cos α 1 π λc1 = dα ≈ 2 √ (B.12) π −π 1 − cos α 1 − 2 sin α 1 π λs1 = dα = 0. (B.13) π −π 1 − sin α Figure B.4 shows the values of coefficients λ0e and λc1 plotted against the relative eccentricity . In general, including only the first two terms of the Fourier series √ 1 1 − 1 − 2 λge ≈ λ0e + λc1 cos α ≈ √ +2 √ cos α (B.14) 1 − 2 1 − 2 In practice, the eccentricity < 0.5, so that on the basis of Figure B.4 λ0e ≈ 1 Since √

1 1−

2

1 ≈ 1 + 2, 2


the term

(B.15) λc1 ≈

(B.16)

DK3193

noise

October 24, 2005

19:9

Appendix B

331

l 0e l c1

2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0

l 0e

l c1

0

0.2

0.1

0.3

0.4 e

0.5

0.6

0.7

0.8

Figure B.4 Coefficients λ0e (Equation B.11) and λc1 (Equation B.12) plotted against the relative eccentricity . Thus, including only the first two terms of Fourier series λge (α) ≈ 1 + cos α

or

ge (α) =

µ0 (1 + cos α) gkC

(B.17)

Similar to Equation B.14 can be written for the dynamic eccentricity. The magnetic flux density in the case of nonuniform air gap is described by the following equation resulting from Equation B.8 b(α, t) = Bm cos( pα − ωt − φ)λge (α, t) = Bm [λ0e + λc1 cos(α + ω t)] cos( pα − ωt − φ) = Bm λ0e cos( pα − ωt − φ) +

Bm λc1 Bm λc1 cos[( p + 1)α − (ω + ω )t − φ] + cos[( p − 1)α + (ω − ω )t + φ] 2 2 ≈ Bm cos( pα − ωt − φ)

+Bm cos[( p+1)α−(ω+ω )t −φ]+Bm cos[( p−1)α+(ω−ω )t +φ]. 2 2

(B.18)

In Equation B.18 the following trigonometric identities cos x cos y = 0.5[cos(x + y)+cos(x − y)] and cos(−x) = cos(x) have been used. Owing to the eccentricity, in addition to the fundamental harmonic of the magnetic flux density, there are also space harmonics with wave numbers ν = p − 1 and ν = p + 1. For a machine with 2 p > 2 the total flux around the periphery from 0 to 2π is zero [41], i.e., 2π

0


b(α, t)dα = 0.

(B.19)

DK3193

noise

October 24, 2005

19:9

332


In 2-pole machines ( p = 1) the integral taken from 0 to 2π over b(α, t) with respect to α is

2π

Bm cos(α − ωt − φ)[1 + cos(α + ω t)]dα = π Bm cos[(ω − ω )t + φ].

0

(B.20)

Equation B.20 clearly shows that a 2-pole machine with eccentricity ( > 0) produces a homopolar axial magnetic flux hom (t) = π Bm cos[(ω − ω )t + φ]. This flux passes through the shaft, bearings, and end bells and in addition to the unbalanced magnetic pull can induce bearing currents. Those induced currents can shorten the lifetime of bearings.


DK3193

noise

October 24, 2005

19:9

Appendix C Magnetic Saturation Equation B.5 expresses the resultant relative permeance of the air gap. The saturation effect due to the main (linkage) magnetic flux can be included by introducing an additional permeance of the air gap sat according to Equations 2.102 and 2.103. Thus, magnetic saturation of the ferromagnetic core reduces the permeance of the air gap λg0 . This is equivalent to increasing the air gap by the coefficient ksat > 1, i.e., • for an induction machine λg0 = µ0

1 1 = µ0 g gkC ksat

(C.1)

• for a permanent magnet (PM) synchronous machine λg0 = µ0

1 1 = µ0 g gkC ksat + h M /µrr ec

(C.2)

where geq is the equivalent air gap increased by the equivalent thickness h M /µrr ec of the PM, h M is the PM thickness, g is the air gap (mechanical clearance), kC is Carter’s coefficient according to Equation 2.79, µ0 is the magnetic permeability of free space, and µrr ec , is the relative recoil magnetic permeability of the PM. The saturation factor of the magnetic circuit is defined as

• for an induction machine ksat = 1 +

2V1t + V1c + 2V2t + V2c 2(Bg /µ0 )gkC

(C.3)

• for a PM synchronous machine ksat = 1 +

2V1t + V1c + V2c 2(Bg /µ0 )gkC

333 Copyright © 2006 Taylor & Francis Group, LLC

(C.4)

DK3193

noise

October 24, 2005

19:9

334


where Bg is the air gap magnetic flux density, V1t is the magnetic voltage drop (MVD) along the stator teeth, V2t is the MVD along the rotor teeth,V1c is the MVD along the stator core (yoke) and V2c is the MVD along the rotor core. MVDs V1t , V2t , V1c , and V2c are calculated as products of magnetic field intensity in each portion of the magnetic circuit times the length of the magnetic circuit. Magnetic field intensities are read on the B–H magnetization curve of the magnetic circuit material for calculated values of the magnetic flux densities. In addition to the air gap permeance, magnetic saturation due to main (linkage flux) affects also the parameter α in Equation 2.105 which expresses Fourier coefficient bµ of the rotor magnetic flux density (Equation 2.104). The magnetic saturation slightly reduces the parameter α as follows αsat ≈

1 1/3 ksat

α.

(C.5)

There is also an influence of the magnetic saturation due to the stator leakage flux on the width of the stator slot opening b14 . As the stator tooth heads become saturated, the relative magnetic permeability of the tooth heads is very small. This effect is equivalent to an increase in the stator slot opening width. Magnetic saturation due to the stator leakage flux can be approximately included according to Norman’s method [167]. According to Norman, the magnetomotive force (MMF) of a single stator slot of an induction motor is expressed as: Fsl = 0.707Ia

Nc w 1 ap

0.75

wc s1 + 0.25 + k p1 kw1 τ s2

(C.6)

where Nc is the number of conductors in a coil, w 1 is the number of stator winding layers, a p is the number of stator winding parallel paths, w c is the stator coil pitch, τ is the pole pitch, k p1 is the stator winding pitch factor for fundamental, kw1 is the stator winding factor for fundamental, s1 is the number of stator slots, and s2 is the number of rotor slots. For a synchronous motor s2 = 2 p, where p is the number of pole pairs. For the MMF Fsl a fictitious leakage magnetic flux density in the air gap that saturates the tooth head is calculated, i.e., Bfg =

1.6 ×

106 g

Fsl

0.64 + 2.5

g t1 +τ

(C.7)

where t1 is the stator slot pitch and τ is the rotor pole pitch, in PM synchronous machines equivalent to the rotor slot pitch. The fictitious magnetic flux density in the air gap B f g can take very large unrealistic values, up to 50 T. The saturation factor κsat due to leakage flux is a function of B f g and can be approximated as follows 5.25 κsat ≈ + 0.25. (C.8) 7 + B 2f g


DK3193

noise

October 24, 2005

19:9

Appendix C

335

The increase in the slot opening due to magnetic saturation t1

b14 = − b14 (1 − κsat ). 3

(C.9)

The slot width with magnetic saturation taken into account b14sat = b14 + b14 where b14 is the slot opening of an unsaturated stator.


(C.10)

DK3193

noise

October 24, 2005

19:9

Appendix D Basics of Vibration D.1

A mass–spring–damper oscillator

A very basic vibration system is the classical harmonic oscillator which consists of two basic elements, a mass and a spring [122]. With an initial disturbance, such a system will vibrate freely and maintain a steady-state vibration without any decay in the amplitude since no energy dissipation mechanism is included in the system. In practice, however, all vibration systems have various energy dissipation mechanisms, such as friction, and material viscosities. This means that unless there is extra energy fed into the system continuously, the vibration will ‘die’ out eventually. Given as an example, a mass–spring–damper system subject to a steady-state force excitation is shown in Figure D.1. The motion equation of the system is, M

d 2 X (t) d X (t) + K X (t) = F(t) +C 2 dt dt

(D.1)

where M is the mass, K is the stiffness of the spring, and C is the damping coefficient. Assume that the system is subject to a sinusoidal force F0 e jωt . For steady-state solutions, we can expect the displacement of the mass being of the form X 0 e jωt . Then Equation D.1 becomes, X 0 (−ω2 M + jωC + K )e jωt = F0 e jωt

(D.2)

The displacement of the mass is thus yielded as X0 =

K−

F0 ω2 M

+ jωC

=

ω

F0 e jϕ (K /ω − ωM)2 + C 2

where tan ϕ =

K /ω − ωM . C 337


(D.3)

DK3193

noise

October 24, 2005

19:9

338

Noise of Polyphase Electric Motors F(t) = F0e jw t X(t) = X0e jw t M

C

K

Figure D.1 Mass–spring–damper oscillator. It is seen that the system response, including the magnitude and the phase angle, is generally frequency and system parameter dependent. Since the velocity of the mass is related to the displacement as V (t) = jωX 0 e jωt , the impedance of the system can be obtained as, F(t) K Z (ω) = =C+ j − ωM . (D.4) V (t) ω By setting the imaginary part of the impedance to zero, the resonant frequency of √ the system can be obtained as ω0 = K /M. This frequency is corresponding to the free vibration of a mass-spring oscillator with no damping mechanism considered. Then the solution of Equation D.1 can be rewritten as X (t) =

F0 e jωt 2 M (ω0 − ω2 ) + jηω0 ω

(D.5)

where η = C/(ω0 M) is the loss factor of the system. It is interesting to examine the energy dissipation within the system. The time averaged input power is Pi =

1 1 1 e(F V ∗ ) = e(Z )|V |2 = Cω2 |X 0 |2 . 2 2 2

(D.6)

It is seen that the input power is directly related to the damping C of the system. It is not difficult to derive that the energy consumed by the damping C is equal to the power fed into the system, i.e., Pd = Pi . This is not surprising though because the power balance is the inevitable consequence of a steady-state vibration process. The time averaged total vibration energy of the system is the sum of the time averaged kinetic and potential energy, i.e., E =

1 1 1 Me(V V ∗ ) + K e(X X ∗ ) = (ω2 M + K )|X 0 |2 . 4 4 4


(D.7)

DK3193

noise

October 24, 2005

19:9

Appendix D

339

When ω = ω0 , i.e., the system is at resonance, the time averaged kinetic energy is equal to the time averaged potential energy. Then, E =

1 1 K |X 0 |2 = ω02 M|X 0 |2 . 2 2

(D.8)

The well-known relationship between the dissipated power and the vibrational energy of the system thus can be obtained, Pd =

D.2

1 Cω02 |X 0 |2 = ω0 η E . 2

(D.9)

Lumped parameter systems

The lumped parameter system generally refers to a simplified modeling process in which the spatial dependence of the variable under consideration can be neglected. For mechanical structures, the distributed mass or inertia is replaced by a finite number of lumped masses or rigid bodies which are connected by massless elastic and damping elements, such as the spring and resistance. The basic principle involved in such simplifications is that a change in the variable, such as vibration displacement or velocity, is considered equal and simultaneous for each element. This, therefore, indicates that the approach is only valid at low frequencies where the wavelength is much longer than the dimension of the structure, so that the spatial variations in a variable are small. Depending on the complexity of the structure and the modeling objectives, the lumped parameter model can be as simple as a mass–spring–damper oscillator, or may have many mass/rigid body, spring, damper elements connected in a complex way. Mathematically, the minimum number of the coordinates necessary to describe the motion of each lumped mass or rigid bodies defines the number of degree of freedom of the system. Corresponding to each degree of freedom, there is one ordinary differential equation, as discussed in Section D.1. It can be seen that one of the advantages for the lumped parameter model is that, instead of solving the partial differential equation for a continuous system, it only deals with a set of ordinary differential equations. The motion equations of a lumped parameter system can be generally obtained from Newton’s second law of motion. Assuming that xi and θi are the generalized coordinates of the system representing the transverse and rotational movement respectively, the motion equations are of the form, m i x¨ i = j Fi j (for mass m i ) (D.10) Ji θï = j Mi j (for rigid body of inertia Ji ) where j Fi j denotes that sum of all forces acting on mass m i , and j Mi j indicates the sum of moments of all forces acting on the rigid body of mass moment of inertia Ji . Other methods used to establish the motion equations include Lagrange’s equation, and the electromechanical analogy technique. Based on the


DK3193

noise

October 24, 2005

19:9

340


Stator

Winding

End-Shield Fan Rotor

Foot

(a) L q1

Rotor (m1)

y1 kb

kb

ks

ks

Stator (m2)

x1

q2 y2 x2

kf

kf

(b)

Figure D.2 Induction motor and its lumped parameter model: (a) motor; (b) model. motion equations, eigenvalues and eigenvectors of the system can be yielded, from which the natural frequencies and the mode shapes of the system can be extracted. A simple example of using a lumped parameter model for the dynamic analysis of a mechanical structure is given here for a typical medium-sized motor structure, as shown in Figure D.2a. A lumped parameter model which only considers the two rigid motions, namely the transverse and rotational motions of the rotor and the stator in the x y plane can be established and shown in Figure D.2b. In the model, the motor feet are simplified as two identical stiffness elements k f ; and the stator and the rotor are represented by two lumped mass elements m 1 , and m 2 . In the real motor structure, both ends of the rotor are sitting on the rotor bearings which are supported by the end-shields. In the lumped parameter model, the two


DK3193

noise

October 24, 2005

19:9

Appendix D

341

serially coupled spring elements kb and ks are connecting the rotor and the stator, representing the bearing stiffness and the in-plane stiffness of the end-shields. Since each of the rigid element is allowed to have the transverse movement in the x direction and rotations in the z direction, two variables are thus needed to describe the motion for either the rotor or the stator, meaning this is a system with four degrees of freedom. The motion equations of the whole system can be established as, 

m1  0   0 0

0 m2 0 0

0 0 J1 0

  x¨ 1 0  x¨ 2  0  + 0   θ¨1  J2 θ¨2



    2k1 0 −2k1 0 0 x1 −2k1 2(k1 + k2 ) x2 0 0 0  =  +  0 −k1 L 2 /2 θ1 0 0 k1 L 2 /2 0 0 0 −k1 L 2 /2 (k1 + k2 )L 2 /2 θ2

(D.11)

where k1 = ks kb /(ks + kb ), k2 = 2k f . It can be seen in Equation D.11 that the transverse and rotational motions are not coupled. This is because the system, for simplicity, is modeled symmetrically in geometry. As a result, the transverse and the rotational vibrations can be solved independently. For example, the motion equations for the transverse motions are,

m1 0 0 m2

x¨ 1 2k1 x1 0 −2k1 + = . −2k1 2(k1 + k2 ) x2 0 x¨ 2

For harmonic solutions,

x1 x2

=

A1 e jωt A2 e jωt

(D.12)

(D.13)

where A1 , and A2 are the vibration amplitudes of the rotor and the stator respectively. Substituting Equation D.13 into Equation D.12,

2k1 − ω2 m 1 −2k1 −2k1 2(k1 + k2 ) − ω2 m 2

A1 A2

=

0 . 0

(D.14)

For nonzero solutions of A1 , and A2 , the determinant has to be zero, i.e., 2k1 − ω2 m 1 −2k1 = 0. (D.15) −2k1 2(k1 + k2 ) − ω2 m 2 The roots of this equation are, therefore, the eigenvalues of the system. 2 ω1,2

=

B±


2 2 B 2 − 4ω11 ω22 2

(D.16)

DK3193

noise

October 24, 2005

19:9

342


2 2 2 2 where B = ω11 + ω22 + 2k1 /m 2 , ω11 = 2k1 /m 1 , ω22 = 2k2 /M2 . The two natural frequencies of the system can be obtained as, 2 2 B ± B 2 − 4ω11 ω22 ω1,2 = . (D.17) 2

For the rotational motions, it can be easily shown that the two natural frequencies 2 2 take the same expressions as Equation D.17 but with B = ω11 +ω22 +k1 L 2 /(2J2 ), 2 2 2 2 ω11 = k1 L /(2J1 ), ω22 = k2 L /(2J2 ).

D.3

Continuous systems

Strictly speaking, all mechanical systems are continuous, being with the mass and stiffness distributed evenly over the system. Due to the complex relationship between the stress and the strain of elastic materials, various types of waves, such as longitudinal, transverse, torsional, and flexural (bending) waves may exist in the system [37]. Consequently, the motion equations that govern the vibration behaviors might be different for different types of vibrations. However, all the motion equations are of the type [190], ∂ ∂ ∂ ∂ 2ξ λ·L , , ξ +ρ 2 =0 (D.18) ∂ x ∂ y ∂z ∂t where ξ is the vibration displacement, ρ is the material density, and L(·) is a differential operator, λ is a stiffness related parameter—both of which takes different forms for different wave types. For example, for a beam in a longitudinal vibration along the x direction, ∂ ∂ ∂ ∂2 λ·L , , (D.19) =E 2 ∂ x ∂ y ∂z ∂x where E is Young’s modulus of the material. If this beam is in bending motion, ∂ ∂ ∂ E I ∂4 λ·L , , (D.20) = ∂ x ∂ y ∂z S ∂x4 where S is the area of the cross section of the beam, and I is the second moment of area of the cross section. For a plate embedded in the x y plane in bending motion, λ·L

∂ ∂ ∂ , , ∂ x ∂ y ∂z

Eh 2 = 12(1 − µ2 )

∂2 ∂2 + ∂x2 ∂ y2

2 (D.21)

where h is the thickness of the plate, and µ is Poisson’s ratio of the material. For a cylindrical shell, the differential operator will take a form that is even more complicated due to the coupling between different types of vibrations.


DK3193

noise

October 24, 2005

19:9

Appendix D

343

It is not always an easy task to solve the motion equations for a particular structure. Only simple and regular structures, such as a beam, a round or a rectangular plate might have analytical solutions. Here, a beam in bending motion is given as an example. E I ∂ 4 ξ(x, t) ∂ 2 ξ(x, t) + ρ = 0. S ∂x4 ∂t 2

(D.22)

For sinusoidal solutions, ξ(x, t) = ξ0 (x)e jωt . The motion equation can be rewritten as, d 4 ξ0 (x) + k 4 ξ0 (x) = 0 (D.23) dx4 where k = ω1/2 (E I /ρ S)−1/4 is the wavenumber of the bending wave in the beam. It is seen that unlike the sound wave in the air, the speed of the bending wave (c = ω/k) in a beam depends on the frequency. The general solution of this differential equation is of the form, ξ0 (x) = Ae− jkx + Be jkx + Ce−kx + Dekx

(D.24)

where A, B, C, D are constants depending on the boundary conditions at both ends. Generally, the first two terms represent two approaching waves traveling in the opposite direction, while the last two terms represent the fields that decrease exponentially with distance from the two ends, thus namely the “near field” solutions. There are three typical boundary conditions in the vibration analysis, namely simply supported, clamped, and free. For a beam, these boundary conditions can be described mathematically as, d 2 ξ0 (x) ξ0 (x)|x=x0 = 0, = 0 for a simply supported end d x 2 x=x0 dξ0 (x) ξ0 (x)|x=x0 = 0, = 0 for a clamped end d x x=x0 d 3 ξ0 (x) d 2 ξ0 (x) = 0, = 0 for a free end. d x 3 x=x0 d x 2 x=x0 These mathematical descriptions are associated with the behavior of four typical field variables for bending waves, the transverse displacement ξ0 (x), angular displacement dξ0 (x)/d x, bending moment ∼ d 2 ξ0 (x)/d x 2 , and the shear force ∼ d 3 ξ0 (x)/d x 3 . With the simply supported condition at both ends of the beam (x0 = 0, l), we have, A = −B,

C = D = 0,

sin kl = 0.

(D.25)

This equation indicates that in order to comply with the boundary conditions, the wavenumber has to take a set of discrete numbers, namely the modal wavenumbers, as, nπ kn = , n = 1, 2, 3, · · · . (D.26) l


DK3193

noise

October 24, 2005

19:9

344


The corresponding set of discrete frequencies, namely the natural frequencies of a simply supported beam, are E I 1/2 nπ 2 ωn = , n = 1, 2, 3, · · · . (D.27) ρS l Associated with each natural frequency ωn , the vibration displacement takes a particular pattern ξ0n (x) = sin(nπ x/l), namely the mode shape functions. In vibration, a mode shape function together with its corresponding natural frequency, and modal wavenumber comprise a so-called “vibration mode.” The vibration response of the beam is, therefore, the linear superposition of these vibration modes. ∞ ∞ nπ ξ0 (x) = An ξ0n (x) = An sin x (D.28) l n=1 n=1 where An are a set of constants determined by the external excitations. If the two ends of the beam are clamped, the mode shape of the beam can be derived as, cosh(kn l) − cos(kn l) [sinh(kn x) + sin(kn x)] sinh(kn l) − sin(kn l) (D.29) where the modal wavenumbers kn are the solution of, ξ0n (x) = cosh(kn x) + cos(kn x) −

cosh(kn l) · cos(kn l) = 1.

(D.30)

The roots of this equation can be shown as 4.73, 7.85, 11.0, 14.14, · · ·, which can be represented approximately by, kn ≈

π nπ − , l 2l

n = 1, 2, 3, · · · .

(D.31)

The results shown here clearly indicate that the modal behaviors, for example, natural frequencies, mode shapes, of a structure very much depend on the boundary conditions. Since the modal behaviors are generally independent of the external excitations, they are the inherent characteristics of a structure in terms of the structural dynamics. In most cases, the forced vibration is of greatest interest. A more general equation of motion with the energy dissipation mechanism considered by introducing the complex stiffness parameter is given as, ∂ ∂ ∂ ∂ 2ξ λ(1 + jη) · L , , (D.32) ξ +ρ 2 = f ∂ x ∂ y ∂z ∂t where f is the distributed force density over/within the structure. For steady-state harmonic solutions ( f = f 0 e jωt , ξ = ξ0 e jωt ), it becomes ∂ ∂ ∂ , , ξ0 − ρω2 ξ0 = f 0 λ(1 + jη) · L (D.33) ∂ x ∂ y ∂z


DK3193

noise

October 24, 2005

19:9

Appendix D

345

Note that the general solution of this equation is of the form ξ0 = ∞ n=1 An ξ0n where ξ0n are the mode shape functions which are determined by boundary conditions and satisfy the relationship, ∂ ∂ ∂ λ·L , , (D.34) ξ0n − ρω2 ξ0n = 0. ∂ x ∂ y ∂z Equation D.33 can be rewritten as, ∞

An ρ[ω02 (1 + jη) − ω2 ]ξ0n = f 0 .

(D.35)

n=1

By mutilplying Equation (D.35) with ξ0m and taking integration over the solution domain (τ ), all terms on the left-hand side vanish due to orthogonality replations among the mode shape functions, except the term for which n = m. Then, 2 An [ω02 (1 + jη) − ω2 ] ρξ0n dτ = f 0 ξ0n dτ. (D.36) τ

τ

The solution to Equation D.32 thus can be obtained as, ξ=

∞ n=1

ξ0n e jωt 2 n [(ω0 − ω2 ) + jηω02 ]

τ

f 0 ξ0n dτ

(D.37)

2 where n = τ ρξ0n dτ is called “norm.” It can be found that the response of a continuous system is a linear superposition of all the modes. Each mode is of the type of a simple oscillator as described in Equation D.5. The contribution of each mode to the total response, also namely the participation factor ( An ), primarily depends on the distribution of external forces with regard to the mode shape functions, and how far away the corresponding natural frequency is from the exciting frequency.


DK3193

noise

November 3, 2005

16:24

Symbols and Abbreviations A Am Amr a ap aw B Bm b bs b f sk bp bsk C c cE c0 cp cT D E Ef Ei e ed ef F F f fc fr fring G g

line current density, A/m peak value of the line current density, A/m magnitude of radial forced vibration displacement radius of cylindrical shell number of parallel current paths of the stator winding number of parallel conductors of the stator winding vector magnetic flux density, T peak value of the magnetic flux density, T instantaneous value of the magnetic flux density; width of slot skew of stator slots skew of PMs pole shoe width skew of stator slots damping coefficient wave velocity; tooth width armature constant (EMF constant) sound velocity in the air structural wave speed torque constant diameter EMF, r ms value; Young modulus (elasticity modulus) EMF per phase induced by the rotor without armature reaction averaged vibrational energy of the subsystem i instantaneous EMF; static eccentricity dynamic eccentricity instantaneous EMF excited by PMs in the stator phase winding force magnetomotive force (MMF) frequency critical frequency; center frequency of the band natural frequency of the r th order ring frequency pearmeance, H; shear modulus air gap (mechanical clearance) 347


DK3193

noise

November 3, 2005

16:24

348

g H h hM hm I I1 I2 i J j1 K k kC kd1 k f skµ ki k p1 ksat ksk ksµ kw1 k0 L LI Li LP LW l1e lM M m m1 m2 N N1 N2 Ncm n nm P Pelm

Symbols and Abbreviations equivalent air gap magnetic field intensity, A/m height height of the PM magnification factor current; sound intensity; moment of inertia stator current rotor current instantaneous value of current moment of inertia current density in the stator winding, A/m2 lumped stiffness, N /m coefficient (general symbol) Carter’s coefficient stator winding distribution factor for fundamental space harmonic PM skew factor stacking factor of laminations stator winding pitch factor for fundamental space harmonic saturation factor of the magnetic circuit due to the main (linkage) magnetic flux stator slot skew factor referred to the pole pitch τ rotor PM skew factor referred to the pole pitch τ stator winding factor kw1 = kd1 k p1 for fundamental harmonic acoustic wave number k02 = k x2 + k 2y in the air inductance; length sound intensity level (SIL), dB effective length of the stator stack sound pressure level (SPL), dB sound power level (SWL), dB length of the one-sided end connection axial length of PM mutual inductance; lumped mass; modal overlap mass; circumferential mode number of stator phases number of rotor phases number of turns number of stator turns in series per phase number of rotor turns in series per phase smallest common multiple number of a higher time harmonic; rotational speed in rpm or rev/s; axial mode; modal density mechanical speed of the rotor (shaft) in rpm or rev/s; active power; acoustic pressure electromagnetic power


DK3193

noise

November 3, 2005

16:24

Symbols and Abbreviations Pi P p pr Q1 q1 R R1 r S SM s s1 s2 T Tc Td T0 Tp Tr Tsh t t1 tr u V V1 v W wM X Z α αi β γ d η λ µ µ0

349

averaged input power to subsystem i active power losses number of pole pairs; sound pressure radial magnetic force (magnetic pressure) of the r th order per unit area number of stator slots per pole number of stator slots per pole per phase resistance; radius stator winding resistance magnetic force order apparent power; surface cross section area of PM S M = w M L M or S M = b p L M slip for fundamental harmonic number of stator teeth or slots number of rotor teeth or slots torque cogging torque developed torque constant of avarage component of the torque period periodic component of the torque shaft torque (output or load torque) time stator slot (tooth) pitch t1 = π D1in /s1 relative torque ripple vibration displacement electric voltage; volume stator input voltage per phase instantaneous value of electric voltage; linear velocity magnetic field energy, J width of PM reactance √ impedance Z = R + j X ; | Z |= Z = R 2 + X 2 electrical angle α = π x/( pτ ); coefficient of approximation of the distribution of the magnetic flux density above the rotor pole effective pole arc coefficient αi = b p /τ overlap angle of pole constant dependent on slot opening and air gap relative static eccentricity relative dynamic eccentricity efficiency; damping loss factor relative permeance ( = G/S), H/m2 relative specific permeance (dimensionless); wavelength number of the rotor µth harmonic magnetic permeability of free space µ0 = 0.4π × 10−6 H/m


DK3193

noise

November 3, 2005

16:24

350 µr µr ec µrr ec ν ρ σ σf τ f φ ζ m ω

Symbols and Abbreviations relative magnetic permeability recoil magnetic permeability relative recoil permeability µrr ec = µr ec /µo number of the stator νth space harmonic; Poisson’s ratio sound power, W specific mass density radiation efficiency; electric conductivity form factor to include the saturation effect pole pitch τ = π D1in /(2 p) magnetic flux, Wb excitation magnetic flux; PM flux, Wb power factor angle; phase angle of the magnetic flux density wave flux linkage = N modal damping ratio (see also η = damping loss factor) angular speed = 2πn mechanical angular speed of the rotor m = 2π n m ; roots of characteristic equation angular frequency ω = 2π f Subscripts

A, B, C phase A, B, or C elm electromagnetic f field, frame g air gap in inner l leakage M magnet m peak value (amplitude) n number of time harmonic; normal components out output, outer s slot; synchronous sat saturation t tooth; tangential components x, y, z Cartesian coordinate system eccentricity 1 stator; fundamental harmonic 2 rotor Superscripts (e) (tr ) (sq)

element trapezoidal square wave


DK3193

noise

November 3, 2005

16:24

Symbols and Abbreviations Abbreviations ANC a.c. BEM CAD CPU DFT DSP d.c. EMF FEM FFT HVAC MFD MMF MVD ODP PM PMBM PSD PWM RMA SEA SIL SED SPL SSC SWL TEFC VVVF WPII

active noise control alternating current boundary-element method computer-aided design central processor unit discrete Fourier transform digital signal processor direct current electromotive force finite element method fast Fourier transform heating, ventilating, and air conditioning magnetic flux density magnetomotive force magnetic voltage drop open drip proof permanent magnet permanent magnet brushless motor power spectral density pulse width modulation reaction mass actuator statistical energy analysis sound intensity level sound energy density sound pressure level solid state converter sound power level totally enclosed fan cooled variable voltage variable frequency water protected type II


351

DK3193

noise

November 3, 2005

16:24

Bibliography [1] Ackerman, B., Janssen, J.H.H., Sottek, R., and van Steen, R.I.: New technique for reducing cogging torque in a class of brushless d.c. motors, IEE Proc. Pt. B, 139(4), 1992, pp. 315–320. [2] Amor, A.H., Timar, P.L., and Poloujadoff, M.: Induction squirrel cage machine design with minimization of electromagnetic noise, IEEE Trans. on EC, Vol. EC-10(4), 1995, pp. 681–687. [3] Astfalck, A., Pulle, D.W.J., and Lai, J.C.S.: Prediction of transient stator force distribution of inverter driven induction motors, AUPEC/EECON’97, Sydney, Australia, 1997, Vol. 1, pp. 1–6. [4] Australian Standard AS 1259 Sound level meters, 1982. [5] Australian Standard AS 2659 Guide to the use of sound measuring equipment, Part 1: Portable sound level meters, 1988; Part 2: Portable equipment for integration of sound signals, 1983. [6] Australian Standard AS 2659. Guide to the use of sound measuring equipment: Portable sound level meters, 1988. [7] Australian Standard AS 2679 Vibration and shock – Mechanical vibration of rotating and reciprocating machinery – Requirements for instruments for measuring vibration severity, 1984. [8] Australian Standard AS 2775 Vibration and shock – Mechanical mounting of accelerometers, 1985. [9] Australian Standard AS 2973 Vibration and shock – Human response vibration-measuring instrumentation, 1987. [10] Astakhov, N.V., Malishev, V.S., and Ovcharenko, N. J.: Magnetic Vibration of Asynchronous Motors (in Russian), Stinca, Kiˇsiniev, 1985. [11] Bies, D. and Hansen, C.H.: Engineering Noise Control, Unwin Hyman, 1988. [12] Bauer, B.B.: A century of microphones, J. Audio Eng. Soc., Vol. 35(4), 1987, pp. 246–258. [13] Belkhayat, D., Roger, D., and Brudny, J.F.: Active reduction of magnetic noise in asynchronous machine controlled by stator current harmonics, Int. Electrical Machines and Drives Conf. EMD’97, Cambridge, U.K., IEE Conf. Publication No. 444, 1997, pp. 400-405. [14] Bendat, J.S. and Piersol, A.: Random data: Analysis and measurement procedures, John Wiley & Sons, 1971. 353 Copyright © 2006 Taylor & Francis Group, LLC

DK3193

noise

November 3, 2005

16:24

354

Bibliography

[15] Berry, A.: A general formulation for the sound radiation from rectangular, baffled plates with arbitrary boundary conditions, J. Acoust. Soc. Am., Vol. 88(6), 1990, pp. 2792–2802. [16] Bianchi, N. and Bolognani, S.: Design techniques for reducing the cogging torque in surface mounted PM motors, IEEE Trans. IA, Vol. 38(5) 2002, pp. 1259–1265. [17] Bies, D.A. and Hamid, S.: In situ determination of loss and coupling loss factors by the power injection method, J. Sound and Vibration, Vol. 70(2), 1980, pp. 187–204. [18] Breton, C., Bartolome, J., Benito, J.A., Tassinario, G., Flotats, I., Lu, C.W., and Chalmers, B.J.: Influence of machine symmetry on reduction of cogging torque in permanent magnet brushless motors, IEEE Trans. MAG Vol. 36(5), 2000, pp. 3819–3823. [19] Broch, J.T.: Mechanical Vibration and Shock Measurements, Brüel & Kjaer Handbook, 1984. [20] Brunsbach, B.J., Henneberger, G., and Klepach, T.: Compensation of torque ripple, Int. Conf. on Electr. Machines ICEM’98, Istanbul, Turkey, 1998, pp. 588–593. [21] Bruskin, D.E., Zorokhoviˇc, A.E., and Khvostov, V.C.: Electrical Machines, ˇ (in Russian), Part 1, Moscow, Vyzˇsaya Skola, 1979. [22] Buckens, F.: Influence of the relative radial thickness of a ring on its natural frequencies, J. Acoust. Soc. Am. Vol. 22(4), 1950, pp. 437–443. [23] Burroughs, C.B., Fischer, R.W., and Kern, F.K.: An introduction to statistical energy analysis, J. Acoust. Soc. Am. Vol. 101(4), 1997, pp. 1779–1789. [24] Buzdugan, G., Mihailescu, E., and Rades, M.: Vibration Measurement, Martinus Nijhoff Publishers, 1986. [25] Cacciolati, C. and Guyader, J.L.: Measurement of SEA coupling loss factors using point mobilities, Phil. Trans. Royal Society London A, Vol. 346, 1994, pp. 465–475. [26] Cai, W., Fulton, D., and K. Reichert, K.: Design of permanent magnet motors with low torque ripple, Int. Conf. on Electr. Machines ICEM’00, Espoo, Finland, Vol. 3, 2000, pp. 1384–1388. [27] Carr, I.E. and Lalor, N: The prediction and reduction of noise from engine covers using energy methods, Proc. of Advances in the Control and Refinement of Vehicle Noise, 1988, pp. 159–165. [28] Carlson, R., Lajoie-Mazenc, M., and Fagundes, J.C.S.: Analysis of torque ripple due to phase commutation in brushless d.c. machines, IEEE Trans. IAS 28(3), 1992, pp. 632–638. [29] Chen, Y.D., Fuh, C.C., and Tung, P.C.: Application of voice coil motors in active dynamic vibration absorbers, IEEE Trans. on MAG, Vol. 41(3), 2005, pp. 1149–1154. [30] Chen, S.X., Low, T.S., Lin, H., and Liu, Z.J.: Design trends of spindle motors for high performance hard disk drives, IEEE Trans. on MAG 32(5), 1996, pp. 3848–3850.


DK3193

noise

November 3, 2005

16:24

Bibliography

355

[31] Cho, D.H. and Kim, K.J.: Modeling of electromagnetic excitation forces of small induction motors for vibration and noise analysis, Proc. IEE, Part B: EPA, Vol. 145(3), 1998, pp. 199–205. [32] Chung, J.Y.: Cross-spectral method of measuring acoustic intensity without error caused by instrument phase mismatch., J. Acoust. Soc. Am., Vol. 64, 1978, pp. 1613–1616. [33] Ciskowski, R.D. and Brebbia, C.A.: Boundary Element Methods in Acoustics, Elsevier Applied Science, 1997. [34] Condenser Microphones and Microphone Preamplifiers for Acoustic Measurement, Brüel & Kjaer, 1982. [35] Coyette, J.P. and Migeot, J.L.: An overview of boundary element approaches for acoustic modelling, Internoise 93, 1993, Leuven, Belgium, pp. 709–714. [36] Covic, G.A. and Boys, J.T.: Noise quieting with random PWM a.c. drives, Proc. IEE, Part B: EPA, Vol. 145, 1998, pp. 1–10. [37] Cremer, L., Heckle, M., and Ungar, E.E.: Structure-Borne Sound: Structural Vibrations and Sound Radiation at Audio Frequencies, 2nd edition, Springer-Verlag, New York, 1988. [38] Crighton, D.G. et al.: Modern Methods in Analytical Acoustics, London, Springer-Verlag, 1992. [39] Crocker, M.J. and Price, A.: Sound transmission using statistical energy analysis., J. of Sound and Vibration, Vol. 9, 1969, pp. 469–486. [40] D¸abrowski, M.: Magnetic Field and Circuits of Electrical Machines (in Polish), Warsaw, WNT, 1971. [41] D¸abrowski, M.: Construction of Electrical Machines (in Polish), WNT, Warsaw, 1977. [42] Davies, H.G. and Wahab, M.A.: Ensemble averages of power flow in randomly excited coupled beams., J. of Sound and Vibration, Vol. 77(3), 1981, pp. 311–321. [43] Delaere, K., Iadevaia, M., Heylen, W., Saa, P., Hameyer, H., and Belmans, R.: Statistical energy analysis of acoustic noise vibration for electric motors: transmission from air gap field to motor frame, IEEE IAS 1999 Conf. and Meeting, Oct. 3–7, 1999, pp.1897–1902. [44] Delaere, K., Heylen, W., Belmans, R., and Hameyer, K.: Introducing magnetostriction into coupled magnetomechanical systems, Int. Conf. on Electr. Machines ICEM’2000, Espoo, Finland, 2002, Vol. 1, pp. 174–178. [45] Delaere, K., Heylen, W., Bellmans, R., and Hameyer, K.: Comparison of induction machine stator vibration spectra induced by reluctance forces and magnetostriction, IEEE Trans. on MAG, Vol. 38(2), 2002, pp. 969–972. [46] de La Ree, J. and Boules, N.: Torque production in permanent magnet synchronous motors, IEEE Trans. on IA Vol. 25(1), 1989, pp. 107–112. [47] Demenko, A.: Time-stepping FE analysis of electric motor dirves with semiconductor converters, IEEE Trans. on MAG, Vol. 30(5), 1994, pp. 3264– 3267. [48] Deodhar, R.P., Staton, D.A., Jahns, T.M., and Miller, T.J.E.: Prediction of cogging torque using the flux-MMF diagram technique, IEEE Trans. on IA Vol. 32(3), 1996, pp. 569–576.


DK3193

noise

November 3, 2005

16:24

356

Bibliography

[49] Dimitriadis, E.K. and Pierce, A.D.: Analytical solution for the power flow exchange between strongly coupled plates under random excitation: A test of statistical energy analysis concepts, J. of Sound and Vibration, Vol. 123(3), 1988, pp. 397–412. [50] Donnell, L.H.: A discussion of thin shell theory, 5th Int. Congress of Appl. Mechanics, Cambridge, MA, U.S.A., 1938, p. 66. [51] Dreyfus, L.: Die Stromwendung Grosse Gleichstrommaschinen, Berlin, Germany, 1929. [52] Elliot, S.: Signal Processing for Active Control, Academia Press, London, U.K., 2001. [53] Ewins, D.J.: Modal Testing: Theory and Practice. Research Studies Press Ltd., 1984. [54] Fahy, F.J.: Measurement of acoustic intensity using the cross-spectral density of two microphone signals, J. Acoust. Soc. Am., Vol. 64, 1977, pp. 1057– 1059. [55] Fahy, F.: Sound and Structural Vibration — Radiation, Transmission, and Response, Academic Press, London, U.K., 1987. [56] Fahy, F. and Mohammed, A.D.: A study of uncertainty in applications of SEA to coupled beam and plate systems. Part 1: Computational experiments, J. of Sound and Vibration, Vol. 158(1), 1992, pp. 45–67. [57] Fahy, F.: Statistical energy analysis: a critical overview, Phil. Trans. R. Soc. Lond. A, Vol. 346, 1994, pp. 431–447. [58] Fahy, F.J.: Sound Intensity, MacMillan, 1995. [59] Favre, E., Cardoletti, L., and Jufer, M.: Permanent magnet synchronous motors: a comprehensive approach to cogging torque suppression, IEEE Trans. on IA Vol. 29(6), 1993, pp. 1141–1149. [60] Finley, W.R.: Noise in induction motors — causes and treatments, IEEE Trans. on IA, Vol. 27(6), 1991, pp. 1204–1213. [61] Finley, W.R. and Burke, R.R.: Proper specifications and installation of induction motors, IEEE IA Magazine, Vol. 3(1), 1997, pp. 56–69. [62] Fredo, C.R.: A SEA-like approach for the derivation of energy flow coefficients with a finite element model, J. of Sound and Vibration, Vol. 199(4), 1997, pp. 645–666. [63] Freise, W. and Jordan, H.: Einseitige magnetische Zugkräfte in Drehstrommaschinen. etz-Archiv., Vol. 83(9), 1962, pp. 299–303. ¨ [64] Frohne, H.: Uber die primären Bestimmungsgrössen der Lautstärke bei Asynchronmaschinen, PhD Thesis, Technischen Hochschule Hannover, Fakultät für Maschinenwesen, Hannover, Germany, 1959. [65] Frydrychowicz-Jastrzebska, G.: Influence of slot skewing on reduction of noise level in single-phase motors (in Polish), Przeglad Elektrotechniczny, Vol. 68(1), 1992, pp. 11–14. [66] Fujiwara, A.: Experimental analysis of screw compressor noise and vibration. 8th Int. Compressor Eng. Conf., Purdue University, West Lafayette, LA, U.S.A., 1986, pp. 566–583.


DK3193

noise

November 3, 2005

16:24

Bibliography

357

[67] Fyfe, K.R. and Ismail, F.: An investigation of the acoustic properties of vibrating finite cylinders, J. Sound Vib., Vol. 128(3), 1989, pp. 361–375. [68] Gade, S.: Sound intensity, Part 1: Theory, Brüel & Kjaer Technical Review, No. 3, 1982, pp. 3–39. [69] Gade, S.: Sound intensity, Part 2: Instrumentation and applications, Brüel & Kjaer Technical Review, No. 4, 1982, pp. 3–32. [70] Gagliardini, L., Houillon, L., Petrinelli, L., and Borello, G.: Virtual SEA: mid-frequency structure-borne noise modeling based on finite element analysis, SAE Noise and Vibration Conference, Traverse City, MI, U.S.A., 2003, Paper No. 2003-01-1555. [71] Garcia-Otero, S. and Devaney, M.: Minimization of acoustic noise in variable speed induction motors using a modified PWM drive, IEEE Trans. on IA, Vol. 30(1), 1994, pp. 111–115. [72] Garvey, S.D.: The vibrational behaviour of laminated component in electrical machines, 4th Int. Conf. on Electr. Machines and Drives, IEEE, London, U.K., 1989, pp. 226–231. [73] Gieras, J.F.: Electrodynamic forces in electromagnetic levitation systems, Acta Technica CSAV, Prague, Czechoslovakia, 1982, No. 5, pp. 532–545. [74] Gieras, J.F.: Analysis of an inverter-fed double-cage induction motor, South African Universities Power Eng. Conf. SAUPEC’92, Durban, South Africa, 1992. [75] Gieras, J.F. and Wing, M.: Permanent Magnet Motors Technology: Design and Applications, 2nd edition, Marcel Dekker, New York, 1996, 2002. [76] Gieras, J.F. and Marler, M.E.: Analytical prediction of torque ripple in permanent magnet brushless motors, Int. Conf. on Electr. Machines ICEM’02, Brugge, Belgium, 2002, pp. 33–34. [77] Gieras, J.F.: Analytical approach to cogging torque calculation of permanent magnet brushless motors, IEEE Trans. on IA, Vol. 40(5), 2004, pp. 1310–1316. [78] Goldie, J.H.: Reaction Mass Actuators at SatCon, SatCon Publication Boston, MA, U.S.A. 2002. [79] Graham, Q., Beckwith, S., and Milliken, F.H.: Magnetic noise in synchronous machines, AIEE Trans., Sept. 1931, pp. 1056–1062. [80] Greiner, H.: Frequency inverters — motors (in German), EMA Magazine, 1997. [81] Grigoriev, N.V.: Vibration of Power Mechanics, Handbook (in Russian), Machinostroenie, Leningrad, 1974. [82] Habetler, T.G. and Divan, D.M.: Acoustic noise reduction in sinusoidal PWM drives using a randomly modulated carrier, IEEE Trans. on PE, Vol. 6(3), 1991, pp. 356–363. [83] Hanselman, D.C.: Effect of skew, pole count and slot count on brushless motor radial force, cogging torque and back EMF, IEE Proc. 144 (5), 1997, pp. 325–330. [84] Hansen, C.H.: Understanding Active Noise Cancellation, Spon Press (Taylor & Francis Group), London – New York, 2001.


DK3193

noise

November 3, 2005

16:24

358

Bibliography

[85] Harris, C. M.: Shock and Vibration Handbook, 3rd edition, McGraw Hill, 1988. [86] Hashemi, N.N. and Lisner, R.P.: A new strategy for active control of acoustic noise in converter-fed induction motors, Australian Universities Power Eng. Conf. AUPEC’2002, Melbourne, VIC, Australia, 2002. [87] Heller, B. and Hamata, V.: Harmonic Field Effects in Induction Machines, Academia (Czechoslovak Academy of Sciences), Prague, 1977. [88] Hickiewicz, J., Lukaniszyn, M., Szymaniec, S., and Wach, P.: On reducing the vibration and noise level of induction machines, Int. Conf. on Electr. Machines ICEM’88, Pisa, Italy, 1988, pp. 25–29. [89] Holmer, C.I. and Heymann, F.J.: Transmission of sound through pipe walls in the presence of flow, J. of Sound and Vibration, Vol. 70(3), 1980, pp. 275– 301. [90] Hoppe, R.: Vibrationen eines Ringes in seiner Ebene, J. Math. (Crelle), Vol. 73, 1871. [91] Hopkins, C.: Statistical energy analysis of coupled plate systems with low modal density and low modal overlap, J. of Sound and Vibration, Vol. 251(2), 2002, pp. 193–214. [92] Ishikawa, T. and Slemon, G.: A method of reducing ripple torque in permanent magnet motors without skewing, IEEE Trans. on MAG 29(3), 1993, pp. 2028–2033. [93] IEC 60034-9 Standard Rotating electrical machines, Part 9: Noise limits, 3rd edition, International Electrotechnical Commission, 1997. [94] IEC 60034-9 Standards Rotating electrical machines — Part 9: Noise limits, 4th edition, International Electrotechnical Commission, 2003. [95] IEC 60942 Standards Ed. 3.0 Electroacoustics — Sound calibrators, International Electrotechnical Commission, 2003. [96] IEC 61672-1 Standards Ed. 1.0 Electroacoustics — Sound level meters — Part 1: Specifications, International Electrotechnical Commission, 2002. [97] International Standard ISO 266 Acoustics — Preferred frequencies, 1997. [98] International Standard ISO 2954 Mechanical vibration of rotating and reciprocating machinery: Requirements for instruments for measuring vibration severity, 1975. [99] International Standard ISO 3740 Acoustics — Determination of sound power levels of noise sources: Guidelines for the use of basic standards, 2000. [100] International Standard ISO 3741 Acoustics — Determination of sound power levels of noise sources using sound pressure: Precision methods for reverberation rooms, 1999. [101] International Standard ISO 3742 Acoustics — Determination of sound power levels of noise sources: Precision methods for discrete-frequency and narrow-band sources in reverberation rooms, 1988. [102] International Standard ISO 3744 Acoustics — Determination of sound power levels of noise sources using sound pressure: Engineering method in an essentially free field over a reflecting plane, 1994.


DK3193

noise

November 3, 2005

16:24

Bibliography

359

[103] International Standard ISO 3745 Acoustics — Determination of sound power levels of noise sources using sound pressure: Precision methods for anechoic and hemi-anechoic rooms, 2003. [104] International Standard ISO 3746 Acoustics — Determination of sound power levels of noise sources using sound pressure: Survey method using an enveloping measurement surface over a reflecting plane, 1995. [105] International Standard ISO 3747 Acoustics — Determination of sound power levels of noise sources: Survey method using a reference sound source, 1987. [106] International Standard ISO 3747 Acoustics — Determination of sound power levels of noise sources using sound pressure — Comparison method in situ, 2000. [107] International Standard ISO 9614-1 Acoustics — Determination of sound power levels of noise sources using sound intensity — Part 1: Measurement at discrete points, 1993. [108] International Standard ISO 9614-2 Acoustics — Determination of sound power levels of noise sources using sound intensity — Part 2: Measurement by scanning, 1996. [109] International Standard ISO 9614-2 Acoustics — Determination of sound power levels of noise sources using sound intensity — Part 3: Precision method for measurement by scanning, 2002. [110] International Standard ISO 5348-3 Mechanical vibration and shock — Mechanical mounting of accelerometers, 1998. [111] International Standard ISO 9614 Acoustics — Determination of the sound power levels of noise sources using sound intensity measurements, 1993. [112] International Standard ISO 3745 Acoustics — Test code for the measurement of airborne noise emitted by rotating electrical machines, 1999. [113] Ishibashi, F., Noda, S., and Mochizuki, M.: Numerical simulation of electromagnetic vibration of small induction motors, Proc. IEE, Part B: Electric Power Appl., Vol. 145(6), 1998, pp. 528–534. [114] Ishibashi, F., Kamimoto,K., Noda, S., and Itomi, K.: Small induction motor noise calculation, IEEE Trans. on EC, Vol. 18(3), 2003, pp. 357–361. [115] Jang, G.H. and Chang, J.H.: Numerical analysis of the electromechanically coupled magnetic field in brushless d.c. motors, J. of Magnetism and Magnetic Materials, Elsevier, Vol. 226–230, 2001, pp. 1223–1225. [116] Jeyapalan, R.K. and Richards, E.J.: Radiation efficiencies of beams in flextural vibration, J. Sound and Vibration, Vol. 67(1), 1979, pp. 55–67. [117] Jahns, T.M.: Torque production in permanent magnet motor drives with rectangular current excitation, IEEE Trans. on IA, Vol. 20(4), 1984, pp. 803–813. [118] Jiao, G. and Rahn, C.D.: Field weakening for radial force reduction in brushless permanent magnet d.c. motors, IEEE Trans. on MAG, Vol. 40(5), 2004, pp. 3286–3292. [119] Jordan, H.: Geräuscharme Elektromotoren, Verlag W. Girardet, Essen, 1950.


DK3193

noise

November 3, 2005

16:24

360

Bibliography

[120] Kaiser, E.R.: Acoustical vibration of rings, J. Acoust. Soc. Am., Vol. 25(4), 1953, pp. 617–623. [121] Kataja, J. and Antila, M.: Selection of active noise control strategy, Joint Baltic-Nordic Acoustic Meeting JBNAM’04, Mariehamn, Aland, Norway, 2004. [122] Kelly, S.G.: Fundamentals of Mechanical Vibrations, 2nd edition, McGraw-Hill Science/Engineering/Math, 2000. ¨ [123] Ketteler, K.H.: Uber den Einfluss der Wicklungsschaltung in Induktionmaschinen auf die Zusatzverluste und den einsetigen magnetischen Zug. etz-Archiv, Vol. 106(3), 1984, pp. 99–106. [124] Kinsler, L.E., Frey, A.R., Coppens, A.B., and Sanders, J.V.: Fundamentals of Acoustics, John Wiley & Sons, New York, 3rd edition, 1982, 4th edition, 2000. [125] Kowal, J., and Ornacki, R.: Analysis of properties of actuators used in vibration control systems, Int. Carpathian, Conf. ICCC’02, Malenovice, Czech Republic, 2002, pp. 791–796. [126] Kubota, Y. and Dowell, E.H.: Experimental investigation of asymptotic modal analysis for a rectangular plate, J. of Sound and Vibration, Vol. 106(2), 1986, pp. 203–216. [127] Kwasnicki, S.: Magnetic Noise of Cage Induction Motors (in Polish), BOBREME Komel, Katowice, Poland, 1998. [128] Lalor, N.: Practical considerations for the measurement of internal and coupling loss factors on complex structures, ISVR Technical Report No.182, June 1990. [129] Langley, R.S.: A derivation of the coupling loss factors used in statistical energy analysis, J. of Sound and Vibration, Vol. 141(2), 1990, pp. 207–219. [130] Langley, R.S. and Heron, K.H.: Elastic wave transmission through plates/beam junctions, J. of Sound and Vibration, Vol. 143(2), 1990, pp. 241–253. [131] Langley, R.S. and Bremer, P.G.: A hybrid method for the vibration analysis of complex structural-acoustic systems, J. Acoust. Soc. Am., Vol. 106, 1999, pp. 1657–1671. [132] Langley, R.S. SEA: Current and future research needs, Proceedings of the 1st Int. Auto SEA Users Conf., San Diego, CA, U.S.A., July 27–28, 2000, pp. 126–135. [133] Langley, R.S. and Cotoni, V.: Response variance prediction in the statistical energy analysis of built-up systems, J. Acoust. Soc. Am., Vol. 115, 2004, pp. 706–718. [134] Laulagnet, B. and Guyader, J.L.: Modal analysis of shell’s acoustic radiation in light and heavy fluids, J. Sound and Vibration, Vol. 131, 1989, pp. 397–415. [135] Leissa, A.: Vibration of Shells, Acoustic Society of America, American Ins. of Physics, 1993. [136] Li,T. and Slemon, G.: Reduction of cogging torque in permanent magnet motors, IEEE Trans. on MAG, Vol. 24(6), 1988, pp. 2901–2903.


DK3193

noise

November 3, 2005

16:24

Bibliography

361

[137] Lisner, R.P. and Timar, P.L.: A new approach to electric motor acoustic noise standards and test procedures, IEEE Trans. on EC, Vol. 14(3), 1999, pp. 692–697. [138] Lo, W.C., Chan, C.C., Zhu, Z.Q., Xu, L., and Howe, D.: Acoustic noise radiated by PWM-controlled induction machine drives, IEEE Trans. IE 47(4), 2000, pp. 880–889. [139] Lyon, R.H. and Maidanik, G.: Power flow between linearly coupled oscillators, J. Acoust. Soc. Am., Vol. 34(5), 1962, pp. 623–639. [140] Lyon, R.H.: Statistical Energy Analysis of Dynamical Systems: Theory and Applications, MIT Press, Cambridge, MA, U.S.A., 1975. [141] Lyon, R.H. and DeJong, R.G.: Theory and Application of Statistical Energy Analysis, 2nd edition, Butterworth-Heinemann, 1995. [142] Mace, B.R.: On the statistical energy analysis hypothesis of coupling power proportionality and some implications of its failure, J. Sound and Vibration, Vol. 178, 1994, pp. 95–112. [143] Mace, B.R.: Frequency averaging effects in SEA, 3rd Int. Congress on Air- and Structure-Borne Sound and Vibration, Montreal, Canada, 1994, pp. 177–184. [144] Mace, B.R. and Shorter, P.J.: Energy flow models from finite element analysis, J. Sound and Vibration, Vol. 233, 2000, pp. 369–389. [145] Mace, B.R.: Statistical energy analysis, energy distribution modes and system modes, J. Sound and Vibration, Vol. 264, 2003, pp. 391–409. [146] Maidanik, G.: Response of ribbed panels to reverberant acoustic field, J. Acoust. Soc. Am., Vol. 34, 1962, pp. 809–826. [147] Malfait, R., Reekmans, R., and Belmasn, R.: Audible noise and losses in variable speed induction motor drives with IGBT inverters — Influence of design and switching frequency, 29th IEEE IAS Annual Meeting, Denver, CO, U.S.A., Vol. 1, 1994, pp. 693–700. [148] Maliˇsev, V.S., Medvedev, V.T., Karwacki, W., and Kostowski, A.: Influence of eccentricity on magnetic forces producing vibrations in induction motors (in Russian), 25th Symp. of Electr. Machines, Lodz, Poland, 1990, pp. 1–7. [149] Maliti, K.C.: Modeling and analysis of magnetic noise in squirrel-cage induction motors, Royal Inst. of Techn. (KTH), Stockholm, Sweden, 2000. [150] Manning, J.E. and Maidanik, G.: Radiation properties of cylindrical shells, J. Acoust. Soc. Am., Vol. 36, 1964, pp. 1691–1698. [151] Manning, J.E.: Formulation of SEA parameters using mobility functions, Phil. Trans. of Royal Society London, Pt. A, Vol. 346, 1994, pp. 477–488. [152] Martinez-Munoz, D. and Lai, J.C.S.: Acoustic noise prediction in a vector controlled induction machine, Int. Electr. Machines and Drives Conf. IEMDC’03, Madison, WI, U.S.A., 2003, pp. 104–110. [153] Martinez-Munoz, D., Lai, J.C.S., and Alakula, M.: A study of the acoustical properties of stators with an outer-surface layer of air gaps, EPE 2003, Toulouse, France. [154] McDevitt, T.E., Campbell, R.L., and Jenkins, D.M.: An investigation of induction motor zeroth-order magnetic stresses, vibration, and sound radiation, IEEE Trans. on MAG, Vol. 40(2), 2004, pp. 774–777.


DK3193

noise

November 3, 2005

16:24

362

Bibliography

[155] Miller, V.R. and Faulkner, L.L.: Predication of aircraft interior noise using the statistical energy analysis method, ASME, J. of Vibration, Acoustics, Stress and Reliability in Design, Vol. 105, 1983, pp. 512–518. [156] Mohan, N., Undeland, T.M., and Robbins, W.P.: Power Electronics: Converters, Applications, and Design, John Wiley & Sons, New York – Chichester – Brisbane – Toronto – Singapore, 1989. [157] Müller, G.G., Black, R., and Davis, T.E.: The diffraction produced by cylindrical and cubical obstacles and by circular and square plates, J. Acoust. Soc. Am., Vol. 10, 1938, p. 6. [158] Mohammed, O.A., Calvert, T., and McConnel, R.: Coupled magnetoelastic finite element formulation including anisotropic reluctivity tensor and magnetostriction effects for machinery applications, IEEE Trans. on MAG, Vol. 37(5), 2001, pp. 3388–3392. [159] Mohammed, O.A., Liu, S., and Ganu, S.C.: Computation of transient magneto-mechanical problems in electrical machines, IEEE Southeast Conf., Columbia, SC, U.S.A., 2002, pp. 187–191. [160] Morse, P.M. and Ingard, K.U.: Theoretical Acoustics, McGraw-Hill, New York, 1968. [161] Nau, S.L. and Gomez Mello, H.G.: Acoustic noise in induction motors: Causes and solutions, 35th IEEE IAS Annual Conf. and Meeting, Rome, Italy, 2000, pp. 253–263. [162] NEMA General standards applying to all machines — Part 9: Rotating Electrical Machines — Sound power limits and measurement procedures, National Electrical Manufacturers Association, 1998. [163] Neves, C.G.C., Carslon, R., Sadowski, N., Bastos, J.P.A., Soeiro, N.S., and Gerges, S.N.Y.: Experimental and numerical analysis of induction motor vibrations, IEEE Trans. MAG, Vol. 35(3), 1999, pp. 1314–1317. [164] Neves, C.G.C., Carslon, R., Sadowski, N., Bastos, J.P.A., Soeiro, N.S., and Gerges, S.N.Y.: Calculation of electromechanic-mechanic-acoustic behavior of a switched reluctance motor, IEEE Trans. MAG, Vol. 36(4), 2000, pp. 1364–1367. [165] Newland, D.E.: Power flow between a class of coupled oscillators, J. Acoust. Soc. Am., Vol. 43(3), 1968, pp. 553–559. [166] Noda, S., Mori, S., Ishibashi, F., and Itomi, K.: Effect of coils on natural frequencies of stator cores in small induction motors, IEEE Trans. on EC., Vol. 2(1), 1987, pp. 93–99. [167] Norman, H.M.: Induction motor locked saturation curves, AIEE Trans. Electr. Eng., April 1934, pp. 536–541. [168] Norton, M.P.: Fundamentals of Noise and Vibration Analysis for Engineers, Cambridge University Press, Sydney, Australia, 1989. [169] Nowacki, W.: Dynamics of Elastic Systems, John Wiley & Sons, New York, 1963. [170] Olson, H.F.: System responsive to the energy flow of sound waves, U.S. Patent No. 1892644, 1932.


DK3193

noise

November 3, 2005

16:24

Bibliography

363

[171] Operation and Maintenance Manual, Teco Westinghouse Motor Company, Round Rock, TX, U.S.A., 2000. [172] Oppenheimer, C.H. and Dubowsky, S.: A radiation efficiency for unbaffled plates with experimental validation, J. Sound and Vibration, Vol. 199(3), 1997, pp. 473–489. [173] Ostovic, V.: Dynamics of Saturated Electric Machines, Springer-Verlag, New York, 1989. [174] Parrett, A.V., Hicks, J.K., Burton, T.E., and Hermans, L.: Statistical energy analysis of airborne and structure-borne automobile interior noise, Paper No. 971970, SAE Noise and Vibration Conf., Traverse City, MI, U.S.A., 1997. [175] Permanent magnet motor with low cogging torque by simulated magnetic field analysis, Shin-Etsu Chemical Co. Ltd., Magnetic Materials Research Center, Takefu-shi, Fukui Prefecture, Japan, 1999. [176] Pierce, A.D.: Acoustics: An Introduction to Its Physical Principles and Applications, McGraw-Hill, Sydney, Australia, 1981. [177] Pillay, P. and Krishnan, R.: An investigation into the torque behavior of a brushless d.c. motor, 23rd IEEE IAS Annual Meeting, New York, 1988, pp. 201–208. [178] Pillay, P. and Krishnan, R.: Modelling, simulation, and analysis of permanent magnet motor drives, Part 1 and 2, IEEE Trans. on IA 25(2), 1989, pp. 265–279. [179] Purkemani, M.: Beitrag zur Erfassung der Sättigungsoberfelder in Drehstromasynchromaschinen, PhD Thesis, Technischen Hochschule Hannover, Fakultät für Maschinenwesen, Hannover, Germany, 1971. [180] Randall, R.: Frequency Analysis, Brüel & Kjaer Technical Handbook, 1987. [181] Rockwood, W.B.: Statistical energy analysis applied to structural borne noise in marine structures, ASME Symposium, Boston, MA, U.S.A., 1987. [182] Sandman, B.E.: Fluid-loading influence coefficients for a finite cylindrical shell, J. Acoust. Soc. Am., Vol. 60, 1976, pp. 1256–1264. [183] Scharton, T.D. and Lyon, R.H.: Power flow and energy sharing in random vibration, J. Acoust. Soc. Am. Vol. 43(6), 1968, pp. 1332–1343. [184] Schenck, H.A.: Improved integral formulation for acoustic radiation surface, J. Acoust. Soc. Am., Vol. 44, 1968, pp. 41–58. [185] Schlichting, H.: Boundary Layer Theory, 6th edition, McGraw Hill, New York, 1966. [186] Serridge, M. and Licht, T.R.: Piezoelectric Accelerometers and Vibration Preamplifiers, Brüel and Kjaer Technical Handbook, 1986. [187] Shubov, I.G.: Noise and Vibration of Electrical Machines (in Russian), Energia, Leningrad, 1974, 2nd edition, Energoatomizdat, Leningrad, 1986. [188] Silbiger, A.J.: Radiation from circular pistons of elliptical profile, J. Acoust. Soc. Am., Vol. 33, 1961, pp. 1515–1592. [189] Smith, P.W.: Statistical models of coupled dynamic systems and the transition from weak to strong coupling, J. Acoust. Soc. Am., Vol. 65, 1979, pp. 695–698.


DK3193

noise

November 3, 2005

16:24

364

Bibliography

[190] Soedel, W.: Vibrations of Shells and Plates, 2nd edition, New York, Marcel Dekker, Inc., 1993. [191] Soize, C.: A model and numerical method in the medium frequency range for vibro-acoustic predictions using the theory of structural fuzzy, J. Acoust. Soc. Am., Vol. 94, 1993, pp. 849–865. [192] Stepanishen, P.R.: Radiated power and radiation loading of cylindrical surfaces with nonuniform velocity distributions, J. Acoust. Soc. Am., Vol. 63, 1978, pp. 328–338. [193] Stepanishen, P.R.: Modal coupling in the vibration of fluid-loaded cylindrical shells, J. Acoust. Soc. Am., Vol. 71, 1982, pp. 813–823. [194] Strelkov, S.P.: Mechanics, Mir Publishers, Moscow, Russia, 1978. [195] Sun, J.C., Wang, C., and Sun, Z.H.: Power flow between three series coupled oscillators, J. Sound and Vibration, Vol. 189(2), 1996, pp. 215–229. [196] Sun, Z.H., Sun, J.C., Wang, C. and Sheng, M.P.: Analysis of the causes of formation of negative loss factors, Chinese Journal of Acoustics, Vol. 15(3), 1996, pp. 239–246. [197] SYSNOISE Manual Book, NIT, Belgium, 1993. [198] Szechenyi, E.: Modal densities and radiation efficiencies of unstiffened cylinders using statistical methods, J. Sound and Vibration, Vol. 19(1), 1971, pp. 65–81. [199] Thomson, W.T.: Theory of Vibration with Applications, 2nd edition, Prentice Hall, 1981. [200] Timar, P.L., Fazekas, A., Kiss, J., Miklos, A. and Yang, S.J.: Noise and Vibration of Electrical Machines, Elsevier, Amsterdam – Oxford – New York – Tokyo, Japan, 1989. [201] Timar, P.L. and Lai, J.C.S.: Acoustic noise of electromagnetic origin in an ideal frequency-converter-driven induction motor, IEEE Proc. Electr. Power Appl., 141(6), 1994, pp. 341–346. [202] Timar, P.L. and Lai, J.C.S.: Analysis of the effects of time harmonics on the radiated sound power of variable speed induction motors, Internoise’96, Liverpool, U.K., 1996, pp. 445–448. [203] Timoshenko, S.: Theory of Plates and Shells, McGraw Hill, 1940. [204] Titan Large A.C. Electric Motors: Installation, Operation, and Maintenance Manual, U.S. Motors Division, Emerson Electric Co., St Louis, MO, U.S.A., 1999. [205] Toader, S.: Combined analytical and finite element approach to the field harmonics and magnetic forces in synchronous machines, Int. Conf. on Electr. Machines ICEM’94, Paris, France, Vol. 3, 1994, pp. 583–588. [206] Tsivitse, P.J. and Weishsmann, P.R.: Polyphase induction motor noise, IEEE Trans. on IGA, Vol. 7(3), 1971, pp. 339–358. [207] Tso, Y.K. and Hansen, C.H.: The prediction of structure-borne noise transmission in ships using statistical energy analysis, Acoustics Australia, Vol. 25(3), 1997, pp. 97–102. [208] Twomey, W.J., Chen, T.L.C., Ojalvo, I.U., and Ting, T.: General method for modifying a finite element model to correlate with modal test data, J. of the American Helicopter Soc., Vol. 36(3), 1991, pp. 48–58.


DK3193

noise

November 3, 2005

16:24

Bibliography

365

[209] Vandevelde, L. and Melkebeek, J.A.A.: Calculation of radial magnetic forces for the analysis of noise and vibrations of squirrel-cage induction motors, 8th IEE Int. Conf. on Electr. Machines and Drives EMD’97, 1997, London, U.K., pp. 86–90. [210] Vandevelde, L. and Melkebeek, J.A.A.: Magnetic forces and magnetostriction in electrical machines and transformer cores, IEEE Trans. on MAG, Vol. 39(3), 2003, pp. 1618–1621. [211] Vandevelde, L., Gyselinck, J., De Wulf, M.A.C., and Melkebeek, J.A.A.: Finite-element computation of the deformation of ferromagnetic material taking into account magnetic forces and magnetostriction, IEEE Trans. on MAG, Vol. 40(2), 2004, pp. 565–568. [212] Ver, I.L. and Holmer, C.I.: In Noise and Vibration Control (Beranek, L.L., ed.), McGraw-Hill, New York, 1971, pp. 287–296. [213] Verdyck, D. and Belmans, R.J.M.: An acoustic model for a permanent magnet machine: Modal shapes and magnetic forces, IEEE Trans. on IA, Vol. 30(6), 1994, pp. 1625–1631. [214] Verma, S.P. and Girgis, R.S.: Method of accurate determination of resonant frequencies and vibration behavior of stators of electrical machines, IEE Proc. Pt. A, Vol. 128(1), 1981, pp. 1–11. [215] Verma, S.P. and Girgis, R.S.: Experimantal verification of resonant frequencies and vibration behavior of stators of electrical machines, IEE Proc. Pt. B, Vol. 128(1), 1981, pp. 12–32. [216] Verma, S.P. and Yang, Y.: Electromagnetic acoustic noise and vibrations of cylindrical rotor synchronous machines, Int. Conf. on Electr. Machines ICEM’94, Paris, France, Vol. 3, 1994, pp. 431–436. [217] Vibration Isolation of Fans, Engineering Data ED-1900, Twin City Fan Companies, Ltd., Minneapolis, MN, U.S.A., 2001. [218] Vlahopoulos, N. and Zhao, X.: Basic development of hybrid finite element method for midfrequency structural vibrations, AIAA Journal, Vol. 37 (11), 1999, pp. 1495–1505. [219] Voldek, A.I.: Electrical Machines (in Russian), 2nd edition, Energia, Leningrad, Russia, 1974. [220] Walker, J.H. and Kerruish, N.: Open-circuit noise in synchronous machines, Proc. IEE, Part A, Vol. 107(36), 1960, pp. 505–512. [221] Wallace, C.E.: Radiation resistance of a rectangular panel, J. Acoust. Soc. Am., Vol. 51(3), 1972, pp. 946–952. [222] Wang, M.Q., Sheng, M.P., and Sun, J.C.: The direct and indirect power flow of three non-conservatively series couples oscillators, J. Sound and Vibration, Vol. 212(2), 1998, pp. 231–251. [223] Wang, C., Lai, J.C.S., and Pulle, D.W.J.: A statistical method for the prediction of acoustic noise radiation from induction motors, European Power Electronics Conf. EPE’99, 1999, Lausanne, Switzerland, Vol. 8, pp. 1–8. [224] Wang, C. and Lai, J.C.S.: Vibration analysis of induction motors, J. Sound and Vibration, Vol. 224(4), 1999, pp. 733–756.


DK3193

noise

November 3, 2005

16:24

366

Bibliography

[225] Wang, C. and Lai, J.C.S.: Modeling the vibration behavior of an induction motor, J. Acoust. Soc. Am., Vol. 103(5) 1998, p. 3084. [226] Wang, C. and Lai, J.C.S.: Prediction of natural frequencies of finite length circular cylindrical shells, Appl. Acoustics, Vol. 59(4), 2000, pp. 385–400. [227] Wang, C. and Lai, J.C.S.: The sound radiation efficiency of finite length acoustically thick circular cylindrical shells under mechanical excitation I: Theoretical analysis, J. Sound and Vibration, Vol. 232(2), 2000, pp. 431– 447. [228] Wang, C., and Lai, J.C.S.: The sound radiation efficiency of finite length circular cylindrical shells under mechanical excitation II: Limitations of the infinite length model, J. Sound and Vibration, Vol. 241(5), 2001, pp. 825– 838. [229] Wang, C., Astfalck, A., and Lai, J.C.S.: Sound power radiated from an inverter driven induction motor: Experimental investigation, IEE Proc. Pt. B. Electr. Power Application, Vol. 149(1), 2002, pp. 46–52. [230] Wang, C., Lai, J.C.S., and Pulle, D.W.J.: Prediction of acoustic noise from variable-speed induction motors: Deterministic versus statistical approaches, IEEE Trans. on IA, Vol. 38(4), 2002, pp. 1037–1044. [231] Wang, C. and Lai, J.C.S.: Discussions on the number of modes required for statistical energy analysis-based calculations, J. Sound and Vibration, 2005, Vol. 281, 2005, pp. 475–481. [232] Wang, C., Lai, J.C.S., and Astfalck, A.: Sound power radiated from an inverter driven induction motor II: Numerical analysis, Proc. IEE: Electr. Power Application, Vol. 151(3), 2004, pp. 341–348. [233] Wang, C. and Lai, J.C.S.: Sound power radiated from an inverter driven induction motor: Statistical energy analysis, IEE Proc., Pt. B Elect. Power Appl., 2005, Vol. 152(3), 2005, pp. 619–626. [234] Waser, M.P. and Crocker, M.J.: Introduction to the two microphone crossspectral method of determining sound intensity, Noise Control Eng., Vol. 22, 1984, pp. 76–84. [235] Watanabe, S., Kenjo, S., Ide, K., Sato, F., and Yamamoto, M.: Natural frequencies and vibration behaviour of motor stators, IEEE Trans. on PAS, Vol. 102(1), 1983, pp. 949–956. [236] Weber, W.: Der Nutungsfaktor in elektrischen Maschinen, ETZ, 1928, pp. 858–861. [237] Wester, E.C.N. and Mace, B.R.: Statistical energy analysis of two edgecoupled rectangular plates: Ensemble averages, J. Sound and Vibration, Vol. 193(4), 1996, pp. 793–822. [238] Williams, W. Parke, F.G. Moran, D.A., and Sherman, C.H.: Acoustic radiation from a finite cylinder, J. Acoust. Soc. Am., Vol. 36, 1964, pp. 2316– 2322. [239] Witczak, P.: Determination of mechanical vibration caused by magnetic forces in induction motors (in Polish), DSc. Thesis, Technical University of Lodz, Zesz. Nauk. No. 725, Lodz, Poland, 1995.


DK3193

noise

November 3, 2005

16:24

Bibliography

367

[240] Witczak, P.: Calculations of force densities distribution in electric machinery by means of the magnetic stress tensor, Archiwum Elektrotechniki, Warsaw, Poland, 1996, No. 2, pp. 145–160. [241] Witczak, P. and Brudny, J.F.: Calculations of magnetic vibration forces in induction motors, Zeszyty Naukowe PL Elektryka No. 789, Lodz, Poland, 1998, pp. 37–44. [242] Witczak, P.: Magnetic vibration forces in induction motors, Int. Workshop on Electr. and Magn. Fields, 1996, Liege, Belgium, pp. 255–260. [243] Woodhouse, J.: An introduction to statistical energy analysis of structural vibration, Appl. Acoustics, Vol. 14, 1981, pp. 455–469. [244] X2C A.C. Inverters, TB Wood’s, Chambersburg, PA, U.S.A., 2003, www.tbwoods.com. [245] Yacamini, R. and Chang, S.C.: Noise and vibration from induction machines fed from harmonic sources, IEEE Trans. on EC 10(2), 1995, pp. 286–292. [246] Yahiaoui, A. and Bouillault, F.: 2D and 3D numerical computation of electrical parameters of an induction motor, IEEE Trans. on MAG, Vol. 30(5), 1994, pp. 3690–3692. [247] Yang, S.J.: Finite element method in evaluating the stator natural frequencies of small machines, Int. Conf. on Electrical Machines ICEM ’78, 1978, Brussels, Belgium. [248] Yang, S.J.: Low-Noise Electrical Motors, Clarendon Press, Oxford, 1981. [249] Yap, F.F. and Woodhouse, J.: Investigation of damping effects on statistical energy analysis of coupled structures, J. Sound and Vibration, Vol. 197(3), 1996, pp. 351–371. [250] Zhang, Q. J. and Sainsbury, M. G.: The energy flow method for strongly coupled systems, 17th Int. Modal Analysis Conf., FL, U.S.A., 1999, pp. 1839– 1845. [251] Zhang, Q.J., Wang, C., Wang, D., Qian, M., Nack, W., and Pamidi, P.R.: Energy flow method for mid-frequency vibration analysis, SAE Noise and Vibration Conf., Traverse City, MI, U.S.A., 2003, Paper No. 2003-01-1454. [252] Zhu, Z.Q. and Howe, D.: Finite element analysis of acoustic power radiated by electrical machines, Proc. Inst. Acoust., Vol. 12(6), 1990, p. 29–36. [253] Zhu, Z.Q. and Howe, D.: Analytical prediction of the cogging torque in radial field permanent magnet brushless motors, IEEE Trans. on MAG, Vol. 28(2), 1992, pp. 1371–1374. [254] Zhu, Z.Q. and Howe, D.: Electromagnetic noise radiated by brushless permanent magnet d.c. drives, Electr. Machines and Drives Conf., Oxford, U.K., 1993, pp. 606–611. [255] Zhu, Z.Q. and Howe, D.: Improved methods for prediction of electromagnetic noise radiated by electrical machines, IEE Proc. of the Electric Power Applications, Vol. 141, 1994, pp. 109–120. [256] Zhu, Z.Q. and Howe, D.: Influence of design parameters on cogging torque in permanent magnet machines, IEEE Trans. on EC 15(4), 2000, pp. 407–412.


Noise of Polyphase Electric Motors

Noise of polyphase electric motors

Handbook of Electric Motors

Handbook of Small Electric Motors

Handbook of Small Electric Motors

ENERGY-EFFICIENT ELECTRIC MOTORS

Electric Motors Drives

Energy-efficient electric motors

Electric Motors in the Home Workshop (Workshop Practice)

Electric Motors and Drives: Fundamentals, Types and Applications

Control of Induction Motors

Control of stepping motors

Noise

Noise

Noise

Noise

Molecular Motors

Molecular Machines & Motors

AC Motors(en)(92s)

Fundamentals of Electric Circuits

Fundamentals of electric drives

Fundamentals of Electric Waves

Physics of Electric Propulsion

White Noise

White Noise

White Noise

Noise Level

White Noise

Quantum noise

Pink Noise

Шум. (Noise)

Noise of Polyphase Electric Motors