Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5

Conference Proceedings of the Society for Experimental Mechanics Series

For other titles published in this series, go to www.springer.com/series/8922

Tom Proulx Editor

Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5 Proceedings of the 2010 Annual Conference on Experimental and Applied Mechanics

Editor Tom Proulx Society for Experimental Mechanics, Inc. 7 School Street Bethel, CT 06801-1405 USA [email protected]

ISSN 2191-5644 e-ISSN 2191-5652 ISBN 978-1-4419-9493-6 e-ISBN 978-1-4419-9798-2 DOI 10.1007/978-1-4419-9798-2 Springer New York Dordrecht Heidelberg London Library of Congress Control Number: 2011927928 © The Society for Experimental Mechanics, Inc. 2011 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Preface

Experimental Mechanics on Emerging Energy Systems and Materials represents one of six tracks of technical papers presented at the Society for Experimental Mechanics Annual Conference & Exposition on Experimental and Applied Mechanics, held at Indianapolis, Indiana, June 7-10, 2010. The full proceedings also include volumes on Dynamic Behavior of Materials, Application of Imaging Techniques, Experimental and Applied Mechanics, along with the 11th International Symposium on MEMS and Nanotechnology, and the Symposium on Time Dependent Constitutive Behavior and Failure/Fracture Processes. Each collection presents early findings from experimental and computational investigations on an important area within Experimental Mechanics. The current volume on the Role of Experimental Mechanics on Emerging Energy Systems and Materials, includes studies on composite Materials for power generation such as wind power generators, fuel cells technology, materials and durability; solar energy – solar cell materials and technology, alternative forms of energy, and new energy phenomena from nature. In recent years, energy has become a hot topic in all walks of life, the SEM community is no exception. Steered by the Research Committee, this track brings together researchers and engineers interested in mechanics aspects of energy systems and materials, and provides a forum to facilitate technical interaction and exchange. We thank the SEM staff and all session organizers for their persistent, devoted efforts as well as the authors, session chairs and presenters in this track who make the ultimate success of the track and the conference. The Society would like to thank the organizers of the track, Bill Y.J. Chao, University of South Carolina; Yu-Ling Lo, National Cheng-Kung University; Ashok K. Ghosh, New Mexico Technology University; Ron Y. Li, General Motors Corporation; David Dillard, Virginia Polytechnic Institute and State University for their efforts. Bethel, Connecticut

Dr. Thomas Proulx Society for Experimental Mechanics, Inc

Contents

1

Experimental Mechanics for Prognosis of Material State Changes in Heterogeneous Materials for Energy Systems K.L. Reifsnider, P.K. Majumdar, P.D. Fazzino, F. Rabbi

1

2

Characterisation of Carbon Nanotube Foam for Improved Gas Storage Capability A. Peña, A. Guerrero, J. Puerta, J.L. Brito, T.K. Heckel

11

3

Thermoresponsive Microcapsules for Autonomic Lithium-ion Battery Shutdown M. Baginska, B.J. Blaiszik, S.A. Odom, A.E. Esser-Kahn, M.M. Caruso, J.S. Moore, N.R. Sottos, S.R. White

17

4

Service Life Prediction of Seal in PEM Fuel Cells T. Cui, C.-W. Lin, C.-H. Chien, Y.-J. Chao, J.V. Zee

25

5

Experimental Validation of a Constitutive Model for Ionomer Membrane in Polymer Electrolyte Membrane Fuel Cell (PEMFC) W. Yoon, X. Huang, R. Solasi

33

Evidence of Piezonuclear Reactions: From Geological and Tectonic Transformations to Neutron Detection and Measurements A. Carpinteri, G. Lacidogna, A. Manuello, O. Borla

41

Energy Dispersive X-ray Spectroscopy Analysis on Rock Samples Subjected to Piezonuclear Tests A. Carpinteri, A. Chiodoni, A. Manuello, R. Sandrone

49

6

7

8

Acoustic and Electromagnetic Emissions in Rocks Under Compression G. Lacidogna, A. Manuello, A. Carpinteri, G. Niccolini, A. Agosto, G. Durin

57

9

Gear With Asymmetric Teeth for use in Wind Turbines S. Ekwaro-Osire, T.-H. Jang, A. Stroud, I. Durukan, F.M. Alemayehu, A. Swift, J. Chapman

65

10

Material Brittleness and the Energetics of Acoustic Emission A.A. Pollock

73

11

b-value of Plain Concrete Beams Based on AE Quanta S. Muralidhara, H. Eskandari, B.K. Raghu Prasad, R.K. Singh

81

12

AE Monitoring of the Syracuse Athena Temple: Scale Invariance in the Timing of Ruptures G. Niccolini, G. Durin, G. Lacidogna, A. Manuello, A. Carpinteri

95

viii

13

Analysis of Energy Released by Elastic Emission in Brittle Materials Under Compression A. Schiavi, G. Niccolini, P. Tarizzo, G. Lacidogna, A. Manuello, A. Carpinteri

103

14

Numerical Simulation of AE Activity in Quasi-brittle Materials Under Compression S. Invernizzi, A. Carpinteri, G. Lacidogna, A. Manuello

109

15

Mechanical Characterization and AE of Translucent Self-compacting Concrete Plates in Bending A. Manuello, A. Carpinteri, S. Invernizzi, G. Lacidogna, S. Pagliolico, A. Torta

117

16

Reconfiguration of Large Leaf Under Wind Load N.-S. Liou, S.-S. Tsai, H.-H. Yen

125

17

Manufacturing Challenges for the Modern Wind Turbine Rotor G. Kanaby

129

18

Structural Monitoring of Wind Turbine Blades Using Fiber Optic Bragg Grating Strain Sensors A. Turner, T.W. Graver, A. Mendez

19

Degradation of Polymer Electrolyte Membranes A. Jones, J. Malladi

20

The Nonlinear Viscoelastic Properties of PFSA Membranes in Water-immersed and Humid Air Conditions L. Yan, T.A. Gray, K.A. Patankar, S.W. Case, M.W. Ellis, R.B. Moore, D.A. Dillard, Y.-H. Lai, Y. Li, C.S. Gittleman

149 155

163

21

Assessing Durability of Elastomeric Seals for Fuel Cell Applications J.E. Klein, G.M. Divoux, H.K. Singh, S.W. Case, D.A. Dillard, J.G. Dillard, W. Kim, R.B. Moore, J.B. Parsons

175

22

Compression of Seals in PEM Fuel Cells C.-H. Chien, C.-W. Lin, Y.-J. Chao, C. Tong, J.V. Zee

183

23

Mechanical Characterization and Modeling of Electrolyte Membranes in Electrolyte-supported SOFCs R. Berke, A. Suresh, M.E. Walter

193

24

Metal Foils in Clean, Renewable Energy Applications M. Robinson

203

25

Experimental and Probabilistic Analysis of Asymmetric Gear Tooth S. Ekwaro-Osire, I. Durukan, F.M. Alemayehu

207

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Experimental Mechanics for Prognosis of Material State Changes in Heterogeneous Materials for Energy Systems

Kenneth L. Reifsnider*, PhD, NAE, Professor (*corresponding author contact: [email protected]) Prasun K. Majumdar, PhD, Research Assistant Professor Paul D. Fazzino, and Fazle Rabbi, Graduate Student Department of Mechanical Engineering, University of South Carolina, 300 Main Street Columbia, SC 29208

ABSTRACT The concept of prognosis is typically discussed in terms of mechanical characteristics such as structural integrity, durability, damage tolerance, fracture toughness, etc. These familiar concepts are usually addressed by considering balance equations, crack growth relationships, and constitutive equations with constant material properties, and constant or cyclically applied load conditions. Loading histories are represented by changing stress (or strain) states, only. But for many situations, especially associated with high performance engineering structures, the local state of the material may also change during service, so that the properties used in those equations are functions of time and history of applied conditions. But for many energy systems, a broader definition is required. For example, in fuel cells, properties such as conductivity and electrochemical character are altered by material degradation, so that “property fields” replace the global constants in multiphysics balance equations, and time and history enter into the governing equations. The present paper will examine a small set of such problems which involve novel experimental methods of following the accumulation of distributed damage and changes in material state. Specifically, the paper discusses this problem in the context of material state changes measured by impedance variations as a method of following and interpreting those changes in terms of functional performance. The application of these concepts is extended from mechanical structures to energy systems, wherein the material state changes result in variations in electrochemical properties that directly control functional performance of devices such as fuel cells.

Introduction Prognosis of Mechanical Performance in Insulating Structural Materials: The traditional concept of structural integrity has a variety of contexts and interpretations. One such scenario is the prediction of the stability or likelihood of collapse from information about response to ‘safe’ loading conditions [1, 2]. Another context has to do with the large field of health monitoring, in which structural characteristics such as stiffness,[3] or more localized information from either passive

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_1, © The Society for Experimental Mechanics, Inc. 2011

1

2

(e.g. acoustic emission) or active (e.g. surface wave excitation response)[4,5] assessment is used to evaluate ‘degradation’ as a method of estimating future performance. And a rather robust literature presents a context of fracture mechanics, focused on localized deformation, crack initiation and propagation, in some cases including specific address of the effects of local constituent particle effects on that sequence on prognosis of performance and life [6]. The focus of the present paper is very different. We are concerned with heterogeneous materials which have meso-phases that are very brittle, e.g., woven fiber reinforced polymer composites. Moreover, we are concerned with the details of local distributed damage development and the manner in which that damage eventually creates a global fracture path. Elements of the mechanical problem: In previous work, we have examined the progression of damage in composite materials and the development of fracture paths during the tensile loading of straight sided coupons of woven glass reinforced composite strips, loaded at different angles to the fiber directions [14-17]. An example of the stress-strain behavior observed under those conditions is shown in Fig. 1 below, for different angles of loading. 9000

0 degree

90o

8000

L o ad (L B )

30 degree 45 degree

7000

Load, lb

15 degree

0o

60 degree 90 degree

6000

5000

15o

4000

30o

60o

3000

2000

45o 1000

0 0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

14.0%

16.0%

18.0%

20.0%

Strain Figure 1 Stress-strain response of woven glass/epoxy specimens of under tensile loading As seen in the figure, the strains to break in the off-axis specimens is large, approaching 20 percent in the case unidirectional loading at orientations of 45 degrees to the fiber directions. For these situations, the resulting global fracture paths eventually followed the fiber directions in a general sense, but the development of the damage that produced those fractures was complex. Figure 2 below shows examples of those fracture patterns. In our earlier work, we were able to show that these large, finite deformations and strains can be followed with proper constitutive formulations, and that when those formulations are correctly used, that the resulting fracture directions can be correctly predicted [15,17] . During the course of that work, we established the fundamental elements of the process of damage development, coalescence, and localization from the perspective of continuum formulations. What is missing, however, is a definitive definition of this process at the material level.

3

Figure 2 Fracture patterns for off-axis specimens loaded at (from the top) 60, 45, 30, 30, 15, 90, and 0 degrees to the fiber directions. The local details of damage development and accumulation control the progress of global property degradation, and they control the development of the final fracture event, i.e., they control the life of the engineering component under sustained or cyclic loading. For that reason, we focus on the local material state changes.

Impedance as an Assessment of Material State Changes Modern methods of “health monitoring” concentrate on the detection and location of flaws and defects, not on interactions or relationships between flaws. Isolated cracks and defects that are comparable in dimension to microstructures are difficult (often impossible) to discern. When they combine to form a macro-crack, most of the life of a component is over. The intermediate stage of development when nano- or micro-cracks begin to form incipient fracture paths is of primary interest, and perhaps least well understood. Material state changes, reflected in changes in properties such as conductivity, are well suited to following such incipient fracture path development. For this discussion, we consider the general property of conductance for this purpose. The concept of using the conductivity of direct current has been used by some investigators to estimate internal integrity, usually in the context of breaking conduction paths in existing conductors, e.g., breaking of carbon fibers in a continuous fiber composite, and “effective property” theories of thermal and electrical conductivity for constant (in time) boundary conditions have been developed [17,18]. However, our approach considers the complex impedance of materials under the excitation of an alternating current. This approach has the advantage of being applicable to materials (or regions of materials) that are not directly conductive, but do have some level of permittivity, i.e., dielectric character (such as the voided regions of defects). Some of the principles of this approach are illustrated in Fig. 3. In that figure, defect development, growth, coalescence, and fracture path development are depicted to represent progressive damage development during repeated or long-term continuous mechanical loading during the life of an engineering composite material. Also indicated is the alternating current excitation of a test element of the material, and the idealized response for the undamaged and fracture-path case, (a) and (e) respectively. For the undamaged case, (a), it is assumed that the material under test is an insulator, such as glass-reinforced epoxy (as will be discussed in the data below). For that situation, the AC response can be expected to be purely capacitive, and (for linear materials, over a wide range of the response) the absolute value of the impedance (the square root of the sum of the squares of the real and imaginary components) will be an essentially linear curve with downward slope, i.e., inversely proportional to the frequency as in Fig. 3 (a).

4

Figure 3 Schematic impedance concept for undamaged materials (a), defect initiation (b), defect coalescence (c), and fracture path initiation (d) with idealized frequency response (e). The other limiting condition is shown in Fig. 3 (e). When a fracture path has developed in the element, by definition, there will be a continuous voided path from one physical face to the other. Let us postulate that air with some hydration or simply diffused moisture (or other gas or fluid with some ionic content) enters that voided space. If that material is a good conductor, and the path is continuous, then the idealized response will not depend on the frequency of excitation (at least to first order), as suggested in Fig. 3 (e). Moreover, if one plots the imaginary component of the impedance vs. the real component, as shown in the last entry of Fig. 3, a general form that is well studied in the literature is obtained, with the high frequency and low frequency intercept and the shape of the response (which may show several arcs) carrying specific information about the specific nature of the conductive regions and their connectivity.

Figure 4 Example of damage development under cyclic end-loaded bending (tensile side) in woven glass reinforced epoxy coupons for loading in the fiber-directions (a) and for off-axis loading (b). To test these concepts, the authors (and others in their group) have conducted a series of exploratory experiments. Woven-glass reinforced polymer composite specimens were used to develop a data set of fatigue results, including examinations of material state changes (especially matrix cracking) as a function of loading angle (relative to the material principal axis), and residual strength under tensile loading (c.f., Fazzino, [19-23]). Norplex-Micarta Inc. provided NP130 E-glass,

5

plain weave thin laminated composite plates with a halogenated epoxy resin matrix for the experimental work. The fiber volume fraction was 55%. Total specimen thickness (for the five ply laminates used) was about 1 mm. End-loaded bending was used to introduce progressive damage; an example of observed micro-cracking is shown in Fig. 4. Impedance spectroscopy measurements were made on such specimens at 0, 25, 50, 75, and 100 percent of life, using a standard (Gamry) impedance measurement device (see references [19-23] for more details).

1.E+07 1.E+06

undamaged 25% of life 50% of life 75% of life

1.E+05 1.E+04

End of life

1.E+08

Z_magnitude (ohm)

Z_magnitude (ohm)

1.E+08

1.E+03

undamaged

1.E+07 1.E+06

25% of life

1.E+05

75% of life

50% of life

1.E+04 1.E+03

End of  life

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06

Frequency (Hz)

Frequency (Hz)

(a) on‐axis

(b) Off axis

Figure 5 Example of damage impedance response as a function of frequency for a) on-axis and b) off-axis loading at 0 (top), 25, 50, 75, and 100 percent (bottom) of life.

-Z_imag (ohm)

Figure 5(a) shows example results for on-axis loading. The initial response is remarkably linear over the total range of frequencies tested, and the end-of life results are largely independent of frequency over nearly all of the frequency range. Figure 5(b) shows similar results for off-axis loading. The damage development in specimens loaded in these directions is much earlier in life and more severe, and the DC conduction path develops much more quickly. Finally, an example plot of the real and imaginary parts of the impedance for these tests is shown in Fig. 6. 2.0E+5 1.8E+5 1.6E+5 1.4E+5 1.2E+5 1.0E+5 8.0E+4 6.0E+4 4.0E+4 2.0E+4 0.0E+0 0.0E+0

1.0E+5

2.0E+5

3.0E+5

4.0E+5

Z_real (ohm)

Figure 6 Example of real and imaginary parts of impedance of response for on-axis loading at 0 (top), 25, 50, 75, and 100 percent (bottom) of life. The sensitivity to the early stages of damage (top and second from top curve) is remarkable and distinct. There is a wealth of information about the specific nature of damage development carried in the results shown in Fig. 5 and Fig. 6, much of which has not yet been explored. However, the

6

basic impedance response of the materials as a function of damage development seems to fit the concept advanced in the discussion around Fig. 3.

Material State Changes in Energy Systems As it happens, heterogeneous (composite) materials are a critical technology for many energy conversion and storage systems, and for countless membrane and thin film applications in separation and reforming operations. Consumer products like high performance batteries would not be possible without the remarkable advances in heterogeneous functional materials that have resulted from new developments in the design and synthesis of nano-structured heterogeneity. For the present discussion, we will focus on just one such application, the electrodes and electrolytes in solid oxide fuel cells (SOFCs). SOFCs are made by bonding together two thin porous ceramic layers (the anode and cathode) on either side of a dense electrolyte layer. The porosity in the electrolytes allows fuel (e.g., hydrogen) to be oxidized and oxygen to be reduced to drive the conversion of the chemical potential of the reactants to electrical power. The principle of operation is sketched in Fig. 7.

Figure 7 Principle of operation of a solid oxide fuel cell (SOFC) The dense electrolyte layer prevents the reactant gasses from combining by direct combustion. This material system is heterogeneous at several levels. At the meso-level, the layers which may have thicknesses of the order of tens of microns, form a composite laminate at that scale. And second, the electrodes (anodes and cathodes) are often heterogeneous at the nano-scale in order to achieve a proper balance of electrical and ionic conduction, and chemical stability. SOFCs are high temperature, solid state devices. The high temperature makes it possible for solid electrolyte (ceramic) layers to conduct ions (with remarkable efficiency). It also enables the device to operate with reactants that are not restricted to pure elements (such as hydrogen and oxygen), so that SOFCs operating on propane, methane, syngas, ethanol, propanol, and even hydrocarbon fuels such as diesel are common. However, the high operating temperature also drives material state changes, which tend to change the operating characteristics of the fuel cells with time and power produced. Impedance spectroscopy is widely used to assess the performance of fuel cells, particularly to assess the polarization losses associated with processes such as activation, Ohmic resistance, and diffusion. The typical interpretation of these data is in terms of equivalent circuits. For example, results are typically interpreted in terms of the ‘effective’ values of resistance, Re and Rt, and the capacitive value Cd in the equivalent circuit shown below. Those values are obtained from fitting the [Z] vs. frequency response data with the equations shown.

7

Figure 8 Equivalent circuit typically used to analyze the impedance data obtained from a solid oxide fuel cell (SOFC) The circuit shown in Fig. 8, and many more discussed in the literature, are useful in the sense that by varying the operating conditions of the fuel cell, the circuit values of various general regions (or more specifically, various processes in those regions) can often be identified and characterized. Since, for example, Re is the real part of [Z] when a very high frequency is applied, if there is a current flowing in the fuel cell during the test, the value of the Ohmic contribution to the polarization loss is generally attributed to Re. If the electrolyte is known to dominate that loss, then Re is assigned to the collective Ohmic resistance of the electrolyte layer, etc. This type of analysis can provide some very valuable general information, but does not define material state changes in any specific way; and in addition, the interpretations are heuristic, in the sense that they are logic drawn from an analogue electrical circuit that is not derived or justified in any way, except that it happens to fit the data to some unspecified level of accuracy.

(a)

(b)

Figure 9 Predicted and observed impedance spectrum before thermal aging (a) and after sintering (b) showing the exact representation of local morphology changes by calculated and observed response to AC excitation.[24] But impedance can be related to specific material properties and morphologies. In earlier publications we have shown that, for example, in a single-phase polycrystalline (electrolyte) material, impedance spectroscopy can clearly identify the contributions of the individual grains and the grain o boundaries. When that material was subjected to thermal aging at 1000 C which had the effect of changing the grain morphology due to sintering, those contributions changed accordingly. To our great surprise, when we used the respective micrographs to solve Maxwell’s equations (using the commercial code COMSOL) set on those two observed microstructures, the analysis correctly predicted the impedance response for those two cases as shown in Fig. 9.[24] More recent results are even more interesting. We have been able to show that the capacitive response of the nano-structured heterogeneous functional materials which are used for the porous electrodes in SOFCs is remarkably sensitive to the geometric details of the local morphology. For example, we have considered an electrode configuration that is an idealized representation of the internal morphology of the Bi-electrode-Supported-Cell (BSC) concept developed by NASA Glenn

8

under the leadership of Tom Cable.[25] That design features a YSZ scaffold of freeze cast porous material on either side of a dense YSZ layer, co-sintered as one unit of material. The electrodes are formed by infiltrating that structure with suitable cathode (on one side) and anode (on the other side) material to form a fuel cell. An idealization of just the YSZ scaffold is sketched in Fig. 10. Catalytic source of electrons

Figure 10 Calculated voltage (charge) distribution in a simulated SOFC fuel cell morphology The charge distribution is created by the “catalytic source of electrons” located as shown in the figure. One of the authors (F.R.) has shown (in a publication under preparation) that the charge distribution and the features of the associated impedance are highly dependent on the local details of the geometry. In an operating SOFC, the internal geometry often ‘ages’ with time, e.g., surfaces between heterogeneous phases change due to diffusion, composition gradients drive grain boundary motion, and transport of chemical species (such as sulphur or carbon) alter not only surfaces but change porosity shape and size. These material state changes, like those we discussed in the case of mechanical degradation, can not only be ‘sensed,’ but also interpreted if we were able to set the proper multi-physics equations on the local geometry so that we could develop an understanding of the processes that cause the state changes. That is our long term objective.

Conclusions and Continuing Work We have outlined a concept and philosophy for the use of impedance spectroscopy to follow the mechanical and electrochemical state changes aeronautical and energy conversion systems. For the case of mechanical systems, the development and coalescence of distributed damage to create a fracture path in fiber-reinforced composite materials was found to have a specific relationship to the details of the changes observed using impedance spectroscopy to characterize the through the thickness impedance losses that develop when micro-damage creates regions of finite capacitive activity in an otherwise insulating material. For the fuel cells considered, which are fundamentally dielectric, impedance spectroscopy provides a tool to follow not only the collective polarization losses of the electrochemical processes associated with the conversion of chemical energy to electrical energy, but we have also demonstrated that the impedance can be used to follow the specific details of local morphology change. This latter capability is invaluable for the determination of the mechanisms that drive performance change (usually degradation) in a fuel cell (which must be a sealed system). But even more important, understanding those changes can form the foundation for a rational design of the details of the internal morphology, a process that is currently almost entirely heuristic. Much work is yet to be done on the specific association of the details of the results with local details. If we could properly set (and solve) the balance equations of mass, momentum, energy and charge, and the associated boundary conditions in terms of the correct material constants on the local morphology of the heterogeneous materials we are discussing, we could, in principle, design the nano-structure of the functional materials for specific performance from first principles calculations. We do this routinely for mechanical structures, but have no comparable foundation for design of

9

electrochemical devices. And if those basic understandings were there, we would also have the ideal foundation for prognosis of future performance, based on the analysis of the physical processes that bring about material state changes during operation. Impedance spectroscopy appears to offer a path to achieving at least some of those objectives.

Acknowledgements The authors gratefully acknowledge the support of the Air Force Office of Scientific Research under grant number FA9550-07-1-0460 and the Office of Naval Research under grant no. N00014-05-1-0103 for support of the research that relates to the mechanical behavior discussed above. The authors further acknowledge the support of the Department of Energy for the Energy Frontiers Center for “Physics Based Design and Synthesis of Nano-Structured Heterogeneous Functional Materials (called HeteroFoaMs),” under grant number DE-SC0001061 which supported the research discussed above that concerns energy systems.

References 1. Stull, C.J., Earls, C.J. and Wilkins, A., “P Posteriori Initial Imperfection Identification in Shell Buckling Problems,” Comput. Methods Appl. Mech. Engrg. 198 (2008) 260-268 2. Hartmann, D., Breidt, M., Nguyen, V., Stangenberg, F., Hohler, S., Schweizerhof, K., Mattern, S., Blankenhorn, G., Moller, B., and Liebscher, M., “Structural Collapse Simulation uinder Consideration of Uncertainty – Fundametnal Concept and Results,” Computers and Structures 86 (2008) 2064-2078 3. Williams, P., “Structural Damage Detection from Transient Responses Using Square-root Unscented Filtering,” ACTA Astronautica 63 (2008) 1259-1272 4. Su, Z.Q., Ye, L., and Lu, Y., “Guided Lamb Waves for Identification of Damage in Composite Structures: A Review,” J. Sound and Vibration 295 #3-5 (2006) 753-780 5. Staszewski, W.J. et al., “Structural Health Monitoring using Scanning Laser Vibrometry: I. Lamb Wave Sensing,” Smart Materials and Structures 13 #2 (2004) 251-260 6. Papazian, J.M., Anagnostou, ElL., Engel, S.J., Hoitsma, D., Madsen, J., Silberstein, R.P., Welsh, G., and whiteside, J.B., “A Structural Integrity Prognosis System,” Engineering Fracture Mechanics (2008) in press 7. Bieler, T.R., Eisenlohr, P., Roters, F., Kumar, D., Mason, D.E., Crimp, M.A., Raabe, D., “The Role of Heterogeneous Deformation on Damage Nucleation at Grain Boundaries in Single Phase Metals,” Intl. J. Plasticity, (2008) in press 8. Ogihara, S. and Reifsnider, K.L., Applied Composite Materials, Vol.9, 249-263, 2002 9. Reifsnider, K.L., Tamuzs, V. and Ogihara, S., Composites Science and Technology, Vol.. 66, 2473-2478, 2006 10. Sun C.T. and Chen, J.L., Composite Materials, 23, 1009-11020, 1989 11. Tamuzs, V.m, Dzelzitis, K. and Reifsnider, K.L., Applied Composite Materials, Vol.11 No.5, 281-293, 2004 12. Tamuzs, V., Dzelzitis, K., and Reifsnider, K.L., Applied Composite Materials, Vol. II, 281-293, 2004 13. Reifsnider, K.L and Case, S., Damage tolerance and durability of material systems, John Wily & Sons, New York, 2002. 14. Xing, Liqun, ”Progressive Failure of Large Deformation Woven Composites under Dynamic Loading,” PhD dissertation, College of Engineering, University of Connecticut, Department of Mechanical Engineering, 2007

10

15. Reifsnider,K., and Xing, L., “Large-Deformation Constitutive Theories for Structural Composites: Rate-Dependent Concepts and Effect of Microstructure,” Strain, 44#1 (2007) 119-125 16. Huffner, David, “Progressive Failure of Woven Polymer-Based Composites under Dynamic Loading; Theory and Analytical Simulation,” Dissertation, Department of Civil and Environmental Engineering, University of Connecticut, 2008 17. Ning,P., and Chou, T.W., “Closed Form Solutions of the In-plane Effective Thermal Conductivities of Woven-Fabric Composites,” Comp. Sci. Tech., Vol. 55 #1 (1995) 41-48 18. Pham, D.C., “Weighted Effective Medium Approximations for Conductivity of Random Composites,” I. J. Heat & Mass Transfer, Vol. 51, #13,14, (2008) 3355-3361 19. Fazzino, P., “Predictive Methods for Large-scale Progressive Damage in Structural Composites for Aircraft Applications,” Masters thesis, Department of Mechanical Engineering, College of Engineering and Computing, University of South Carolina, November, 2008 20. Fazzino, P., and Reifsnider, K.L., “Electrochemical Impedance Spectroscopy Detection of Damage in Out of Plane Fatigued Fiber Reinforced Composite Materials,” Applied Composite Materials, 15#3 (2008) 127-138. 21. Fazzino, P.D., K.L. Reifsnider, and P. Majumdar, Impedance spectroscopy for progressive damage analysis in woven composites. Composites Science and Technology, 2009. 69(11-12): p. 2008-14. 22. Fazzino, P., Reifsnider, K., Majumdar, P. Impedance Spectroscopy of Fabric Reinforced Composites. Published in the Proceedings of SAMPE 2009 Conference. May 18-21, 2009, Baltimore, MD. 23. K. Reifsnider, P. Majumdar & P. Fazzino, "Material State Changes as a Basis for Prognosis in Aeronautical Structures," J. Aeronautical Society, Vol. 113 #1150 (2009). 24. Ju, G., Reifsnider, K., Xinyu Huang, (2004) “Time dependent properties and performance of a tubular solid oxide fuel cell”, ASME Journal of fuel cell science and Technology, Vol.1 p35. 25. T. Cable and S. Sophie, “A Symmetrical Planar SOFC Design for NASA’s High Specific Power Density Requirements,” Journal of Power Sources, 174 # 1 (2007) 221-227.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Characterisation of Carbon Nanotube Foam for Improved Gas Storage Capability

Armando Peña1, Aimé Guerrero2, Julio Puerta3, Joaquín L. Brito4, Thomas K. Heckel5 1

Universidad Simón Bolívar, Laboratorio de Carbón y Residuales de Petróleo, 89000, Caracas 1080, Venezuela 2 Universidad Simón Bolívar, Departamento de Ciencias de los Materiales, 89000, Caracas 1080, Venezuela 3 Universidad Simón Bolívar, Departamento de Física, 89000, Caracas 1080, Venezuela 4 Instituto Venezolano de Investigaciones Científicas, Centro de Química, Apartado 20632, Caracas 1020-A, Venezuela 5 University of Siegen, Research Group for Material Science and Material Testing, Siegen, Germany

Corresponding authors: [email protected], [email protected] ABSTRACT Advanced fuel cells require efficient hydrogen storage tanks. This study presents preliminary results on a novel compound based on an alumina substrate coated with carbon nanotube foam (CNF) that is expected to improve substantially the hydrogen storage capability. A catalytic chemical vapour deposition (CCVD) technique was applied for obtaining the desired structure. It involved the organometallic compound ferrocene (a simultaneous source of iron and carbon), H2 as reducing gas, and Ar as dragging gas. The CNF-alumina system formed was characterised by means of scanning and transmission electron microscopy (SEM, TEM, resp.). Applying the BET method with N2 as carrier gas, it was found that the novel compound exhibits a high specific surface area, due to the porous morphology, and a high thermal stability. These aspects are very promising for the application intended. The sponge-like structure of the CNF may store hydrogen (or other gases) due to physical adsorption in much larger quantities as compared to conventional storage tanks.

Keywords: Carbon nanotubes foam, Metal precursor; Ferrocene; Catalytic Chemical vapour deposition

1. Introduction Nanotubes have generated great interest since their special geometry results in amazing properties [1,2]. The possible disposition of carbon atoms to form nanotubes, as well as their ability to generate defects, expansions, contractions, twistings and interruptions, is studied continuously. As a consequence a way to define the nanotubes’ parameters, i.e. identifying diameter, quirality and semiconductor or metallic character, has been presented [3]. The production of carbon nanostructures and their characteristic morphology will determine their subsequent applications, e.g. as transmitters of field, in storage of gases (cells of H2), sensors and biosensors, etc [4].

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_2, © The Society for Experimental Mechanics, Inc. 2011

11

12

Particularly, the exploration of some applications requires the knowledge of the surface characteristics of carbon nanotubes, that depend on the routes and synthesis strategies and/or of subsequent processing [5,6,7]. The high adsorption capacity, chemical inertia and facility in the regeneration of carbon nanotubes qualify this material as potential adsorption systems [6]. To. V. Roll, et al., (1997) have produced a new allotropic form of carbon, the carbon nanofoam CNF [8,9]. This carbon nanofoam is composed by a network of carbon nanotubes and was discovered accidentally trying to synthesize nanotubes and fullerenes. This work deals with the synthesis of carbon nanotube foam [9,10] in alumina pipes in a CVD reactor to study its surface properties [11,12]. A simple and direct method is presented of the synthesis of carbon nanotube foam (CNF)[13,14] as internal coating of alumina pipes (system CNF-Al2O3) at 900ºC. The coating with CNF was carried out by catalytic chemical vapor deposition (CCVD) from ferrocene (a simultaneous source of Fe and carbon), H2 as reducing gas, and Ar as dragging gas. The compound that forms consists of ENC-alumina; it was characterized by means of scanning electron microscopy, transmission electron microscopy and adsorption of N2 by the BET method to study its superficial properties. The results show promising characteristics for the possible application as gas storage by means of physisorption [15].

2. Experimental Section

In order to synthesize the CNF the technique of catalytic chemical vapor deposition CCVD process has been applied. Ferrocene has been used simultaneously as metallic precursor (Fe) and as carbon source, hydrogen and argon have been employed as reducing and dragging gas, respectively, and alumina pipes have been applied for collecting the reaction products. Ferrocene has been sublimated previously in a pre-heating chamber at 150ºC, and was dragged to the cracking furnace via H2 and Ar. The operational parameter conditions of the reactor have been: pyrolisis temperature 1000ºC, total gas flow 10 mL/min, gas composition 1:1 v/v of Ar: H2. The experimental set up is shown in figure 1.

Figure 1. Scheme of the reactor CVD employed the synthesis of the CNF Cross sections of the ceramic substrate (Al2O3 pipe) have been characterized before and after the CCVD treatment, by means of scanning electron microscopy SEM, 30kV, and adsorption of N2 at 77K by BET method with a degasification temperature at 350ºC. The carbon nanofoam synthesized in the interior of the pipe was also was characterized with transmission electron microscopy 200kV (TEM).

3. Results and discussion The Al2O3 pipes have been characterized by means of SEM in order to analyze the morphology and composition of the CNF deposited. In figure 2 a section of an alumina pipe is visible, which is covered with coarse grains of a

13

material, having a porous structure. The composition of this structure was determined by means of Energy dispersive X-ray spectroscopy, EDS. Only Al and O were detected. The specific surface area of pipe was determined by BET and revealed180 m2/g.

Figure 2. SEM Al2O3 pipe cross section micrograph. The SEM micrographs in figure 3 visualize that the deposited carbon has been arranged to a large extent to carbon nanotubes inside of the Al2O3 pipes (left image). To a smaller degree nanotube foams have accumulated (right image). The framework consists of carbon nanotubes covered with nanoespheres or nanoparticles also of carbon. The composition of these structures was determined carrying out a chemical analysis by EDS. Only C, Fe, and O were detected. The presence of oxygen is attributed to the oxidation of the surface of the Fe nanoparticle by environmental exposure. The nanotubes grew homogeneously and continuously from the entire Al2O3 surface. The nanotube surface density was approximately 109/m2. The morphology curve and random distribution of the nanotubes are influenced by metal-assisted interaction, the nature of the Al2O3 substrate and surface conditions (porosity, topography) [16,17].

Figure 3. SEM micrograph of carbon nanotubes nanofoam (CNF) on Al2O3, at 20X and 10000X. TEM analysis was carried out on the nanofoam that was presented in Figure 3. The morphology of the foam results from the structure of the nanotubes. In Figure 4, TEM images show a nanocapusle of carbon (left) and a nanowire of carbon (right). The dark regions in the interior in both images are composed of nanoparticles of iron from which the growth of the carbon nanocapsule and nanowire by catalytic reaction can be observed. It is also visible that graphene aligns and grows perfectly in multiple layers around the nanocapsule and –wire. The external average diameter of the carbon nanotubes of which the nanofoam is composed is 22 nm approximately.

14

5 nm

5 nm

Figure 4. TEM micrograph of carbon nanocapsule (left) and multiwalled carbon nanowire (right) The growth of carbon nanotubes from nanocapsules or –wires can be distinguished clearly, as the TEM micrographs in figure 5 illustrate.

20 nm

10 nm

Figure 5. TEM Micrographs of nanocapsule placed on nanowire both of carbon The micrographs that are presented in Figure 6 show overlapping carbon nanocapsules (top and middle) and a high resolution image on how the atoms arrange within a nanocapsule (bottom). The two first micrographs, just like those presented in figure 3, permit to observe the adhesion between the nanoparticles of Fe and the carbonaceous network and also the cohesion between nanocapsules.

20 nm

10 nm

Figure 6. TEM micrographs of nanocapsules of united carbon among if (top and middle) and transformed of Fourier of the interior of the nanocapsules (bottom) The cross section of the alumina pipe with the nanotubes placed was characterized through BET method. The resulting surface area is in the order of (1000±30) m2/g. The calculated specific area of the CNF is 820 m2/g. This suggests that the material is able for gas storage applications [4,5,6,11].

15

In Figure 7 a schematic representation of the elaborate system of this work is presented: a pipe of Al2O3 is shown before and after the deposition of the carbon nanofoam. This pipe of CNF-Al2O3 can be utilized as internal unit of gas storage tanks, specifically in conserving of H2 for fuel cells. The system CNF-Al2O3 prepared shows excellent chemical stability, high surface area and low density.

(Ferrocene)

9000C H2 ,Ar

Al2O3 Emptytubebefore CCVDprocess

Al2O3 Tube with CNFafter CCVDprocess

Figure 7. Schematic representation of the alumina pipe. Before and after deposition in CVD reactor. (Right) alumina tube empty. (Left) Nanofoam/Al2O3 system. In order to verify the system’s mechanical stability (bonding between nanofoam and Al2O3) [18], dry air has been circulated through the pipe. Up to a flow rate of 500 mL/min the CNF remained adherent to the pipe, at a flow rate of 600 mL/min, measurements show that 10 mass% of CNF have been removed. In order to further improve the storage capability, additional porous plugs of alumina were placed at the ends of the pipe, preventing detachment of the nanotubes – Figure 8. These porous covers admit the air flow or another gas at a high rate and can be used as a filter.

Figure 8. Schematic representation of the mechanical stability test. The CNF is maintained adhered to the pipe until a air dry flow of 500mL/min. Conclusions In conclusion, a system was obtained, consisting of CNF/Al2O3 that can be utilized as internal unit of gas storage, in particular as reservoir of H2 for fuel cells. Due to the promising characteristics in the physisorption of gases, a specific area of 1000±30 m2/g of carbon nanofoam placed inside the Al2O3 pipe was obtained in this research work. In general, the carbon nanofoam found in the interior of the alumina pipe consists of multilayered nanotubes and nanocapsules. The distribution of the carbon nanofoam in the alumina pipe is homogeneous, with a nanotubes surface density of 109/m2 approximately. The distribution of the external diameter of the carbon nanotubes is some 22nm. The mechanical stability of the nanofoam inside the alumina pipe was tested satisfactorily up to 500mL/min.

16

References: 1. Dai, Hohgjie. Carbon nanotubes: Synthesis, Integration, and Properties. Accounts of Chemical Research, 35(12):1035-1044, 2002. 2. Mamalis, A.G.; Vogtländer, L.O.G.; Markopoulus, A. Nanotechnology and nanostructured materials: trends in carbon nanotubos. Precision Engineering, 28:16-30, 2004. 3. Endo, M.; Hayashi, T.; Kim, Y.A.; Muramatsu, H. Development and Application of Carbon Nanotubes. Japanese Journal of Applied Physics, 45(6A):4883-4892, 2006. 4. Mann, D. Synthesis of carbon nanotubes. In O’Connell, M.J. editor; Carbon nanotubes Properties and Applications. Taylor & Francis Group, Boca Raton, Florida, 2006. 5. Rakov, E. Chemistry of Carbon Nanotubes. En Gogotsi, Y. editor; Carbon Nanomaterials. Taylor & Francis Group, Boca Raton, Florida, 2006. 6. Hennrich, F.; Chan, C.; Moore, V.; Rolandi, M.; O´Connell, M. The element carbon. En O’Connell, M.J. editor; Carbon nanotubes Properties and Applications. Taylor & Francis Group, Boca Raton, Florida, 2006. 7. Guerrero, A.; Puerta, J.; Gómez, F.; Blanco; F. Synthesis of carbon nanotubes by laser ablation in graphite substrate of industrial arc electrodes. Phys. Scr.T131:1-4, 2008. 8. Rode, A.; Hyde, S.; Gamaly, E.; Elliman, R.; McKenzie, D.; BulcocK, S. Structural analysis of a carbon foam formed by high pulse-rate laser ablation. Applied Physics A, 69:S755-S758, 1999. 9. Gamaly, E.G.; Rode, A.V. Nanostructures Created by Lasers. Encyclopedia of Nanoscience and Nanotechnology. Volume X:1-26, 2004 (ISBN: 1-58883-001-2/$35.00). 10. Chinthaginjala, J.K.; Seshan, K.; Lefferts, L. Preparation and Application of Carbon-Nanofiber Based Microstructured Materials as Catalyst Supports. Ind. Eng. Chem., 46:3968-3978, 2007. 11. Peña A, Guerrero A, Puerta J, Brito J, Cañizales E. Carbon Nanotubes as Catalyst Supports. XVII Catalysis Venezuelan Congress, Memories in ISBN:978-980-12-3931-4, Choroní, Estado Aragua, Venezuela, 2009. 12. Peña A, Guerrero A, Puerta J, Cañizales E, Brito J. Carbon nanowires coated with iron filled nanocage. IX International Congress of Venezuelan Chemistry Society. Cumaná, Estado Sucre, Venezuela, 2009. 13. Rode, A.; Hyde, S.; Gamaly, E.; Elliman, R.; McKenzie, D.; BulcocK, S. Structural analysis of a carbón foam formed by high pulse-rate laser ablation. Applied Physics A, 69:S755-S758, 1999. 14. Gibson, L.; Ashby, M. Cellular Solids: Structure and Properties. 2nd Edition. Cambridge University Press, Cambridge, 1997. 15. R.R. Bacsa, Ch. Laurent, A. Peigney, W.S. Bacsa, Th. Vaugien. High specific surface area carbon nanotubes from catalytic chemical vapor deposition process. Chem. Phys. Lett. 323: 566-571. 2000. 16. Terrado, E.; Redrado, M.; Muñoz, E.; Maser, W.K.; Benito, A.M.; Martínez, M.T. Aligned carbon nanotubes grown on alumina and quartz substrates by a simple thermal CVD process. Diamond &Related Materials 15: 1059-1063, 2006. 17. Castell, A.M.; Raymakers,J.A; Kong, J.; Dai. H. Large Scale CVD Synthesis of Single-Walled Carbon Nanotubes. J. Phys. Chem. B 103:6484-6492, 1999. 18. Guerrero A, Puerta J, Lira J. Estudio experimental de la adhesión interfacial del sistema Sn/Pb-Al2O3. Revista Mexicana de Física S52 (3) 48-50.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Thermoresponsive Microcapsules for Autonomic Lithium-ion Battery Shutdown M. Baginska1, B.J. Blaiszik2, S.A. Odom3, A.E. Esser-Kahn3, M.M. Caruso3, J.S. Moore3,4, N.R. Sottos2,4, S.R. White1,4 1

Department of Aerospace Engineering, University of Illinois at Urbana-Champaign 2 Department of Materials Science, University of Illinois at Urbana-Champaign 3 Department of Chemistry, University of Illinois at Urbana-Champaign 4 Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign ABSTRACT Lithium-ion batteries are commonly used in consumer electronics applications such as cellular phones and computers. However, there are safety concerns, such as external heating, over-charging, high current charging, or physical damage, which prevent their full market acceptance. Autonomic shutdown of lithium-ion batteries, through functionalization of battery electrodes with thermoresponsive microcapsules, is proposed as a fail-safe mechanism. The proposed concept relies on monomer-filled microcapsules that can be triggered to rupture within a desired temperature range and deliver an electrically isolating core to the electrode surface to shut down the battery cell. Preparation of thermoresponsive microcapsules that can be triggered to rupture using a low-boiling point solvent and deliver a thermally polymerizable core is described. Additionally, we demonstrate that the rupture temperature can be controlled by appropriate selection of microencapsulated trigger solvents. Initial work on the coating of battery materials with thermoresponsive spheres is also described. KEYWORDS: lithium-ion batteries, microcapsules, emulsion polymerization, thermoresponsive, autonomic Introduction Li-ion batteries are the focus of significant research due to their potential for high energy density, lack of 1 memory effect and ability to handle hundreds of charge/discharge cycles . Li-ion batteries are predominantly used in consumer electronics; however, improvements in safety are required before their full market acceptance. Fail-safe shutdown in the event of overcharge, short circuits, or exposure to hightemperature environments could make Li-ion batteries a viable choice for defense, automotive, and aerospace applications while improving the safety of devices. The current defense against internal battery 2 overheating, the shutdown separator , is not ideal since localized heating can result in separator shrinking and melting, leading to physical electrode contact and potential explosions. One approach to improved safety in Li-ion batteries is the incorporation of functional microcapsules. 3 4 5 6 Previously, microcapsules used in fragrances , food processing , textiles , and self-healing materials have relied mainly on mechanical rupture to break the shell wall and deliver their contents. However, improvement in microencapsulation techniques has led to the development of microcapsules capable of 7 8 9 10 responding to environmental stimuli, such as pH , electric field , temperature , photosensitivity , and 11 magnetic fields . Li-ion battery safety could be enhanced by the incorporation of thermoresponsive microcapsules in order to trigger shutdown mechanisms. The proposed concept is shown schematically in Figure 1 and can be achieved by either of two approaches. In one approach, battery electrodes are functionalized with monomer-filled microcapsules, which are then triggered to rupture within a desired temperature range and deliver a thermally polymerizable core to the electrode surface, electrically isolating the electrodes and effectively shutting down the battery cell. In the second approach, the electrode is coated with capsules that undergo a phase transformation (melt) at a predetermined temperature, and coating the electrodes with a non-conductive separator.

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_3, © The Society for Experimental Mechanics, Inc. 2011

17

18

(a) (b) Figure 1: Schematic representation of proposed shutdown concept for lithium-ion batteries.

The temperature profile which dictates the design of thermoresponsive microcapsules for shutdown of a Li-ion battery is shown schematically in Figure 2. The normal operational temperature for a Li-ion battery is between temperature of 40°C and 60°C. However, thermal hotspots (>100°C) may occasionally develop on battery electrodes. It is desirable to quench these thermal hotspots, by having capsules in the immediate vicinity trigger, though shutting the battery down completely may not be necessary. When the internal battery temperature continues to increase (>130°C), global shutdown should be initiated.

Figure 2: Overview of areas of intervention in failing lithium-ion batteries.

Initial work on triggering mechanisms has been completed, including rupture experiments of several microcapsule core-shell chemistries conducted with direct observation of microcapsules on a microscopemounted hot stage. The current triggering mechanism consists of encapsulating low-boiling point solvents which volatilize, increasing the internal pressure in the capsules until rupture occurs. While the low-boiling point solvent is the trigger, the rest of the core contains the thermally polymerizable monomer to shut down a battery cell undergoing thermal runaway. Initial work on thermal transition cores has also been conducted, including a screening of candidate meltable polymers with melting points at approximately 100°C. Materials and Methods Microcapsule Preparation Thermoresponsive microcapsules containing a low-boiling point solvent, pentane (b.p. 36°C), and thermally polymerizable monomer, divinylbenzene (DVB) were prepared by interfacial polymerization of polyurethane and water. Polyurethane was chosen as the shell wall material because of its robustness compared to brittle urea-formaldehyde polymer. Further, urea-formaldehyde encapsulations require

19 heating to 55°C; however, using pentane as a low-boiling point solvent trigger prohibits this. Polyurethane encapsulations can be achieved at room temperature, enabling the encapsulation of low-boiling point solvents. Polyurethane was dissolved in a solution of divinylbenzene, pentane, and cyclohexanone, and added to an aqueous surfactant solution of poly(ethylene-co-maleic anhydride) stirred at 450 rpm. Microcapsules were allowed to react for 6 hours and then were filter dried into a coarse powder. Thermal transition spheres were prepared by pouring a molten microcrystalline wax (m.p. ~95°C) into an aqueous solution of Brij 76 and sodium dodecyl sulfate (SDS) surfactants heated to 97°C. The resulting suspension of wax droplets was stirred at 4000 rpm and then rapidly cooled with ice water to solidify the wax. The resulting wax microspheres were centrifuged 5x to rinse off the surfactant solution, filter dried, and sieved to yield a fine powder. Electron Microscopy Scanning electron microscopy (Philips XL30 ESEM-FEG) was used to investigate the morphology of the capsules and battery materials. Samples were prepared by mounting on an SEM stage coated with carbon tape and sputter coated (Denton Vacuum Desk II TSC Turbo Sputter Coater) with a coating of gold-palladium. The ESEM imaging was conducted at 5.0 kV accelerating voltage with a spot size of 3.0 to minimize sample charging. Hot Stage Techniques The temperature increase in a battery experiencing thermal runaway can be as high as 150°C in one 12 minute . Initial hot stage experiments were conducted using a Linkam HFS 91 hot stage in conjunction with a TMS 92 temperature controller. The rupture of microcapsules was studied at the maximum heating rates of 130°C/min. Microcapsules were heated until rupture in silicone oil on a glass substrate. The amount of intact microcapsules as a function of temperature was recorded using custom-developed LabView software. Results and Discussion Capsule Morphology The synthesized polyurethane microcapsules were spherical in shape and had a dense, smooth, shell wall as shown in Figure 3. The average diameter was 236 µm with a standard deviation of 67.5 µm as measured by optical inspection. The shell wall thickness was an average 3.5 µm, as determined by SEM observation of shell wall cross-sections. The shell wall thickness can be tailored by varying the amount of 13 polyurethane in the microcapsule preparation . Microcapsules can also be scaled down in size by 13 increasing the stir rate in the preparation bath depending on application requirements .

(a) (b) Figure 3: a) Smooth shell wall of a polyurethane microcapsule. b) Cross-section of microcapsule to illustrate shell wall thickness.

20 The thermal transition spheres that were produced using the procedure described in the Methods section and are shown in Figure 4 below. Surfactant choice was critical in the successful production of spheres; a combination of a 1 wt% aqueous solution of SDS and a 1 wt% aqueous solution of Brij 76 surfactants were necessary to obtain well-formed spheres.

Figure 4: Thermal transition sphere morphology.

Microcapsule Trigger Evaluation Based on hot stage thermal rupture testing, thermoresponsive microcapsules do not share a distinct, common rupture temperature, but instead rupture gradually, as shown in Figure 5. The rupture profile for the pentane/DVB thermoresponsive microcapsules is shown by the blue curve, and rupture begins at a temperature of approximately 110°C. Microcapsules with an ethyl-phenyl acetate core (b.p. 228 °C) and divinylbenzene were prepared as a high-boiling point solvent and heated at the same rate as the thermoresponsive capsules. The ethyl-phenyl acetate (EPA) microcapsules, whose rupture profile is shown by the red curve in Figure 5, demonstrated a more uniform rupture, occurring sharply at around 250°C. As seen from Figure 5 below, the rupture onset temperature for the two capsule types varies by approximately 125°C.

Figure 5: Percentage of intact pentane and EPA-triggered thermoresponsive microcapsules as a function of temperature.

Figure 6a contains an optical micrograph of microcapsules containing pentane, the low-boiling point solvent, and DVB, which serves as the thermally polymerizable monomer. In order to obtain maximum

21 core delivery, microcapsule rupture must occur prior to the thermal polymerization temperature of the monomer contained inside. Figure 6b shows a microcapsule post-rupture, and as well as the partially dispersed polymerized film. DVB was shown to thermally polymerize starting at 150°C by differential scanning calorimetry. The capsule shown in Figure 6b likely ruptured at a temperature close to 150°C, and it is evident that not all of the monomer was delivered. The dark region within the broken shell wall is a result of DVB that has polymerized prior to being delivered. This can still be optimized by selecting a solvent that would induce rupture well ahead of the anticipated thermal polymerization.

(a)

(b)

Figure 6: a) Intact thermoresponsive microcapsules. b) Ruptured microcapsule after heating at 130°C/min.

Thermal Transition Spheres Thermal transition spheres were spin coated on a polyethylene separator using a 2 wt% spheres in Nmethylpyrrolidone (NMP) solution and polyvinylpyrrolidone (PVP) as a binder (Figure 7a). The spheres appeared to adhere to the separator, but further studies are necessary to determine their effectiveness at preventing ionic conductivity in a lithium-ion battery in solid or molten form. Figure 7b shows a dense, molten wax film on a smooth polyethylene separator. The molten wax has a rough, grooved morphology and when molten, must sufficiently interfere with ionic conductivity that ions cannot flow throw the separator pores.

(a) (b) Figure 7: a) Monolayer of thermal transition microspheres spin coated onto a polyethylene separator. b) Detailed view of molten wax film dispersed over a polyethylene separator.

Conclusions Thermoresponsive microcapsules containing pentane, a low-boiling point solvent and DVB, thermally polymerizable monomer were produced. Under similar heating conditions (130°C/min), thermoresponsive

22 microcapsules were observed to begin to rupture over 100 degrees earlier than comparable in size ethylphenyl acetate microcapsules. The rupture was observed to be progressive, though onset temperature varied depending on the encapsulated solvent. A film of polymerized divinylbenzene near broken microcapsules was observed in both capsule types, verifying the delivery of the monomer. However, the correlation between capsule diameter, shell wall thickness, and rupture temperature must still be investigated. The thermoresponsive microcapsule system can be tuned to the application requirements by selecting the appropriate low boiling point solvent for the desired rupture temperature, as well as a monomer that thermally polymerizes at the desired temperature. In order to extend these microcapsules to lithium-ion battery applications, microcapsules must be able to withstand the physical, chemical, and electrochemical battery environment. For example, microcapsules 14 must have sufficient electrolyte stability in salts such as LiPF6 or LiClO4 and solvents such as ethylene carbonate (EC), ethyl methyl carbonate (EMC) or dimethyl carbonate (DMC). Additionally, it is vital that capsules incorporated into batteries are neither oxidized nor reduced and lithium intercalation does not occur in the polymer being examined. These experiments are currently in progress. Thermal transition spheres were synthesized using a microcrystalline wax and a procedure for forming a monolayer on a battery separator was developed. After heating, a dense, rough film was observed over the polyethylene substrate. The effectiveness of this film in preventing ionic conductivity is currently being studied. Acknowledgements This research was supported as part of the Center for Electrical Energy Storage: Tailored Interfaces, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences. We would also like thank the autonomic materials group member David McIlroy for developing LabView testing software and Scott Robinson at the Beckman Institute Imaging Technology Group for his help with scanning electron microscopy. References 1. Armand, M., Tarascon, J. Building better batteries. Nature 451, 652-657(2008). 2. Sheng, Z.S., A review on the separators of liquid electrolyte Li-ion batteries. Journal of Power Sources 164, 351-64(2007). 3. Zhao, Q., Li, B. pH-controlled drug loading and release from biodegradable microcapsules. Nanomedicine: Nanotechnology, Biology and Medicine 4, 302-310(2008). 4. Desai, K.G.H., Park, H.J. Recent developments in microencapsulation of food ingredients. Drying Technology: An International Journal 23, 1361(2005). 5. Sánchez, P. et al. Development of thermo-regulating textiles using paraffin wax microcapsules. Thermochimica Acta 498, 16-21(2010). 6. Blaiszik, B. et al. Microcapsules filled with reactive solutions for self-healing materials. Polymer 50, 990-997(2009). 7. Qi, W. et al. Triggered release of insulin from glucose-sensitive enzyme multilayer shells. Biomaterials 30, 2799-2806(2009). 8. Yoshida, M. et al. Permeability control in electro-sensitive microcapsules with immobilized ferroelectric liquid crystalline segments. Journal of Polymer Science Part A: Polymer Chemistry 46, 1749-1757(2008). 9. Chu, L. et al. Preparation of micron-sized monodispersed thermoresponsive core−shell microcapsules. Langmuir 18, 1856-1864(2002). 10. Wang, X. et al. A novel method to control microcapsule release behavior via photo-crosslink polyurethane acrylate shells. Journal of Applied Polymer Science 113, 1008-1016(2009). 11. Hu, S. et al. Controlled rupture of magnetic polyelectrolyte microcapsules for drug delivery. Langmuir 24, 11811-11818(2008).

23 12. Yufei C., Li S., Evans J., Modeling studies on battery thermal behaviour, thermal runaway, st thermal management, and energy efficiency. IECEC 96. Proceedings of the 31 Intersociety Energy Conversion Engineering Conference. 11-16 Aug. 1996. Vol. 2. 1465-70(1996). 13. Caruso, M. et al. Robust, double-walled microcapsules for self-healing polymeric materials. Submitted to ACS Applied Materials and Interfaces. 2010. 14. G.E., Blomgren, Liquid electrolytes for lithium and lithium-ion batteries, Journal of Power Sources, 2003, pp. 326-329.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Service Life Prediction of Seal in PEM Fuel Cells

Tong Cui1 1

C.H. Chien2

Y.J. Chaoa, 3

J. Van Zee1

Department of Chemical Engineering, University of South Carolina, Columbia, SC 29208, USA 2

3

C-W. Lin2

Department of Mechanical and Electro-Mechanical Engineering, National Sun Yat-Sen University, Taiwan

Department of Mechanical Engineering, University of South Carolina, Columbia, SC 29208, USA

ABSTRACT Proton exchange membrane fuel cell (PEMFC) is a promising power source for automobiles in the near future. During operation, there are gases and liquids inside the fuel cell. Sealing around the peripherals of the cell is therefore required to prevent the gases/liquids inside the cell from leaking. Polymers are usually used as the sealing or gasket materials. They in general possess the property of viscoelasticity. The stress relaxation behavior of liquid silicon rubber, a type of polymer, is studied in this article. Applying the time-temperature superposition, master curve is generated for prediction of service life of this material used as seals in PEMFC. Keywords: Stress relaxation, WLF equation, lifetime prediction. 1. INTRODUCTION PEMFC is a promising power source for potable, stationary, and automotive applications because of its high efficiency and clean reactants. Currently, most researches in this area are focusing on catalyst and membrane that are the heart of the PEMFC [1-2]. Of equal importance to any commercial product is the durability. According to US Department of Energy, PEMFC for transportation applications must operate reliably for at least 5,000 hours at a temperature range from - 40 to 80°C [3]. To satisfy this requirement, all components in the PEMFC including the seals have to be durable enough to bear the load at the normal operation temperature. Little work on gasket or seals in fuel cells had been reported in the literature. Effective design of seals should ensure not only is it effective initially, but also durable in long term. However, field tests often show that one of the limiting factors in durability of PEMFC is the leaking of the liquids and/or gas from inside. This leaking or loss of functionality of seals can be either from chemical attack, long term mechanical degradation, or both to the seal material. Chemical and mechanical degradation of gasket materials in either PEMFC solution or an accelerated solution was studied by Tan, et al. [4-7]. Gasket materials such as ethylene propylene diene Monomer (EPDM) rubber and Fluoroelastomer were first aged in simulated PEM fuel cell environment. Chemical and mechanical properties were then assessed. Furthermore, leachants into the solution from the gasket materials, which may be detrimental to the electrical-chemical reaction, were also studied. a

Author for correspondence, [email protected], Tel: 803-777-5869

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_4, © The Society for Experimental Mechanics, Inc. 2011

25

26

Many seals are made of polymeric materials. Polymers are well known to exhibit strong creep or stress relaxation behavior. Relaxation of stress of a polymeric seal reduces the effectiveness of the sealing function. The present paper is concerned with the stress relaxation behavior of liquid silicone rubber (LSR) at different temperatures. The objective is to predict the sealing life in PEMFC applications when LSR is used as the seal or gasket materials. Compression stress relaxation tests were performed on LSR in both ambient air and wet conditions, at different applied strain levels, and at elevated temperatures. A stress relaxation model is adopted for this material. The Williams-Landel-Ferry (WLF) methodology [8] is used for time-temperature shift at various conditions to form the master curve at a reference temperature. Service life of the seals was then predicted based on the master curve. 2. MODELS 2.1 Viscoelastic model Viscoelastic materials have the properties of both viscosity and elasticity. A mathematical model for viscoelastic behavior therefore would include elements representing both. Typical models, either linear or nonlinear, include springs and dashpots. Spring is for elastic deformation and it can respond simultaneously to any applied load; whereas, dashpot reacts like viscous fluids, and moves at a rate proportional to the stress. A basic model for viscoelastic materials, called the Maxwell model, is represented by a dashpot and a spring connected in series. Several extensions to the Maxwell model have been made [9]. For example, Rabinovich [10] used a nonlinear Maxwell’s equation to predict the stress relaxation. Alfrey [11] connected the springs and dashpots in parallel to generate a new model to describe the relaxation curve. These models have been successfully applied to predict the stress relaxation behavior at certain conditions. A generalized Maxwell model would result in a solution as n

 (t )       i exp(  i 1

t ) i

(1)

  is the stress stabilized after a long time,  i depends on the applied strain lever and material properties, t is time and  i is a material constant. This equation is often called the Prony series and is the most

Where

popular linear model for stress relaxation of polymeric materials. As shown later, the SLR material studied in this work exhibits nearly linear behavior up to 25% strain which is in the range of interest in sealing application. As a consequence, the Prony series is adopted in this work.

2.2 Time-temperature superposition Stress relaxation is, to a large extent, influenced by temperature. For polymeric materials, the Williams-LandelFerry (WLF) equation [8], as shown in equation (2), describes the relation between time t and temperature T.

G r (t , T2 )  G r ( A(T1 , T 2 )t , T1 )

(2)

C1 (T2  T1 ) C 2  (T2  T1 )

(3)

where A is given by log[ A(T1 , T2 )]  

As shown above, there are two parameters in the WLF function, namely C1 and C2, and they are only dependent

27

on the reference temperature T1. The essence of the WLF equation is that time and temperature are equivalent and therefore the time function of stress at one certain temperature has the same shape of the same function at another temperature. It is a mathematical application of the Boltzmann's superposition principle. In practice, the stress relaxation curve determined at certain temperature can be horizontally shifted to another temperature when presented in logarithmic scale. The beauty of this time and temperature superposition lies in that it can avoid the long term tests on creep or stress relaxation of polymers at a certain temperature, by testing the material at a higher temperature but with shorter time. The WLF time-temperature superposition has been extensively used in studying creep and stress relaxation of polymeric materials. 3. EXPERIMENT 3.1 Material and preparation Liquid silicon rubber (LSR), a commercially available high purity platinum-cure silicone, is a good candidate for gasket in PEMFC. It is inexpensive and easy to mould into custom shapes and design. Its glass transition temperature (-40°C) is relatively low and therefore it remains flexible, elastic, and retains its properties up to 300°C. Although its surface chemistry may change over time when exposed to high concentration PEMFC solutions, its bulk mechanical properties remain unchanged [5]. However, like other polymers, LSR also shows viscoelastic behavior especially at high temperatures. LSR in sheet form was obtained from manufacturer and cut into round buttons for compression stress relaxation (CSR) tests. A cylindrical disc with a 13 mm diameter (D) and 6.3 mm height (H) was used according to ASTM D6147 [12]. 3.2 Instrument A stress relaxation tester (Elastocon AB, Sweden) was used in this experiment. This instrument consists of three independent rigs. Each rig has a load cell and a chamber that can test a sample at a certain temperature. Liquid or humidity can also be applied in the container inside which the test specimen is compressed. The rigs are connected to a data acquisition device and the temperature and current force applied to the specimen can be recorded by a computer. 3.3 Experimental procedures Table 1 lists the stress relaxation tests performed. They are generally at high temperatures and various strain levels. The wet condition was done by putting de-ionized water (DI water) in the container so the specimen was completely immersed in the DI water throughout the test. This procedure is validated by the fact that the specimens were still immersed in DI water after completing the test. The test procedure follows the ASTM D6147, such as pre-conditioning of the test specimen and the 25% strain applied at the test temperature. Table 1 Compression stress relaxation tests Temperature Strain level (%) (°C) 25 25 In ambient air 70 25 120 25 70 10 70 15 In DI water 70 25 100 25 120 25

28

4. RESULT AND DISCUSSION 4.1 Material model for compression stress relaxation Figure 1 shows the stress relaxation curves obtained from strain levels at 10%, 15% and 25% for specimens in DI water at 70°C. In general, they all show a nearly exponential decay of stress with time. The isochronous stress strain curves generated from the data in figure 1 are shown in figure 2. It can be seen that at times from 0 to 1600 hours stress is nearly proportional to strain. And, therefore, the linear Prony series shown as eqn.(1), is proved to be adequate for the current work.

1.4 10% strain 15% strain 25% strain

Yong's Modulus (MPa)

1.2 1.0 0.8 0.6 0.4

Liquid silicone rubber In DI water

0.2 0.0 0

1000

2000

3000

4000

5000

Time (hour)

Figure 1 Stress relaxation curves of LSR at three strain levels, 10%, 15% and 25% in DI water It is noted that when the Prony series can be rewritten as

E (t ) 

n  (t ) 1 t  [     i exp(  )] 0 0  i 1 i n

 E   Ei exp(  i 1

(4)

t ) i

The “Young’s Modulus E(t)” is therefore a function of time only, regardless of the level of the applied constant strain  0 . Figure 3 shows the time dependent Young’s modulus at the three different strain levels. Although not perfect, the three curves are pretty close to each other.

29

1.4

0 hour 100 hours 1000 hours 1600 hours

1.2

Stress (MPa)

1.0 0.8 0.6 0.4 0.2 0.0

0

5

10

15

20

25

30

Strain (%)

Figure 2 Isochronous stress-strain plot of LSR at different times (with fitted lines) Table 2 the coefficients of three term E∞ (MPa) E1 (MPa) E2 (MPa) 0.168802 0.133157 0.770912 τ1 (Hour) τ2 (Hour) 18.577 254.065

Prony series E3 (MPa) 0.176865 τ3 (Hour) 2500

Young's Modulus (MPa)

The Prony series of equation (1) with three exponential terms are selected to fit with the stress relaxation curve at 25% strain. The coefficients in the fitted Prony series are shown in Table 2. From the practical point of view, these results demonstrate that the test data at one strain level can be transferred or applied to another strain level through the Maxwell model of equation (1) or (4). Therefore, for convenience one strain level in stress relaxation test is sufficient in laboratory for studying the sealing behavior of PEMFC.

Strain 10% Strain 15% Strain 25%

6

4

Liquid silicone rubber In DI water

2

0 0

1000

2000

3000

4000

5000

Time (hour)

Figure 3 Time dependent Young’s Modulus E(t) at three different strain levels 4.1 Compression stress relaxation test results with specimens in DI Water

30

Figure 4 shows the normalized stress relaxation curves at three temperatures, 70 °C, 100 °C, and 120 °C. The tests for 70 °C and 100 °C were ended after about 1600 hours, since the relaxation rate at this time is sufficiently low. Specifically, the incremental relaxation rate at 1600 hrs is at 0.002 %/hr for the case of 70 °C and 0.001 %/hr for the case of 100 °C which would introduce no significant additional relaxation. For instance, in another 1000 hours after 1600 hours, the additional stress relaxation would be less than 2 % and 1 %, respectively, for the two cases. It is noted that the test at 120°C was stopped at around 800 hrs due to leakage of the high water vapor pressure in the container. Figure 4 shows the exponential decay nature of the stress with time similar to that described by the Maxwell model. It also shows that higher temperature contributes to faster stress relaxation. Figure 5 shows the same data as in figure 4, but in logarithmic scale. Linearity of the 70 °C curve indicating physical relaxation is observed throughout the test time. Linearity for the 100°C curve is only up to 700 hours. After that, the data gradually deviated from the linearity indicating chemical relaxation has occurred. The linear region for the 120°C curve is only limited in about 400 hours. These data demonstrate that chemical relaxation is prone to occur at higher temperature.

Normalized Stress (%)

120

Liquid silicone rubber, In DI water with 25% strain

100

o

70 C, initial stress 1.24MPa

80 60 40

100 oC, initial stress 1.34MPa o

20

120 C, initial stress 1.44MPa

0 0

200 400 600 800 1000 1200 1400 1600 1800 Time (hour)

Normalized Stress (%)

Figure 4 Stress relaxation data of liquid silicone rubber in DI water at three temperatures

70 oC 100 oC o 120 C

100

Liquid silicone rubber, In DI water with 25% strain 10 1

10

100

1000

10000

Time (hour)

Figure 5 Stress relaxation data of liquid silicone rubber in DI water at three temperatures in logarithmic scale Using the WLF time-temperature shift, data in figure 5 is curve-fitted to equation (4) for the determination of the coefficient C1 and C2 with 70°C as its reference temperature T1. Figure 6 shows the curves shifted to the reference temperature of 70oC. . In this figure, all three curves are overlapped to form the master curve at this

31

particular reference temperature. The master curve can then be shifted to any other temperature with the WLF function. Note that to avoid long time tests, e.g. years, for service life prediction, laboratory tests are typically conducted within a reasonable test time, e.g. weeks, at a higher temperature than its operating temperature. Applying the WLF time-temperature superposition principle, the test curve at the higher temperature can then be shifted to a lower temperature, e.g. the operating temperature, corresponding to a longer test time. This is the beauty of time and temperature superposition and WLF function. For example, assuming the LSR is used as the seal in FC and the operating temperature is at 70°C, the service life of the seal would be predicted around 6000 hours (8.3 months) from figure 6, if 60% of the initial sealing stress has to be maintained. In this example, we have set the criterion that when the stress in the seal is fewer than 60% of its initially applied stress the seal will not hold and leakage would occur. The choice of this cutoff percentage or stress obviously depends upon the internal pressure of the FC and other practical considerations.

Normalized Stress (%)

o

70 C o 100 C o 120 C

100 60

Liquid silicone rubber In DI water with 25% strain

10 101

102

103

104

105

Time (Hour)

Figure 6 Master curve of stress relaxation of LSR at a reference temperature of 70° 4.2 Compression stress relaxation test results with specimens in ambient air LSR specimens were also tested in ambient air at three temperatures, i.e. 25°C, 70°C and 120°C, with 25% applied strain. Results will be reported elsewhere. 5

CONCLUSION

This paper presents experimental studies of stress relaxation behavior of LSR in DI water and ambient air, different temperature, and different applied strain levels. From the experimental data and analysis, it is concluded: 1. A linear Maxwell viscoelastic model represented by the Prony series is sufficient for describing the stress relaxation of this material. 2. Applying the time-temperature superposition principle and WLF function, master stress relaxation curves can be constructed for any given temperature. 3. The master curves can be used to estimate the service life of the LSR seal when a cutoff criterion is set. 6. ACKNOWLEDGEMENTS This study is sponsored by Graduate Students Research Abroad Program (Grants no.97-2917-I-110-108) from National Research Council, Taiwan, and National Sun Yat-Sen University Study Abroad Scholarship Award, to the

32

second author. In addition, the support from the US Department of Energy (DE-FC36-06G086041 and DE-FG3608GO88116) to the University of South Carolina Research Foundation and the NSF Industry/University Cooperative Research Center for Fuel Cells at the University of South Carolina are greatly appreciated. REFERENCES [1] Patankar, K.A., Dillard, D.A., Case, S.W., Ellis, M.W., Lai, Y-H., Budinski, M.K. and Gittleman, C.S. “Hydrothermal Characterization of The Viscoelastic Properties of Gore-Secelt® 57 Proton Exchange Membrane”, Mechanics of Time-Dependent Material, 12, p.221-236, 2008. [2] Patankar, K.A., Dillard, D.A., Ellis, M.W., Budinski, M., Case, S. W., Lai, Y-H, and Gittleman, C., “Nonlinear Viscoelastic Characterization and Modeling of Proton Exchange Membranes”, Proceedings of Fuel Cell 2009, The The 7th International Conference on Fuel Cell Science, Engineering and Technology, June 8-10, 2009. [3] US Department of Energy, “Development and Demonstration Plan Planned Activities for 2003–2010, Sect”, 3.4.4, p. 7., 2003. [4] Tan, J.Z., Chao, Y.J., Yang, M., Lee, W. K. and Van Zee, J.W., “Chemical and Mechanical Stability of a Silicone Gasket Material Exposed to PEM Fuel Cell Environment”, International Journal of Hydrogen Energy, xxx (in press), pp. 1-7., 2010. [5] Tan, J. Z., Chao, Y.J., Van Zee, J.W. and Lee, W.K., “Degradation of Elastomeric Gasket Materials in PEM Fuel Cells”, Materials Science and Engineering, A 445–446, pp. 669–675, 2007. [6] Tan, J.Z., Chao, Y.J., Van Zee, J.W., Li, Xiaodong, Wang, Xinnan and Yang, Ming, “Assessment of Mechanical Properties of Fluoroelastomer and EPDM in a Simulated PEM Fuel Cell Environment by Microindentation Test”, Materials Science and Engineering, A 496, pp. 464-470, 2008. [7] Tan, J.Z., Chao, Y.J., Li, Xiaodong and Van Zee, J.W., “Degradation of Silicone Rubber under Compression in a Simulated PEM Fuel Cell Environment”, Journal of Power Source, 172, pp. 782-789, 2007. [8] Sperling, L.H., Introduction to Physical Polymer Science. John Wiley & Sons, Inc. New York. 1986. [9] Junisbekov, T.M., Kestelman, V.V. and Malinin, N.I., Stress Relaxation in Viscoelastic Material, Science Publishers, Inc., New Hampshire, 2003. [10] Rabinovich, A.L., Introduction to the Mechanics of Reinforced Polymers, Nauka, Moscow, pp.482, 1970, [11] Alfrey, T., Mechanical Properties of High Polymers, Foreign Lit. Publ., Moscow, pp.619, 1952, [12] ASTM D6147, Test Method for Vulcanized Rubber and Thermoplastic Elastomer-Determination of Force Decay (Stress Relaxation) in Compression, ASTM International, West Conshohochen, PA. 2002.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Experimental Validation of A Constitutive Model for Ionomer Membrane in Polymer Electrolyte Membrane Fuel Cell (PEMFC)

Wonseok Yoon1∗, Xinyu Huang2, and Roham Solasi3

1

Florida Solar Energy center Department of Mechanical, Materials, and Aerospace Engineering University of Central Florida 4000 Central Florida Blvd, Orlando, FL32816-2450 Email : [email protected] 2

Mechanical Engineering Department and SOFC Program College of Engineering and Computing University of South Carolina 300 Main Street Columbia, SC 29208 Email : [email protected] 3

Sensata Technologies 529 Pleasant St. MS B-37 Attleboro, MA 02703 Email : [email protected]

∗

Corresponding Author; Wonseok Yoon (E-mail : [email protected])

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_5, © The Society for Experimental Mechanics, Inc. 2011

33

34

ABSTRACT The authors have implemented a nonlinear constitutive modeling for ionomer membranes with application in polymer electrolyte membrane fuel cells. The constitutive model features multiplicative decomposition of viscoelastic and plastic deformation gradient tensors, micro-mechanism inspired viscous flow rule, nonlinear viscoelastic Bergström-Boyce model, and hydration and temperature dependent elastic modulus of ionomer membrane. In this work, the authors attempted to experimentally validate the constitutive model. Experimental results obtained from uniaxial tension tests of perfluorosulfonic acid membrane (e.g Nafion® from DuPont.) under wellcontrolled environments were used to fit the model parameters, which are subsequently used in a finite element (FEM) code to predict stress and deformation of ionomers in complex loading cases. The proposed model after validation showed fairly good predictive capabilities for the large deformation behavior of Nafion membrane subjected to the creep and relaxation at different strain rates in a wide range of relative humidity. 1. INTRODUCTION Nafion, a solid electrolyte in polymer electrolyte membrane fuel cells (PEMFCs), has a unique microstructure. It consists of hydrophobic PTFE (polytetrafluoroethylene)-like back bone and water clusters surrounded by hydrophilic sulfonic acid side chains. This types of membranes selectively conduct protons and prevents reactant gases from crossover. In PEMFCs, the gas crossover negatively affects membrane durability as well as fuel cell performance. Particularly, membrane can be mechanically weakened without significant loss of materials resulting from chemical degradation after open circuit voltage (OCV) condition[1] and mechanical defects such as cracks and pinholes generated due to the degradation can result in the catastrophic failure of fuel cell. In fuel cell operation, typical operating temperature can range from 60°C to 90°C with humidified gas feeds. Water absorption/desorption plays a critical role in ionomer membranes’ mechanical behavior as well as proton conductivity. A repeated temperature and humidity fluctuation (hygro-thermal cycling) negatively affects the membrane stability through the reduction in mechanical properties such as strength and toughness. The membrane electrode assembly (MEA) is typically constrained by gas diffusion media and bipolar plates in a PEMFC; membrane dry-out induces biaxial tensile stress and this stress level can be amplified due to stress concentration at the membrane edge[2]. Therefore, the prediction of mechanical stress in the membrane is necessary to better understand the membrane failure mechanisms in PEMFCs. The authors have recently implemented a constitutive model for ionomer membrane based on Bergström and Boyce’s model[3]. In the previous paper[4], stress-strain data of Nafion 111 membrane in vapor and liquid equilibrated condition were compared with those from finite element calculation and the FEM results showed fairly good agreement with an experimental data at varying temperature and humidities. The viscoelastic response of Nafion subjected to creep has been investigated by Benziger’s group [5, 6], and it has been found that the microstructure changes in the hydrophilic domain of Nafion are responsible for the complex effects of temperature and water activity on mechanical properties. In relaxation tests[7], the response is not only dependent on temperature and water activity, but also on the imposed strain level. They have concluded that these responses result from the microphase restructuring induced by the combination of temperature, absorbed water, and applied stress acting together. Creep was also considered as a mechanism for the pinhole formation in the fuel cell membrane through membrane thinning as a result of creep. Therefore, understanding on the time-dependent behavior of membrane is a critical to predict the material lifetime. In this paper, the author attempted to further validate the previous model using stress and strain response of Nafion under complex loading conditions such as creep and relaxation in different temperature, relative humidities, and loading rates. 2. EXPERIMENTAL Stress and strain data from uniaxial tensile tests were measured in controlled humidity and temperature environments. The chamber was mounted on an MTS Tytron servo-mechanical test frame as can be seen in ® Figure 1a. All creep and relaxation experiments have been conducted on dispersion-cast Nafion NR111 in the acid form. Approximately, the samples with gauge length of 30 mm and width of 6 mm were cut by sharp scalpels or roller blades. All the samples were given a sufficient time in the chamber that is required for membrane to reach equilibrium with the desired environmental conditions of temperature and relative humidity before the tests. In the relaxation tests, the specimens were extended with different initial strain rates up to the engineering strain of 0.5 and then displacement was fixed for one hour in order to observe stress relaxation. For the creep tests, stress was applied uniaxially to the sample by hanging a dead weight to one side of the specimen with nylon fishing string in Figure 1b. The nylon string was passed through pulleys and a plastic container was attached to it.

35

The weight was a container filled with steel shot to give an appropriate mass. Creep strain was measured with a linear variable displacement transducer (LVDT) connecting to the bottom of the plastic container and the LabView software and data acquisition hardware were used to record the data. Load Cell

Grips Nylon String

Rod

Specimen

Pulley

Teflon Block Weight

Weight Load Cell

LVDT

Specimen Chamber LVDT

a

b

Figure 1. (a) Picture of the creep setup along with the environment chamber (b) Schematics of the creep testing setup 3. CONSTITUTIVE MODEL OF IONOMER MEMBRANE The proposed model was inspired by micromechanism of polymer deformation such as reptational dynamics[8, 9] and microstructure of ionomer membranes such as amorphous network (entanglements), crystallite, and water clusters. Nafion® is microphase separated into hydrophobic domains which consist of the relatively stiff PTFE backbone and hydrophilic domains confined by aggregates of sulfonate end groups at ionomer side chains. It is, therefore, likely that physical entanglements, intermolecular interactions between the PTFE backbone, and ionic interactions between the side chains would determine the mechanical behavior of a membrane. The structural changes in Nafion® film in an ambient environment and under stretching has been investigated using small/wide-angle X-ray scattering and birefringence experiments[10, 11]. In these reports, Nafion® structure is considered a collection of elongated polymeric aggregates in bundles. Upon stretching, large bundles rotate so that the elongated aggregates are aligned and ordered with each other at small strains. At high strains, the aggregates within the bundles are aligned due to sliding or disentangling of the aggregates from each other. Relationship between birefringence and mechanical properties in relaxation experiments also indicated that the stress relaxed much faster than the birefringence at small strains[11]. The authors conjectured that the fast elastic response is caused by the electro-static interaction between the aggregates and is strongly influenced by the nature of the counter-ion. Based on the hypothetical deformation mechanism of Nafion® above, an one-dimensional rheological representation of the model is proposed, as shown in Figure 2.

36

Figure 2. One dimensional rheological representation of the constitutive model for ionomer membrane This model is adapted from a Dual Network Fluoropolymer (DNF) model, originally developed by Bergström[12]. The original model was modified in a way that the temperature and hydration dependence on mechanical properties of Nafion® membrane can be taken into account. In this study, the effect of the hygro-thermal deformation on the total deformation in the uniaxial tension tests was neglected for model simplicity. The overall mechanical behavior of the membrane against an external force can be decomposed into a viscoplastic response and viscoelastic response; the viscoelastic part is further decomposed into two networks, A and B. The viscoplastic dashpot captures the irreversible molecular chain slippage, the network A in the viscoelastic response represents the chain stretching and locking of the amorphous phase in the membrane, and the network B represents the deformation resistance due to the intermolecular and ionic interactions. The Cauchy stress is a function of the Cauchy Green deformation tensor, obtained from the Bergström and Boyce’s model[9] and the total stress is the sum of the stress contributions from the network A and B. The deformation gradient F is multiplicatively decomposed into viscoelastic and viscoplastic parts and the viscoelastic deformation gradient is further decomposed into elastic and viscous parts:

F = Fve F p (1) ve e v F =F F (2) p ve A B e v where F is the plastic deformation gradient, F = F = F , F the elastic deformation gradient and F the viscous deformation gradient tensor.

& ⋅ F . The The rate kinematics are described by the spatial velocity gradient L which is given by L = F corresponding rate kinematics can be decomposed into viscoelastic and viscoplastic contributions: -1

~ L = F& ⋅ F −1 = Lve + L p (3) ~p ~p ~ p ve ve − ve ve p − ve & where L = F F and L = F L F =D +W . ~p ~p The rate of deformation D and spin tensor W are defined as the symmetric and skew-symmetric parts of ~ L p . Likewise, the velocity gradient of viscoelastic parts Lve can be further decomposed into elastic and viscous components:

~ Lve = F& ve ⋅ F-ve = Le + Lv ~ ~ ~ where Lv = D v + W v .

(4)

The rate of spin tensors were taken to be zero, which means that the flow is irrotational[12], and the plastic and

det(F p ) = 1 . In addition, the volume change of the membrane during the deformation was assumed to be insignificant, i.e., det(F ) ≈ 1 . v

viscous deformation were assumed to be incompressible, i.e., det(F ) = 1 and

The rate of viscoplastic flow of network A and B can be constitutively described by the multiplication of a plastic deformation rate and direction tensor of deviatoric stress components:

37

~ D v = γ& v N v

(5)

~ D p = γ& p N p

(6) The details of the expressions for the eq.5 and 6 are described in the author’s recent paper[4] and the reference[12]. The plastic and viscous deformation gradient are updated by the eq.(5) and (6) at every time step. The Cauchy stress tensor for network A and B are given by the eight-chain representation[12-14].

T A = f 8ch ( F ve ) =

µA

⋅

L-1 (λve / λlock ) dev[Bve∗ ] + κ [ J ve − 1]1 -1 lock L (1 / λ )

J ve λve ⎛ µ ⎞ L-1 (λe / λlock ) e∗ e ⎟ T B = s B ⋅ f 8ch ( F e ) = s B ⋅ ⎜ B ⋅ -1 dev[B ] + κ [ J − 1 ] 1 ⎜ J e λe L (1 / λlock ) ⎟ ⎝ ⎠

(7)

(8)

In order to incorporate the hydration and temperature dependent mechanical property such as initial shear modulus µ A and µ B for ionomer membranes, an empirical relationship of elastic modulus as a function of membrane water content,

λm

and temperature,

θ ( °C ) was used. The initial shear modulus can be computed

from the product of the elastic modulus, E and scalar factors s0 B and s 0 A , which can be considered specific material parameters.

E (λ , θ ) = exp{( A1 ⋅ θ + B1 ) ⋅ λm + ( A2 ⋅ θ + B2 )} µ B = s0 B ⋅ E (λ ,θ ) = s0 B ⋅ exp{( A1 ⋅ θ + B1 ) ⋅ λm + ( A2 ⋅ θ + B2 )} µ A = s0 A ⋅ µ B

(9) (10) (11)

Table 1 Constants for elastic modulus equation

A1 0.000645

B1 -0.058673

A2 -0.014673

B2 10.534189

4. RESULTS AND DISCUSSIONS The proposed model was implemented in COMSOL Multiphysics v3.5. The built-In structural mechanics module equations were modified accordingly so that the stress components can be calculated from the constitutive equations that we have developed above. The rate equations for governing the viscous and plastic deformation were modeled in PDE general form and a time-dependent solver (Backward differentiation formula, BDF) was used to update viscous and plastic deformation gradient. In each time step, COMSOL Multiphysics solves a nonlinear system of equations using a damped Newton method [15]. Figure 3 shows the comparison of FEM and experimental results at 25°C and 50%RH for different stress levels. The constitutive model can capture the creep responses fairly well at different applied stress levels. At a stress levels over 8 MPa, the membrane can rupture in less than 2 hours. Water absorbed by Nafion can plasticize the membrane at temperature below 90°C [6]. In Figure 4, the creep responses at 90% RH and 5% RH are compared with that at 25°C and 50%RH. At 90%RH, the creep strain increased more than that at 50% RH with a similar stress level and the FEM results are in good agreement with experimental results. Creep is considered thermally activated process in most engineering materials, but hydration factor plays a significant role in creep response for Nafion membrane. Figure 5 plots the evolution of true stress components computed from the FEM calculation of creep response for the membrane at 25°C and 50%RH. Due to the geometry changes during the creep, total true stress increased with time. However, the stress component from the network B decreases with time; the network B represents molecular interactions from ionic clusters and intermolecular polymer chains. This result indicates that the interactions become weak with time because of change of microstructure of the ionomer by altering size and shape of the hydrophilic domains and the rearrangement and sliding of backbone chains.

38

Figure 3 Comparison of FEM results with experimental results of creep response for Nafion N111 membrane with respect to applied stresses at room temperature and 50% RH

Figure 4 Comparison of FEM results with experimental results of creep response for Nafion N111 membrane with respect to applied stresses at room temperature and different RH values

Figure 5 Evolution of true stress components during creep for membrane at 25°C and 50%RH

39

The relaxation behavior of membranes under fuel cell operating condition is important to determine the design parameters such as hydration/temperature range, clamping pressure and the appropriate selection of seal materials. Figure 6 shows the experimental results of relaxation tests and simulated results by FEM at different strain rates at 25°C and 50%RH. Our constitutive model can capture long-term relaxation behavior of the membrane. The overall behavior of stress-strain curves before the displacement is fixed also matches the experimental results. In the same manner as the creep response, the true stress evolution from two different molecular networks is presented in Figure 7. Due to the fixed displacement, stress from the network A doesn’t change with time, however, stress from the network B relaxed significantly with time. In the stress-strain curve of Nafion membrane, stress response at early strain typically results from deformation of ionic domain and rotation of bundle clusters[10]. Membranes in the stress relaxation test were stretched up to an engineering strain of 0.5 and the membrane is loaded in the post yield region before rupture, which occurs at strain between 1.5 and 2.5 depending on environmental conditions used. Therefore, the total true stress before the displacement is fixed is anticipated to result from mostly elastic response of the network B, rather than the network A representing stretching and locking of chain entanglements. As can noticed in Figure 7, true stress from the network A is much lower than that from the network B at a short period of time, but the stress component from the network B decays with time. Similar to the creep response, during relaxation, the membrane resistance to deformation becomes weak due to the disentangling and sliding of polymer chain and the relaxation of ionic domains.

Figure 6 Comparison of FEM results with experimental results of relaxation response for Nafion N111 membrane with respect to strain rate at room temperature and 50% RH

Figure 7 Evolution of true stress components during relaxation for membrane at 25°C and 50%RH

40

5. CONCLUSIONS The Nafion material model that we have developed was validated by comparing solutions from FEM analysis with experimental results. It was demonstrated that the model can capture quite well the time dependent creep and relaxation behavior at various relative humidities and strain rates. Total true stress component can be decomposed into two different molecular responses, namely network A and network B. During creep tests, the stress from network A which is assumed to be the response from chain entanglement keeps increase with time, however, the stress from network B decreases slightly with time. In the relaxation tests, due to the fixed strain, 0.5, stress from network A doesn’t change much with time, which is conceptually believed to result from chain entanglement. In contrary, the stress from network B decreased significantly with time. Based on the micromechanisms of Nafion deformation, bundles of backbone chain are believed to rotate and align with stretching at small strain, and some of the chains are disentangled and ionic clusters are deformed as strain increases. The stress decay with time observed from network B is believed to be related to chain sliding and stress recovery from the deformation of ionic clusters. The future work will focus on improving the constitutive model considering the volumetric expansion of membrane in the kinematic equation to better capture the nonlinear, time-dependent, hydration- and temperature-dependent behavior of ionomer membranes. REFERENCE [1] Yoon W., Huang X., Study of Polymer Electrolyte Membrane Degradation under OCV Hold using Bi-layer MEAs, J ElectrochemSoc, Reviewed and Accepted, 2009. [2] Huang X., Solasi R., Zou Y., Feshler M., Reifsnider K., Condit D., Burlatsky S., Madden T., Mechanical Endurance of Polymer Electrolyte Membrane and PEM Fuel Cell Durability, Journal of Polymer Science: Part B, 44, 2346-57, 2006. [3] Bergstrom J.S., Boyce M.C., Constitutive modeling of the large strain time-dependent behavior of elastromers, JMechPhysSolids, 46(5), 931-54, 1998. [4] Yoon W., Huang X., A Nonlinear Viscoelastic-Viscoplastic Constitutive Model for Ionomer Membranes in Polymer Electrolyte Membrane Fuel Cells(PEMFC), To be submitted, 2009. [5] Majsztrik P.W., Bocarsly A.B., Benziger J.B., An instrument for environmental control of vapor pressure and temperature for tensile creep and other mechanical property measurements, Review of Scientific Instruments, 78, 103904, 2007. [6] Majsztrik P.W., Bocarsly A.B., Benziger J.B., Viscoelastic Response of Nafion. Effects of Temperature and Hydration on Tensile Creep, Macromolecules, 41(24), 9849-62, 2008. [7] Satterfield M.B., Benziger J.B., Viscoealstic Properties of Nafion at Elevated Temperature and Humidity, J of Polymer Science:Part B, 47, 11-24, 2009. [8] Doi M., Edwards S.F. The Theory of Polymer Dynamics. New York: Oxford University Press 1986. [9] Bergstrom J.S. Large Strain Time-Dependent Behavior of Elastomeric Materials. MIT; 1999. [10] van der Heijden P.C., Rubatat L., Diat O., Orientation of Drawn Nafion at Molecular and Mesoscopic Scales, Macromolecules, 37, 5327-36, 2004. [11] van der Heijden P.C., de la Rosa A., Gebel G., Diat O., Relaxation of drawn Nafion films studied with birefringence experiments, Polym Adv Technol, 16, 102-7, 2005. [12] Bergstrom J.S., Hilbert Jr L.B., A constitutive model for predicting the large deformation thermomechanical behavior of fluoropolymers, Mechanics of Materials 37, 899-913, 2005. [13] Arruda E.M., Boyce M.C., A Three-Dimensional Constitutive Model For The Large Stretch Behavior of Rubber Elastic Materials, JMechPhysSolids, 41(2), 389-412, 1993. [14] Bergstrom J.S., Boyce M.C., Large strain time-dependent behavior of filled elastomers, Mechanics of Materials, 32, 627-44, 2000. [15] COMSOL Multiphysics User's Guide.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Evidence of piezonuclear reactions: From geological and tectonic transformations to neutron detection and measurements A. Carpinteri1, G. Lacidogna1, A. Manuello1, O. Borla1,2 1

Politecnico di Torino, Department of Structural Engineering & Geotechnics Corso Duca degli Abruzzi 24 – 10129 Torino, Italy 2

Istituto Nazionale di Fisica Nucleare, INFN sez. Torino Via Pietro Giuria 1 – 10125 Torino, Italy E-mail: [email protected]

ABSTRACT Neutron emission measurements, by means of helium-3 and neutron bubble detectors, were performed on solid specimens during three different kinds of mechanical tests: compression tests under displacement control, under cyclic loading, and by ultrasonic vibration. The material used for the tests was Green Luserna granite. Since the analyzed material contains iron, our conjecture was that piezonuclear fission reactions involving fission of iron into aluminum, and of iron into magnesium and silicon, should have occurred during compression damage and failure. It is also interesting to emphasize that the present natural abundances of aluminum (∼8%), and silicon (28%) and scarcity of iron (∼4%) in the continental Earth’s crust should be possibly due to the piezonuclear fission reactions considered above. INTRODUCTION We deal with a new topic in the scientific literature: piezonuclear neutron emissions from very brittle rock specimens in compression. In the scientific community some studies have been already conducted on the different forms of energy emitted during the failure of brittle materials. They are based on the signals captured by the acoustic emission measurement systems, or on the detection of the electromagnetic charge. But only very recently piezonuclear neutron emissions from very brittle rock specimens in compression have been discovered [1-3]. In this paper, the authors analyse this phenomenon from an experimental point of view. In particular, we present new experiments, by means of helium-3 neutron detectors and bubble type BD thermodynamic neutron detectors, performed on brittle rock test specimens. We carried out three different kinds of mechanical tests: compression tests under displacement control, under cyclic loading, and by ultrasonic solicitations. The material used for the compression tests was non-radioactive Green Luserna granite, with different specimen size and shape and consequently with different brittleness numbers. The compression tests were performed at the Fracture Mechanics Laboratory of the Politecnico of Torino, while the ultrasonic test at the Medical and Environmental Physics Laboratory of the University of Torino. For specimens of larger dimensions, neutron emissions, detected by helium-3, were found to be of about one order of magnitude higher than the ordinary natural background level at the time of the catastrophic failure. As regards test specimens with more ductile behaviour, neutron emissions significantly higher than the background level were found. This piezonuclear reactions are due to the different modalities of energy release during the tests. For specimens with sufficiently large size and slenderness, relatively large energy release is expected, and hence a higher probability of neutron emissions at the time of failure. Furthermore, during compression tests under cyclic loading, an equivalent neutron dose, analysed by neutron bubble detectors, about two times higher than the ordinary background level was found at the end of the test. Finally, by using an ultrasonic horn suitably joined with the specimen, an ultrasonic test was carried out on a Green Luserna granite specimen in order to produce continuing vibration at 20 kHz. Three hours after the start of the test, an equivalent neutron dose of about three times higher than the background level was found. Since the analyzed material contains iron, our conjecture was that piezonuclear fission reactions involving fission of iron into aluminum, or into magnesium and silicon, should have occurred during compression on the tested specimens. The present natural abundances of aluminum (∼8%), and silicon (28%) and scarcity of T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_6, © The Society for Experimental Mechanics, Inc. 2011

41

42 iron (∼4%) in the continental Earth’s crust are possibly due to the piezonuclear fission reactions considered above. This reaction would be activated where the environment conditions (pressure and temperature) are particularly severe, and mechanical phenomena of fracture, crushing, fragmentation, comminution, erosion, friction, etc., may occur. If we consider the evolution of the percentages of the most abundant elements in the Earth crust during the last 4.5 billion years, we realize that iron and nickel have drastically diminished, whereas aluminum, silicon and magnesium have as much increased. It is also interesting to realize that such increases have developed mainly in the tectonic regions, where frictional phenomena between the continental plates occurred [1-3]. MATERIAL AND METHODS He proportional counter 3 3 The He detector used in the tests is a He type (Xeram, France) with electronics of preamplification, amplification, and discrimination directly connected to the detector tube. The detector is powered with high voltage power supply (about 1.3 kV) via NIM (Nuclear Instrument Module) module. The logic output producing the TTL (through the lens) pulses is connected to a NIM counter. The logic output of the detector is enabled for analog signals exceeding 300 mV. This discrimination threshold is a consequence of the sensitivity of the helium-3 detector to the gamma rays ensuing neutron emission in ordinary nuclear processes. This value has been determined by measuring the analog signal of the detector by means of a Co-60 gamma source. The detector is also calibrated for the measurement of thermal neutrons; its sensitivity 2 is 65 cps/nthermal, i.e., the flux of thermal neutrons was 1 thermal neutron/s cm , corresponding to a count rate of 65 cps. 3

Neutron Bubble detectors A set of passive neutron detectors insensitive to electromagnetic noise and with zero gamma sensitivity was used. The dosimeters, based on superheated bubble detectors (BTI, Ontario, Canada) (BUBBLE TECHNOLOGY INDUSTRIES (1992)) [4], are calibrated at the factory against an AmBe (AmericiumBeryllium) source in terms of NCRP38 [5]. Bubble detectors are the most sensitive, accurate neutron dosimeters available that provide instant visible detection and measurement of neutron dose. Each detector is composed of a polycarbonate vial filled with elastic tissue equivalent polymer, in which droplets of a superheated gas (freon) are dispersed. When a neutron strikes a droplet , the droplet immediately vaporizes, forming a visible gas bubble trapped in the gel. The number of droplets provides a direct measurement of the equivalent neutron dose with an efficiency of about 20%. These detectors are suitable for neutron dose measurements, in the energy range of thermal neutrons (E = 0,025eV, BDT type) and fast neutrons (E > 100 keV, BD-PND type). The servo-hydraulic press The servo-controlled press employed works by means of a digital type electronic control unit. The management software is TESTXPERTII by Zwick/Roel (Zwick/Roel Group, Ulm, Germany), while the mechanical parts are manufactured by Baldwin (Instron Industrial Products Group, Grove City, PA, USA). The force applied is determined by measuring the pressure in the loading cylinder by means of a transducer. The margin of error in the determination of the force is 1%, which makes it a class 1 mechanical press. The platen connected to the stroke of the press is in contact with the test specimen and controlled by means of a wire-type potentiometric displacement transducer. Sonotrode Ultrasonic oscillation was generated by an high intensity ultrasonic horn (Bandelin HD 2200) working at 20kHz. The device guarantees a constant amplitude (ranging from 10% to 100%) independently of changing conditions within the sample. The apparatus consists of a generator that converts electrical energy to 20 kHz ultrasound, and of a transducer that switches this energy into mechanical longitudinal vibration of the same frequency. EXPERIMENTAL SET-UP Compression tests under displacement control Neutron emissions were measured on nine Green Luserna granite cylindrical specimens, of different size and shape (Table 1), denoted with P1, P2,…, P9. The specimens were arranged with the two smaller surfaces in contact with the press platens, without coupling materials in-between, according to the testing modalities known as “test by means of rigid platens with friction”. The tests were performed under displacement control, with the planned displacement velocities ranging from 0.001 to 0.01 mm/s. The helium-3 neutron detector was switched on at least one hour before the beginning of each compression test, in order to reach the thermal equilibrium of electronics, and to make sure that the behaviour of the devices were stable with respect to intrinsic thermal effects. The detector was placed in front of the test

43 specimen at a distance of 20 cm and it was enclosed in a polystyrene case of 10 cm of thickness in order to avoid “spurious” signals coming from impact and vibration. A relative measurement of natural neutron background was performed in order to assess the average background affecting data acquisition in experimental room condition. The helium-3 device was positioned in the same condition of the experimental set up and the background measures were performed fixing at 60 s the acquisition time, during a preliminary period of more than three hours, for a total number of 200 counts. -4 -4 -2 -1 The average measured background level is ranging from (3.17±0.32)·10 to (4.74±0.46)·10 nthermalcm s (see Table 2). Geometry of the specimen

Granite Specimen

D (mm)

H (mm)

λ=H/D

P1 P2 P3 P4 P5 P6 P7 P8 P9

28 28 28 53 53 53 112 112 112

14 28 56 25 50 101 60 112 224

0.5 1 2 0.5 1 2 0.5 1 2

Displacement velocity (mm/s) 0.001 0.001 0.001 0.001 0.001 0.001 0.01 0.01 0.01

Peak Load (kN)

Time at the peak load (s)

52.19 33.46 41.28 129.00 139.10 206.50 1099.30 1077.10 897.80

735.0 1239.0 1089.0 960.0 2460.0 1180.0 231.3 263.5 218.6

Table 1: Characteristics of compression tests under displacement control on Green Luserna granite specimens. Compression test under cyclic loading A Green Luserna granite specimen (D=53mm, H=53mm, λ=1) was used. The cyclic loading was programmed at a frequency of 2 Hz and with a load excursion from a minimum load of 10 kN to a maximum of 60 kN. With respect to the tests performed under displacement control, neutron emissions from compression test under cyclic loading were performed by using neutron bubble detectors. Due to their isotropic angular response, three BDT and three BD-PND detectors were positioned at a distance of about 5 cm, all around the specimen. The detectors were previously activated, unscrewing the protection cap, in order to reach the suitable thermal equilibrium, and they were kept active for all the test duration. Furthermore, a BDT and a BD-PND detector were used as background control during the test. Ultrasonic test A Green Luserna granite specimen (D=53mm, H=100mm, λ=2) was connected to the ultrasonic horn by a glued screw inserted in a 5 mm deep hole. This kind of connection was made in order to achieve a resonance condition, considering the speed of sound in Luserna stone, and the length of the specimen. Ultrasonic irradiation of the specimen was carried out for 3 hours. After the switching on of the transducer, 10% of the maximum power was reached in 20 min. Successively, the transducer power increased to 20% after one hour, and next reached a maximum level of about 30% after 2 hours. Then, the transducer worked in the same power condition up to the end of the test. EXPERIMENTAL RESULTS Compression tests under displacement control The helium-3 device was switched on at least one hour before the beginning of each test, in order to reach a suitable electronic thermal equilibrium. Additional background measurements were repeated before each test, fixing an acquisition time of 60 s in order to check possible variation of natural background. Neutron measurements of specimen P2, P3, P4, P7 yielded values comparable with the ordinary natural background, while in specimens P1 and P5 the experimental data exceeded the background value by about four times. Instead, for specimen P6, P8 and P9, the neutron emissions achieved values higher than one order of magnitude with respect to ordinary background. In Fig. 1, for specimens P6, P8, and P9, the load vs. time diagram, and the neutron count rate evolution are shown. In Table 2, experimental data concerning compression tests on the nine Green Luserna granite specimens are synthesized. The experimental results seem to demonstrate that neutron emissions follow an anisotropical distribution with an impulsive release from a specific zone of the specimen. Moreover, it is a matter of fact that the detected neutron flux, and consequently neutron dose, are inversely proportional to the square of the distance from the source. For these reasons, helium-3 device could have underestimated neutron flux intensity, in any specimen. A possible solution to avoid underestimated data acquisition is an experimental measurement by using more than one helium-3 detector and more bubble dosimeters placed around the test specimens.

250

35

Cps - Green Luserna granite (D=53mm, H=101mm) -2 Average Neutron Background (4,74±0,46)x10 cps

30

Load (kN)

200

25

20

-2

Load (kN)

Specimen P6 150

Cps (10 )

44

15

100

(a)

10 50 5

0 0

250

500

750

1000

1250

1500

0 1750

Time (s)

1400

1200

45 40

Specimen P8

1000

35 30 -2

800

Cps (10 )

Load (kN)

50

Cps - Green Luserna granite (D=112mm, H=112mm) -2 Average Neutron Background (4,20±0,80)x10 cps Load (kN)

25 600

20

(b)

15

400

10 200

5

0 0

1200

100

150

200

Time (s)

250

300

350

45

Load (kN)

40

Specimen P9

800

0 400

50

Cps - Green Luserna granite (D=112mm, H=224mm) -2 Average Neutron Background (4,20±0,80)x10 cps

1000

35 30 -2

Cps (10 )

Load (kN)

50

600

25 20

400

(c)

15 10

200

5 0 0

50

100

150

200

Time (s)

250

300

350

0 400

Figure 1: Specimens P6, P8, P9. Load vs. time diagrams, and neutron emissions count rate. Granite Specimen P1 P2 P3 P4 P5 P6 P7 P8 P9

Time at neutron emission (s) 570 ------2460 1440 --270 225

Count rate at the neutron emission -2 (10 cps) 8.33±3.73 background background background 11.67±4.08 25.00±6.01 background 30.00±11.10 30.00±10.00

Corresponding thermal neutron flux -4 -2 -1 (10 cm s ) 12.81±5.74 background background background 17.95±6.28 38.46±9.25 background 46.15±17.08 46.15±15.38

Average neutron background -4 -2 -1 (10 cm s ) 3.17±0.32 3.17±0.32 3.17±0.32 3.83±0.37 3.84±0.37 4.74±0.46 4.20±0.80 4.20±0.80 4.20±0.80

Table 2: Compression tests under displacement control. Neutron emissions experimental data on Green Luserna granite specimens.

45 Compression test under cyclic loading Droplets counting was performed every 12 hours and the equivalent neutron dose was calculated. In the same way, the natural background was estimated by means of the two bubble dosimeters used for assessment. The ordinary background was found to be (13.98±2.76) nSv/h. In Fig. 2 neutron equivalent dose variation, evaluated during the cyclic compression test, is reported. An increment of more than twice with respect to the background level was detected at specimen failure. No significant variations in neutron emissions were observed before the failure. The equivalent neutron dose, at the end of the test, was (28.74±5.75) nSv/h. Compression test under cyclic loading on Green Luserna granite specimen

40

Equivalent Neutron Dose (nSv/h)

35

Equivalent Neutron Dose Average Neutron Background (13.98±2.76) nSv/h

30 25 20 15 10 5 0 0

12

24

36

48

60

72

84

96

Time (hrs)

Figure 2: Compression test under cyclic loading. Equivalent neutron dose variation on Green Luserna granite specimen. Ultrasonic test The ultrasonic test on Green Luserna granite specimen (D=53mm, H=100mm, λ=2) was carried out at the Medical and Environmental Physics Laboratory of Experimental Physics Department of the University of Torino. A relative natural background measurement was performed by means of the helium-3 detector for -3 more than 6 hours. The average natural background was of (6.50±0.85)·10 cps, for a corresponding thermal -4 -2 -1 neutron flux of (1.00±0.13)·10 nthermalcm s . This natural background level, lower than the one calculated during the compression tests at the Fracture Mechanics Laboratory of the Politecnico of Torino, is in agreement with the location of the experimental Physics Laboratory, which is three floors below the ground level. During the ultrasonic test, the specimen temperature was monitored by using a multimeter/thermometer (Tektronix mod. S3910). The temperature reached 50°C after 20 min, and then increased up to a maximum level of 100°C at the end of the ultrasonic test. I n Fig. 3, the neutron emissions detected are compared with the transducer power trend and the specimen temperature. A significant increment in neutron activity after 130 min from the beginning of the test was measured. At this time, the transducer power reached 30% of the maximum, with a specimen temperature of about 90°C. Some neutron variations were detected during the first hour of the test, but they may be due to ordinary fluctuations of natural background. At the switching off of the sonotrode, the neutron activity decreased to the typical background value. Ultrasonic Test on Green Luserna granite specimen

40

Cps - Luserna Granite (D=53mm, H=100m) -3 Average Neutron Background (6.50±0.85)x10 cps Transducer Power Specimen Temperature

16 14

35

25

-3

Cps (10 )

12 10

20

8

15

6 10 4 5

2 Sonotrode switch on

0 20

40

60

80

60

40

20

Sonotrode switch off

0 0

100

30

Temperature (°C)

18

120

Transducer Power (%)

20

80

100

120

140

160

180

0

200

Time (min)

Figure 3: Ultrasonic test. Neutron emissions compared with the specimen temperature, and with the transducer power trend.

46 PIEZONUCLEAR REACTIONS: FROM THE LABORATORY TO THE EARTH SCALE From the results shown in the previous sections and the experimental evidence reported in recent papers [13], it can be clearly seen that piezonuclear reactions are possible in inert non-radioactive solids. To this purpose, an important aspect that should be taken into account is the composition of the materials in which the piezonuclear reactions have occurred. Green Luserna granite contains a considerable amount of iron oxides (∼3% of Fe2O3) [6], and the iron content of the green Luserna granite used in the piezonuclear experiments may contribute to the phenomenon in question. After the fracture experiments on green Luserna Granite specimens [1-3], the analysis of the fracture surfaces, conducted by energy dispersive X-ray spectroscopy, have shown a considerable reduction in the iron content (–25%). This iron decrease is counterbalanced by an increase in aluminum, silicon, and magnesium. In particular, the increase in aluminum content corresponds to the eighty-five percent of the iron decrease. Therefore, the piezonuclear fission reactions: 27 (1) Fe56 26 → 2Al13 + 2 neutrons 24 28 Fe56 26 → Mg12 + Si14 + 4 neutrons

(2)

have occurred [1-3]. Considering that granite is a common and widely occurring type of intrusive, Sialic, igneous rock and that it is characterized by an extensive concentration in the rocks that make up the Earth’s crust (∼60% of the Earth’s crust), the piezonuclear fission reactions expressed above can be generalized from the laboratory to the Earth’s crust scale, where mechanical phenomena of brittle fracture, due to fault collision and subduction, take place continuously in most seismic areas. This hypothesis seems to find surprising evidence and confirmation from both the geomechanical and the geochemical points of view. The neutron emissions involved in piezonuclear reactions can be detected not only in laboratory experiments, as shown in this paper, and in [1-3], but also at the Earth’s crust scale. Recent neutron emission detections by Kuzhevskij et al. [7,8] have led to consider also the Earth’s crust, in addition to cosmic rays, as being a relevant source of neutron flux variations. Neutron emissions measured near the Earth’s surface exceeded the neutron background by about one order of magnitude in correspondence to seismic activity and rather appreciable earthquakes [9]. This relationship between the processes in the Earth’s crust and neutron flux variations has allowed increasing tectonic activity to be detected and methods for short-term prediction and monitoring of earthquakes to be developed [7,8]. Neutron flux variations, in correspondence to seismic activity, may be evidence of changes in the chemical composition of the crust, as a result of piezonuclear reactions. The present natural abundances of aluminum (∼8%), and silicon (28%) and scarcity of iron (∼4%) in the continental Earth’s crust are possibly due to the piezonuclear fission reactions considered above.

HETEROGENEITY IN THE COMPOSITION OF THE EARTH’S CRUST: Fe AND Al RESERVOIR LOCATIONS The location of Al and Fe mineral reservoirs seems to be closely connected to the geological periods when different continental zones were formed [10-16]. This fact would seem to suggest that our planet has undergone a continuous evolution from the most ancient geological regions, which currently reflect the continental cores that are rich in Fe reservoirs, to more recent or contemporary areas of the Earth’s crust where the concentrations of Si and Al oxides present very high mass percentages [10]. The main iron reservoir locations (Magnetite and Hematite mines) are reported in Fig. 4(a). The main concentrations of Aloxides and rocky andesitic formations (the Rocky Mountains and the Andes, with a strong concentration of Al2O3 minerals) are shown in Fig. 4(b) together with the most important subduction lines, plate tectonic trenches and rifts [10,14]. The geographical locations of main bauxite mines show that the largest concentrations of Al reservoirs can be found in correspondence to the most seismic areas of the Earth (Fig.4(b)). The main iron mines are instead exclusively located in the oldest and interior parts of continents (formed through the eruptive activity of the proto-Earth), in geographic areas with a reduced seismic risk and always far from the main fault lines. From this point of view, the close correlation between bauxite and andesitic reservoirs and the subduction and most seismic areas of the Earth’s crust provides very impressive evidence of piezonuclear effects at the planetary scale. GEOCHEMICAL EVIDENCE OF PIEZONUCLEAR REACTIONS IN THE EVOLUTION OF THE EARTH’S CRUST From 4.0 to 2.0 Gyrs ago, Fe could be considered one of the most common bio-essential elements required for the metabolic action of all living organisms. Today, the deficiency of this nutrient suggests it as a limiting factor for the development of marine phytoplankton and life on Earth [13]. Elements such as Fe and Ni in the Earth’s protocrust had higher concentrations in the Hadean (4.5−3.8 Gyr ago) and Archean (3.8−2.5 Gyr ago) periods compared to the present values. The Si and Al concentrations instead were lower than those of today [10-12]. In Fig. 5, the evolution in mass percentage concentration of Si, Al, Fe and Ni in the Earth protocrust and crust over the last 4.5 Billion years is reported. A clear transition

47 from a more basaltic condition (high concentrations of Fe and Ni) to a Sialic one (high concentrations of Al and Si) can be observed during the life time of our planet. The most abrupt changes in element concentrations shown in Fig. 5 appear to be intimately connected to the tectonic activity of the Earth. In particular, the abrupt transitions of 2.5 Gyrs ago coincide with the period of the Earth’s largest tectonic activity [11,12].

(a)

(b)

Figure 4: (a) Locations of the largest iron mines in the world [17-20]. Iron ore reservoirs (Magnetite and Hematite mines) are located in geographic areas with reduced seismic risks and always far from fault lines. (b) The largest aluminum (bauxite) reservoirs are reported together with the main Andesitic formations and most important subduction lines and plate tectonic trenches [10].

Figure 5: Evolution in mass percentage concentration of Si, Al, Fe and Ni in the Earth crust during the last 4.5 Billion years (age of the planet Earth) [10-12,15, 21-27].

CONCLUSIONS Neutron emission measurements were performed on Green Luserna granite specimens during mechanical tests. From these experiments, it can be clearly seen that piezonuclear reactions giving rise to neutron emissions are possible in inert non-radioactive solids under loading. In particular, during compression tests of specimens with sufficiently large size, the neutron flux was found to be of about one order of magnitude higher than the background level at the time of catastrophic failure. For test specimens with more ductile behaviour, neutron emissions significantly higher than the background were found. Neutron detection is also confirmed in compression test under cyclic loading and during ultrasonic vibration. Our conjecture is that piezonuclear fission reactions involving fission of iron into aluminum, or into magnesium and silicon, should have occurred during compression on the tested specimens. This hypothesis seems to find surprising evidence and confirmation at the Earth crust scale from both geomechanical and geochemical points of view.

48 ACKNOWLEDGEMENTS The financial support provided by the Regione Piemonte (Italy) RE-FRESCOS Project, is gratefully acknowledged. Special thanks are due to Daniele Madonna Ripa and Alessandro Troia from the National Research Institute of Metrology – INRiM, for their indispensable assistance during the ultrasonic tests. REFERENCES [1] Cardone, F., Carpinteri, A., Lacidogna, G. Piezonuclear neutrons from fracturing of inert solids. Physics Letters A. 373. 4158-4163. (2009). [2] Carpinteri, A., Cardone, F., Lacidogna G. Piezonuclear neutrons from brittle fracture: Early results of mechanical compression tests. Strain. 45. 332-339. (2009). [3] Carpinteri, A., Cardone, F., Lacidogna G. Energy emissions from failure phenomena: Mechanical, electromagnetic, nuclear. Experimental Mechanics. doi: 10.1007/s11340-009-9325-7. [4] Bubble Technology Industries, Instruction manual for the Bubble detector, Chalk River, Ontario, Canada (1992). [5] National Council on Radiation Protection and Measurements, Protection Against Neutron Radiation, NCRP Report 38 (1971). [6] Vola, G., Marchi, M. Mineralogical and petrographic quantitative analysis of a recycled aggregate from quarry wastes. The Luserna stone case-study. Proc of the 12th Euroseminar on Microscopy Applied to Building Materials, 15-19 September 2009, Dortmund, Germany. (2009). [7] Kuzhevskij, B. M., Yu. Nechaev, O., Sigaeva, E. A., Zakharov, V. A. Neutron flux variations near the Earth’s crust. A possible tectonic activity detection. Natural Hazards and Earth System Sciences. 3. 637–645. (2003). [8] Kuzhevskij, B. M., Yu. Nechaev, O., Sigaeva, E. A. Distribution of neutrons near the Earth’s surface. Natural Hazards and Earth System Sciences. 3. 255–262. (2003) [9] Volodichev, N.N., Kuzhevskij, B.M., Nechaev, O.Yu., Panasyuk, M.I., Podorolsky, A.N.,. Shavrin, P.I. Sun-Moon-Earth connections: The neutron intensity splashes and seismic activity. Astron. Vestnik. 34(2). 188–190. (2000). [10] Favero, G., and Jobstraibizer, P. The distribution of aluminum in the Earth: from cosmogenesis to Sial evolution. Coordination Chemistry Reviews. 149. 467–400. (1996). [11] Taylor, S. R. and McLennan, S. M. The geochemical evolution of the continental crust. Reviews of Geophysics. 33(2). 241–265. (1995). [12] Taylor, S.R. and McLennan, S. M. Planetary Crusts: Their Composition, Origin and Evolution, Cambridge University Press, Cambridge. (2009). [13] Anbar, A. D. Elements and evolution. Science. 322. 1481–1482. (2008). [14] Lunine E, J. I. Earth: Evolution of a Habitable World. Cambridge University Press, Cambridge, New York, Melbourne. (1998). [15] Hazen et al. Mineral evolution, American Mineralogist. 93. 1693–1720. (2008). [16] Condie, K. C. Plate Tectonics and crustal evolution. Pergamon Press, New York, Toronto, Oxford, Sydney, Braunshweig, Paris. (1976). [17] Roy, I., Sarkar, B. C., Chattopadhyay, A. MINFO-a prototype mineral information database for iron ore resourcers of India. Computers and Geosciences. 27. 357–361. (2001) [18] World Iron Ore producers. Available at http://www.mapsofworld.com/minerals/world-iron-oreproducers.html; last accessed October 2009. [19] World Mineral Resources Map. Available at http://www.mapsofworld.com/world-mineral-map.htm; last accessed October 2009. [20] Key Iron Deposits of the World. Available at http://www.portergeo.com.au/tours/iron2002/iron2002depm2b.asp; last accessed October 2009. [21] Konhauser, K.O. et al. Oceanic nickel depletion and a methanogen famine before the Great Oxidation Event. Nature. 458. 750–754. (2009). [22] Saito, M. A. Less nickel for more oxygen. Nature. 458. 714–715. (2009). [23] Rudnick, R. L. and Fountain, D. M. Nature and composition of the continental crust: A lower crustal perspective. Reviews of Geophysics. 33(3). 267–309. (1995). [24] Egami, F. Minor elements and evolution. Journal of Molecular Evolution. 4(2). 113–120. (1975). [25] Natl. Academy of Sciences. Medical and Biological Effects of Environmental Pollutants: Nickel. Proc. Natl Acad Sci. Washington, DC. (1975). [26] Doglioni, C. Interno della Terra, Treccani, Enciclopedia Scienza e Tecnica, 595–605. (2007). [27] Foing, B. Earth’s childhood attic. Astrobiological Magazine: Retrospection (on-line) February 23. (2005).

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Energy dispersive X-ray spectroscopy analysis on rock samples subjected to piezonuclear tests A. Carpinteri1, A. Chiodoni2, A. Manuello1, R. Sandrone3 1

Politecnico di Torino, Department of Structural Engineering & Geotechnics, Corso Duca degli Abruzzi 24 – 10129 Torino, Italy E-mail: [email protected]

2

Politecnico di Torino, Italian Institut of Tecnology (IIT) Center for Space Human Robotics, Corso Trento 21 – 10129 Torino, Italy 3

Politecnico di Torino, Department of Land, Environment and Geo-Engineering, Corso Duca degli Abruzzi 24 – 10129 Torino, Italy

ABSTRACT In the present paper, Energy Dispersive X-ray Spectroscopy (EDS) was performed on different samples of external or fracture surfaces coming from specimens used in piezonuclear tests [1,2]. For each sample, different measurements of the same crystalline phases (phengite and biotite) were performed in order to get averaged information of the chemical composition and to detect possible piezonuclear transmutations from iron to lighter elements. The results of EDS analyses show that, in the fracture surface samples, a considerable reduction in the iron content (∼25%) seems to be counterbalanced by an increase in Al, Si, and Mg concentrations. INTRODUCTION It has been shown that pressure, exerted on radioactive or inert media, can generate nuclear reactions and reproducible neutron emissions. In particular, low energy nuclear reactions and neutron emissions have been verified in pressurized deuterium gas by Arata et al. [3,4], and in radioactive deuterium-containing liquids during ultrasounds and cavitation by Taleyarkhan et al. [5]. The experiments recently proposed by Carpinteri et al. [1] and by Cardone et al. [2] follow a different path from those of other research teams and represent the first evidence of piezonuclear reactions and neutron emissions in inert, stable and nonradioactive solids under compression, as well as in nonradioactive liquids during ultrasound cavitation [6,7]. The analyses of the present paper are in strict connection with the results of piezonuclear tests presented by Carpinteri et al. [1] and by Cardone et al. [2]. Neutron emission measurements, by means of helium-3 neutron detectors, have recently been performed on solid test specimens during crushing failure [1,2]. Neutron emissions from “Luserna stone” test specimens were found to be of about one order of magnitude higher than the natural background level at the time of failure. These neutron emissions should be caused by nucleolysis or piezonuclear “fissions” occurred in the granitic gneiss samples, transforming heavier (Fe) into lighter (Mg, Al, Si) atoms in correspondence to brittle failure of the specimens. These reactions —less infrequent than we could think— would be activated where the environment conditions (pressure and temperature) are particularly severe, and mechanical phenomena of fracture, crushing, fragmentation, comminution, erosion, friction, etc., may occur. [1,2]. In the present paper, in order to correlate the neutron emission from the Luserna stone with the variations in rock composition due to brittle failure of the granitic gneiss specimens, energy dispersive X-ray spectroscopy (EDS) was performed on different samples of external or fracture surfaces belonging to the same two specimens used in the piezonuclear tests by Carpinteri et al. [1,2]. These analyses lead to get averaged information of the mineral and chemical composition and to detect possible piezonuclear transmutations from iron to lighter elements. The quantitative elemental analyses were performed by a ZEISS Supra 40 Field Emission Scanning Electron Microscope (FESEM) equipped with an Oxford X-rays microanalysis. The samples were carefully chosen to investigate and compare the same crystalline phases both before and after the crushing failure. In particular, two

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_7, © The Society for Experimental Mechanics, Inc. 2011

49

50 crystalline phases, phengite and biotite, that are quite common in the Luserna stone (20% and 2%, respectively), are considered owing to the high iron concentration in their chemical compositions [8]. GRANITIC GNEISS COMPOSITION AND PIEZONUCLEAR TEST Luserna stone is a leucogranitic orthogneiss, probably from the Lower Permian Age, that outcrops in the LusernaInfernotto basin (Cottian Alps, Piedmont) at the border between the Turin and Cuneo provinces (North-western Italy) [9]. Characterized by a micro “Augen” texture, it is grey-greenish or locally pale blue in colour. Geologically, Luserna stone pertains to the Dora-Maira Massif [8,10], that represents a part of the ancient European margin annexed to the Cottian Alps during Alpine orogenesis. From a petrographic point of view, it is the metamorphic result of a late-Ercinian leucogranitic rock transformation [8,11] The Luserna Stone has a sub-horizontal attitude, with a marked fine-grained foliation that is mostly associated with visible lineation. The mineralogical composition includes K-feldspar (10-25 Wt. %), quartz (30-40 Wt. %), albite (15-25 Wt. %) and phengite (10-20 Wt. %); subordinated biotite, chlorite, zoisite and/or clinozoisite/epidote (less then 5%). In addition to common accessory phases (ores, titanite, apatite and zircon), tourmaline, carbonates, rare axinite and frequent fluorite are present [8,12]. In the fundamental papers of Carpinteri et al. [1] and Cardone et al. [2], the materials selected for the compression tests were Carrara marble and Luserna stone. This choice was prompted by the consideration that, test specimen dimensions being the same, different brittleness numbers [13] would cause catastrophic failure in granite, not in marble. These early results on piezonuclear reactions from brittle fractures of Luserna stone specimens may be accounted for by the catastrophic nature of the failure [13-15]. In this case, another important aspect that should be taken into account is also the composition of the materials in which the piezonuclear reactions have occurred. The marble used in the piezonuclear tests [1,2] contains only iron impurities (not more than 0.07% of Fe 2O3 as total Fe), Luserna stone instead contains a considerable amount of iron oxides (∼3% of Fe2O3 as total Fe). The iron content of the Luserna granite used in the piezonuclear experiments could contribute to the phenomenon in question, in analogy with the case of piezonuclear reactions in liquids [6,7]. Piezonuclear reactions with neutron emissions have in fact been obtained in liquids containing iron chloride or iron nitrate and subjected to ultrasounds and cavitation [6,7]. In these experiments on liquid solutions, aluminum atoms appeared at the end in a final quantity as large as about seven times the small initial quantity [16]. For this reason, the iron content in the Luserna stone must be considered, together with the brittle nature of failures, as a very important factor for the occurrence of piezonuclear reactions and neutron flux emissions. In this context, the compositional variations found between external and fracture surfaces may be particularly important to assess “nucleolysis” or piezonuclear “fissions”, occurred in the tested material, transforming heavier (Fe) into lighter (Mg, Al, Si) atoms. In consequence of Luserna stone being a very heterogeneous rock, and in order to assess mass percentage variations in chemical elements such as Fe, Al, Si and Mg, the EDS analyses have been focused on two crystalline phases: phengite an biotite. These two minerals of granitic gneiss, that are quite common in the Luserna stone (20% and 2%, respectively), show a mineral chemistry in which the iron content is largely diffused (see Figs. 1a and 1b).

Figure 1: (a) The chemical composition of phengite includes: SiO2 (∼56%), Al2O3 (∼24%), Fe2O3 and FeO (∼8%) MgO (∼1.5%), Na2O (∼0.2%) and K2O (∼10%). (b) The chemical composition of biotite includes: SiO2 (∼35%), Al2O3 (∼16%), Fe2O3 and FeO (∼33%), MgO (∼3.5%), TiO2 (∼1.5%), and K2O (∼10%).

51 X-RAY SPECTROSCOPY ANALYSIS In order to correlate the composition of Luserna rock with the neutron emission from the fracture surfaces, semiquantitative compositional analyses were carried out by means of a ZEISS Supra 40 Field Emission Scanning Electron Microscope equipped with an Oxford INCA Energy Dispersive X- Ray detector (Si(Li)) with a resolution of 133 eV @ (MnKα). Each sample was carefully chosen and preliminary characterised by using an optical microscope in order to precisely find the minerals of interest, i.e. biotite and phengite. The samples were covered by a 5 nm thick Cr layer, in order to improve their conductivity and properly distinguish the phases of interest. The energy used for the analyses was 20 KeV. Using a nanometric electron beam probe, the measurements were carried out on different zones of the samples. This kind of analysis involves tenths of cubic microns for each acquisition point in the investigated sample. Two different kinds of samples were examined: (i) polished thin sections, prepared with a standard petrographic sample preparation, covered by Cr, for what concerns the outer surface of the Luserna stone; (ii) small portions of fracture surfaces without any kind of preparation, apart from the Cr covering, in order to avoid any kind of modification of the composition due to the slice preparation, for what concerns the fracture surface. Semi-quantitative standardless analysis was performed on the collected spectra, fixing the stoichiometry of the oxides, in order to correlate the oxides content with the specific crystalline phase. The Cr lines were excluded from the semi-quantitative evaluation. In Fig. 2a, two polished thin sections obtained from the external surfaces of an integer and uncracked portion of the specimens are shown. The polished thin sections present a rectangular geometry (45X27 mm) and are 30 µm thick. In Fig. 2b, two portions of fracture surfaces taken from the tested specimens are shown. For the EDS analyses, several phengite and biotite sites were localized on the surface of the polished thin sections and on the fracture surfaces. Sixty measurements of phengite crystalline phase and thirty of biotite were selected and analysed. In Figs. 3a and 3b, two electron microscope images of phengite and biotite sites, the first in the external sample (polished thin section 1) and the second on the fracture surface (fracture surface 2), are shown. In Tab. 1, six representative analyses of chemical composition and atomic proportions of phengite (3 for external samples and 3 for fracture surfaces) are shown. In the same Table, 4 representative analyses (2 for external samples and 2 for fracture surfaces) of biotite are also reported.

Figure 2: (a) Polished thin sections obtained by the external surface of an integer and non fractured portion of the testes specimens [1,2]. (b) Fracture surface taken by the tested specimens [1,2].

(a)

(b)

Figure 3: FESEM images of phengite and biotite in the case of external (a) and fractured sample.

52 Tab. 1: Representative chemical analyses and atomic proportions (basis 22 oxygens) for phengite and biotite. Six representative analyses in the case of phengite and four in the case of biotite are choosen.

EDS RESULTS OF PHENGITE In Fig. 4a and b, the results for the Fe concentrations obtained from 60 measurements of phengite crystalline phase are shown. Thirty of these measurements have been selected on the polished thin sections as representatives of the external surface samples, whereas the other thirty measurements are selected on fracture surfaces. It can be observed that the distribution of Fe concentrations for the external surfaces, represented in the graph by squares, show an average value of the distribution (calculated as the arithmetic mean value) equal to 6.20%. In the same graph the distribution of Fe concentrations on the fracture samples (indicated by triangles) shows significant variations. It can be seen that the mean value of the distribution of measurements performed on fracture surfaces is equal to 4.0% and it is considerably lower than the mean value of external surface measurements (6.20%). It is also interesting to note that the two Fe value distributions are separated by at least two standard deviations (σ= 0.37 in the case of external surfaces and σ=0.52 in the case of fracture surfaces). The iron decrease, considering the mean values of the distributions of phengite composition, is about 2.20%. This iron content reduction corresponds to an absolute decrease of 35% with respect to the previous Fe content (6.20% in phengite) (see also Table 2). Similarly to Fig. 4a, in Fig 4b the Al mass percentage concentrations are considered in both the cases of external and fracture surfaces. For Al contents, the observed variations show a mass percentage increase approximately equal to that of Fe (compare Fig. 4a and 4b). The average increase of the distribution, corresponding to the fracture surfaces (indicated by triangles), is about 2.00% of the phengite composition. The average value of Al concentrations changes from 12.50% on the external surface to 14.50% on the fracture surface. The absolute increase in Al content is equal to 16%. The evidence emerging from the EDS analyses, that the two values for the iron decrease (-2.20%) and for the Al increase (+2.0%) are approximately equal, is really impressive. This fact is even more evident considering the trends of the other chemical elements constituting the mineral chemistry (excluding H and O) in phengite. In Fig. 5a, b and c the Si, Mg and K concentration distributions are reported for external and fracture surfaces. In this case no appreciable variations can be recognized between the average values. These evidence of Fe and Al variations in fractured samples, coming from the same specimens used for the piezonuclear tests [1,2], together with the neutron emission measurements shown by Carpinteri et al. [1] e da Cardone et al. [2], lead to the conclusion that the piezonuclear reaction: 27 (1) Fe56 26 → 2 Al13 + 2 neutrons , should have occurred [1,2].

53

Analysis

(a)

Analysis

(b) Figure 4: Fe and Al concentrations in phengite: (a) Fe concentrations for the external surfaces (squares) and for fracture surfaces (triangles). The Fe decrease considering the two mean values of the distributions is equal to 2.20%. (b) Al concentrations for the external surfaces (indicated by squares) and for fracture surfaces (indicated by triangles). The Al increase, considering the two mean values of the distributions, is equal to 2.0%.

54

Analysis

Analysis

Analysis

Figure 5: Si, Mg, K concentrations in phengite: Si (a), Mg (b), and K (c) concentration distributions are reported for external and fracture surfaces. In this case, no appreciable variations can be recognized in fracture surfaces. EDS RESULTS OF BIOTITE In Fig. 6a-e the results for Fe, Al, Si, Mg and K concentrations measured on 30 acquisition points of biotite crystalline phase are shown. These measurements were selected on the polished thin sections as representatives of the uncracked material samples (15 measurements) and on fracture surfaces (15 measurements). It can be observed that the distribution of Fe concentrations for the external surfaces, represented in Fig. 5a by squares, shows an average value of the distribution (calculated as the arithmetic mean value) equal to 21.20%. On the other hand, considering in the same graph the distribution of Fe concentrations on fractured samples (indicated by triangles), it can be seen that the mean value of the distribution of measurements in fractured samples drops to 18.20%. In this case, the iron decrease, considering the mean values of the distributions of biotite composition, is about 3.00%. This iron content reduction (-3.00%) corresponds to an absolute decrease of 14% with respect to the previous Fe content (21.20% in biotite) (see Table 3). Similarly to Fig. 6a, in Fig. 6b the Al mass percentage concentrations are considered in both cases of external and fractured samples. For Al contents the observed variations show an average increase of about 1.50% in the phengite composition. The average value of Al concentrations changes from 8.10% in the external surface to 9.60% in the fracture surface, with an absolute increase in Al content equal to 18%. In Fig. 6c and 6d it is shown that, in the case of biotite, also Si and Mg contents present considerable variations, whereas only K does not show appreciable variations (Fig. 6e). Fig. 6c shows that the mass percentage concentration of Si changes from a mean value of 18.4% (external surface) to a mean value of 19.60% (fracture surface) with an increase of 1.20%. Similarly in Fig. 6d the Mg concentration distributions show that Mg content mean value changes from 1.50% (external surface) to a mean value of 2.20% (fracture surface). Therefore, the iron decrease (-3.00%) in biotite is counterbalanced by an increase in aluminum (+1.50%), silicon (+1.20%), and magnesium (+0.70%). Considering these evidence for the content variations in Fe, Al, Si and Mg for the biotite, in analogy to the results of EDS analyses discussed in the previous section, it is possible to conjecture that another piezonuclear reaction, in addition to (1), should have occurred in biotite crystalline phase during the piezonuclear tests [1,2]: 24 28 Fe56 (2) 26 → Mg12 + Si14 + 4 neutrons

55

Analysis

Analysis

(b)

(a)

Analysis

Analysis

(c)

Analysis

(d)

(e)

Figure 6: Fe, Al, Si, Mg, and K concentrations in biotite: Fe (a), Al (b), Si (c), Mg (d) and K (e) concentrations in external and fracture surface for biotite are reported. The iron decrease (-3.00%) in biotite crystalline phase is counterbalanced by an increase in aluminum (+1.50%), silicon (+1.20%), and magnesium (+0.70%). In case of K no appreciable variations can be recognized between the uncracked and the fractured samples (see also Tab. 3).

56 CONCLUSIONS In the present paper Energy Dispersive X-ray Spectroscopy (EDS) was performed on different samples of external and fracture surfaces coming from specimens used in the piezonuclear tests recently shown in Carpinteri et al. [1] and Cardone et al. [2]. For each sample, different measurements of the same crystalline phases (phengite and biotite) were performed in order to get averaged information of the chemical composition and to detect possible piezonuclear transitions from iron to lighter elements. Considering the results for phengite and biotite, and considering also their abundances in the Luserna stone composition, a considerable reduction in the iron content (∼25%) is observed. This iron decrease is counterbalanced by an increase in aluminum, silicon, and magnesium. In particular, the increase in aluminum content corresponds to the eighty-five percent of the iron decrease. Therefore, the Author’s opinion is that piezonuclear fission reactions (1) and (2) should have occurred in granitic gneiss during the piezonuclear tests [1,2]. Finally, considering that granite is characterized by an extensive concentration in the rocks that make up the Earth’s continental crust (∼60% of the Earth’s continental crust), the piezonuclear fission reactions considered above can be generalized from the laboratory to the Earth’s crust scale, where mechanical phenomena of brittle fracture, due to tectonic activity, take place continuously in most seismic areas. ACKNOWLEDGEMENTS The financial support provided by Regione Piemonte (RE-FRESCOS project) is gratefully acknowledged. Special thanks are due to Prof. Alessandro Borghi of the Departement of Mineralogical and Petrological Sciences (Università di Torino) for the helpful discussions and suggestions on chemical and atomic proportion analyses. REFERENCES [1] Carpinteri, A., Cardone, F., Lacidogna G. (2009) Piezonuclear neutrons from brittle fracture: Early results of mechanical compression tests. Strain. 45, 332–339. [2] Cardone, F., Carpinteri, A., Lacidogna, G. (2009) Piezonuclear neutrons from fracturing of inert solids. Physics Letters A. 373, 4158–4163. [3] Arata, Y., Zhang, Y. (1995) Achievement of solid-state plasma fusion (‘‘cold-fusion’’). Proc. Japan. Acad. 71, Ser. B, 304–309. [4] Arata, Y., Fujita, H., Zhang, Y-C. (2002) Intense deuterium nuclear fusion of pycnodeuterium-lumps coagulated locally within highly deuterated atom clusters. Proc. Japan. Acad.78, Ser. B, 201–204. [5] Taleyarkhan, R. P., West, C. D., Cho, J. S., Lahey Jr., R. T., Nigmatulin, R. I., Block, R. C. (2002) Evidence for Nuclear Emissions During Acoustic Cavitation. Science. 295, 1978–1293. [6] Cardone, F., Cherubini, G., Petrucci, A. (2009) Piezonuclear neutrons. Phys. Lett. A. 373(8-9), 862-866. See also: F. Cardone et al. http://www.arxiv.org/abs/0710.5115. [7] Cardone, F., Mignani, R., (2007) Deformed Spacetime, Springer, Dordrecht, Chapters 16-17. [8] Vola, G., Marchi, M., (2009) Mineralogical and petrographic quantitative analysis of a recycled aggregate from quarry wastes. The Luserna stone case-study. Proc of the 12th Euroseminar on Microscopy Applied to Building Materials, 15-19 September 2009, Dortmund, Germany. [9] Sandrone R., Cadoppi P., Sacchi R., Vialon P. (1993) The Dora-Maira Massif. In: Von Raumer J.F., Neubauer F. (Eds.). Pre-Mesozoic geology in the Alps. Springer, Berlin, 317-325. [10] Sandrone, R., Borghi, A. (1992) Zoned garnets in the northern Dora-Maria Massif and their contribution to a reconstruction of the regional metamorphic evolution. European Journal of Minerals. 4,465-474. [11] Compagnoni R., Crisci G.M., Sandrone R. (1982-83) Caratterizzazione chimica e petrografica degli “gneiss di Luserna” (Massiccio cristallino Dora-Maira, Alpi Occidentali). Rend. Soc. It. Min. Petr. 38, 498. [12] Sandrone R. (2001) La Pietra di Luserna nella letteratura tecnico-scientifica. Sem. Int. Le Pietre Ornamentali della Montagna Europea, Luserna San Giovanni-Torre Pellice (TO), 10-12 giugno 2001, 333339. [13] Carpinteri, A. (1989) Cusp catastrophe interpretation of fracture instability. Journal of the Mechanics and Physics of Solids. 37, 567-582. [14] Carpinteri, A. (1990) A catastrophe theory approach to fracture mechanics. International Journal of Fracture. 44, 57– 69. [15] Carpinteri, A., Corrado, M. (2009) An extended (fractal) overlapping crack model to describe crushing sizescale effects in compression. Eng. Failure Analysis. 16, 2530–2540. [16] Cardone, F., Mignani, R., Petrucci, A., private communication.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Acoustic and electromagnetic emissions in rocks under compression

G. Lacidogna1 A. Manuello1, A. Carpinteri1, G. Niccolini2, A. Agosto2, G. Durin2 1

Politecnico di Torino, Department of Structural Engineering & Geotechnics Corso Duca degli Abruzzi 24 – 10129 Torino, Italy 2

National Institute of Metrological Research - INRIM Strada delle Cacce 91 – 10135 Torino, Italy E-mail: [email protected]

ABSTRACT The present paper focuses on Acoustic Emission (AE) and Electromagnetic Emission (EME) detected during laboratory compression tests. We investigated the mechanical behaviour of granite rock (Luserna stone) specimens with different size and shape. The recorded AE and EME signals were related to the time history of the load applied to the specimens. The results show the frequency range in which the EME due to fracture phenomena take place. In addition, the experimental evidence demonstrates how AE can be considered as a fracture precursor, since it precedes EME events, which accompany stress drops and related discontinuous fracture advancements. INTRODUCTION In this work we measured the electromagnetic field, given by the moving charges, during laboratory fracture experiments on specimens made of granitic gneiss with different shape and size. In particular, the mechanical behaviour of cylindrical “Luserna stone” specimens loaded up to their failure were monitored by Electromagnetic Emission (EME). All specimens were tested in compression at a constant displacement rate. One of these specimens was also monitored by piezoelectric (PZT) transducers for AE data acquisition. For all the specimens the investigation of magnetic activity was performed by two different measuring devices calibrated according to metrological requirements. These two EME devices differ according to the frequency range of acquisition. In all the considered cases, the magnetic signals were generally observed only in correspondence to sharp stress drops or to the final collapse. A number of laboratory studies have revealed the existence of electromagnetic emission (EME) during fracture experiments carried out on a wide range of materials [1-8]. The EME during failure of materials is analogous to the anomalous radiation of geoelectromagnetic waves observed before major earthquakes [9-11], reinforcing the idea that the EME effect can be applied as a forecasting tool for earthquakes. The present paper focuses on EME and acoustic emission (AE) detected during laboratory compression tests on concrete and rock specimens. While the mechanism of AE is fully understood, being provided by transient elastic waves due to stress redistribution following fracture propagation [12-17], the origin of EME from fracture is not completely clear and different attempts have been made to explain it. An explanation of the EME origin was related to dislocation phenomena [18,19], which however are not able to explain EME from fracture in brittle materials where the motion of dislocations can be neglected [5]. The weakness of the “dislocation movement hypothesis” was confirmed in some experiments showing that the EME amplitude increased with the brittleness of the investigated materials [20]. In brittle materials the fracture propagation occurs suddenly and is accompanied by abrupt stress drops in the stress-strain curve related to sudden loss in the specimen stiffness. Another relevant attempt to explain the EME origin was made through the “capacitor model”, where EME is assumed to be caused by net charges of opposite sign appearing on the vibrating faces of opening cracks [3,21]. This model is not widely accepted since an accelerated electric dipole created by charged opening cracks apparently does not explain EME from shearing cracks, which indeed are experimentally observed [5]. However, T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_8, © The Society for Experimental Mechanics, Inc. 2011

57

58 the general validity of this model can be maintained, since a certain separation between charged faces is guaranteed even in shear fractures. Frid et al. [5] and Rabinovitch et al. [8] recently proposed a model of the EME origin where, following the rupture of bonds during the cracks growth, mechanical and electrical equilibrium are broken at the fracture surfaces with creation of ions moving collectively as a surface wave on both faces. Lines of positive ions on both newly created faces (which maintain their charge neutrality unlike the capacitor model) oscillate collectively around their equilibrium positions in opposite phase to the negative ones (see Fig. 1). The resulting oscillating dipoles created on both faces of the propagating fracture act as the source of EME. Moreover, when a brittle material, like rocks or concrete, is mechanically excited, the most general motion of the system is a superposition of its normal modes, in which all parts of the system move sinusoidally with the same frequency (natural frequency or resonant frequency). The way specified in Fig. 1 is the atomic motion, which contributes to the general motion of the system and is the source of EM radiation. According to this model, the EME amplitude increases as long as a fracture propagates, since the rupture of new atomic bonds contributes to it. When fracture stops, the waves and the EME decay by relaxation. Since larger fracture advancements produce larger stress drops, this model agrees with the results in compression tests on rock specimens obtained by Fukui et al. [7], which establish a relationship of proportionality between stress drop and intensity of related EME.

Figure 1: (a) Picture of crack propagation and crack surfaces at a specific time. Crack surfaces are in the xz plane and the crack propagates in the x direction. (b) Schematic representation at a specific time of surface waves propagating on the two newly formed crack surfaces. Layers of atoms move together generating surface vibrational waves on each face, where positive charges vibrate in opposite phase to the negative ones. EXPERIMENTAL SETUP Four specimens made of Luserna granite were examined in this study (Fig. 2). They were subjected to uniaxial compression using a servo-controlled machine (MTS) with a maximum capacity of 250 kN and load measurement accuracy of ± 1.0%. This machine is equipped with control electronics which makes it possible to carry out tests in either load control or displacement control. Each test was performed in piston travel displacement control by setting constant piston velocity. The test specimens were arranged in contact with the press platens without any coupling materials, according to the testing modalities known as “test by means of rigid platens with friction”. The shapes and sizes of the specimens, and the employed piston velocities are listed in Table 1. The selected range of the piston velocities, reported in the table, are in the authors’ experience the most suitable values to evaluate AE and EME activity in quasi-brittle materials such as concrete and rocks [15,22-24].

59

(a)

(b) Figure 2: Specimens C1-C3 (a), specimen C4 (b).

Specimen

Diameter [mm]

Height [mm]

Slenderness λ=H/D

Piston velocity –1 [m s ]

C1

28

14

0.5

1.0×10

C2

28

28

1

1.0×10

C3

28

56

2

1.0×10

C4

53

100

2

2.0×10

–6 –6 –6 –6

Table 1: Geometry of the specimens, and compression test piston velocities. AE AND EME MEASUREMENTS The EME was detected using two different measuring devices calibrated according to metrological requirements. These two devices differ in the frequency range of acquisition. The first adopted device (Narda ELT-400 exposure level tester) works in the frequency range between 10 Hz and 400 kHz, the measurement range is between 1 nT 2 and 80 mT, and the three-axial measurement system has a 100 cm magnetic field sensor for each axis. This particular frequency range was chosen to avoid disturbances due to radio waves that operate on medium frequencies of 300-3000 kHz, or other electronic devices that generally operate on frequencies above 5-30 MHz. Narda ELT-400 device was placed 1 m away from the specimens (see Fig. 3a). The second EME device is constituted by three winding loops with different number of turns that can be positioned around the monitored specimen (Fig. 3b). This instrumentation, realized at the National Institute of Metrological Research (INRIM), acquires data in a larger frequency domain and in particular extends the bandwidth up to 4 MHz. The working principle is based on the induction Faraday’s law. It states that the electromagnetic force (voltage) in a closed circuit (loop) is proportional to the change of the magnetic flux in the windings section. Based on this principle, the number of turns was chosen to evaluate the magnetic field frequency. In fact, the three coaxial coils with an increasing number of turns are capable to perform the measurement from very low (Hertz) to high (MHz) frequencies in the magnetic field. The first coil, constituted by 5 turns, works in a frequency range from 300 kHz to 4 MHz. The other two coils constituted by 125 turns and 500 turns, work in the frequency range from 0 kHz to 20 kHz, and from 0 kHz to 1 kHz, respectively. Each turn is realized by a 0.2 mm copper wire, mounted on two coaxial PVC tubes embedded in a two components resin, in order to allow a large range of measurements. Specific tests were conducted to assess the potential EM environmental noise, and what affecting the EM signals due to the MTS test machine electronic control. In particular, the EM probe was used to detect the EM background noise for about five hours before the beginning of each compressive test. The background noise was estimated at about 40 nT in the frequency range 10 Hz - 400 kHz. Data acquisition of the EME signals was triggered when the magnetic field exceeded the threshold fixed at 0.2 µT, after the preliminary measurements to filter out the magnetic noise in the laboratory. As mentioned in the Introduction, only for one specimen Acoustic Emission was detected by applying to the sample surface a piezoelectric (PZT) transducer. This sensor is sensitive in the frequency range from 50 to 500 kHz for detection of high-frequency acoustic emission (AE) (Fig. 3a). In this case (specimen C4), both the recorded AE and EME signals were related to the time history of the load applied to the specimen.

60

(a)

(b)

Figure 3: (a) The electromagnetic antenna (Narda ELT-400) and the AE acquisition system used during the tests. (b) The electromagnetic coaxial coils device positioned around one of the monitored specimens. TEST RESULTS All specimens were tested in compression up to failure. The tests have shown a brittle response with a rapid decrease in load-carrying capacity beyond the peak load. In Figs. 4-6, for specimens C1-3 the load vs. time diagrams and the EM signals, detected by NARDA ELT-400 probe, are reported. The test specimens have the same diameter with a different slenderness λ, respectively of 0.5, 1 and 2. Specimen C1 presents a more ductile behaviour (Fig. 4) characterized by the descending branch of the load vs. time diagram. During the compression test three EM signals, with constant peak amplitude of 1.2 µT, were detected at 180s, 500s and 750s (see Fig. 4a). All these signals were anyway detected at stress-drops. During the post-peak stage, i.e., softening branch in the load vs. time diagram, no further EME signals were detected. In fact, at the peak load the fracture is completely formed and the subsequent stages are characterized only by opening of the fracture surfaces. According to the model proposed by Frid et al. [5] and Rabinovitch et al. [8], this means that no newly broken atomic bonds can contribute to EME. In Fig. 4b the FFT analysis for signal 1 is shown. The main detected frequency is close to 160 kHz according to the working frequency range of the Narda ELT-400. Otherwise, specimens C2 and C3 show a brittle behaviour, characterized by abrupt stress drops after reaching the peak load. The experimental results are reported in Figs. 5 and 6. Five EM signals were detected during the specimen C2 test. In specimen C3 four EM signals were observed. Only in the case of specimen C2 the signals amplitude (in µT) seem to be proportional to the stress-drop values. The experimental results obtained in the described test are in good agreement with those obtained in [25]. In Fig. 7a the load vs. time diagram for specimen C4, monitored by the EM coaxial coils device, is shown. This specimen has a diameter equal to 53 mm and a slenderness λ= 2. The load vs. time diagram is almost linear up to failure. At 70% of the peak load a significant increase in the AE rate is observed (see the slope of the dashed line in Fig. 7a). During this test a EME signal with amplitude equal to 16 µT was detected in correspondence to the peak load (1220 s). In Fig. 7b the signal amplitude and its FFT analysis are shown. From the FFT a main frequency of about 5 MHz is identified. Therefore, as shown in a recent paper by the authors [25], the results shown in the present paper demonstrate again that AE can be considered as a fracture precursor, since it precedes EME events, which accompany stress drops and related discontinuous fracture advancements.

61

(a)

(b) Figure 4: (a) Load vs. time curve of the Luserna stone specimen C1. (b) Amplitude and FFT analysis for the detected signal 1.

Figure 5: Load vs. time curve of the Luserna stone specimen C2. Five EM signals were detected during the test.

62

Figure 6: Load vs. time curve of the Luserna stone specimen C3. Four EM signals were detected during the test.

1000

Load

600 400

Cumulated AE

Cumulated AE

Load (kN)

800

200

0

(a)

Time (s)

EM coaxial coils device

(b) Figure 7: (a) Load vs. time curve and AE cumulated number (dashed line) for specimen C4. A signal of 16 µT is detected in correspondence to the peak load. (b) Amplitude and FFT analysis for the detected signal.

63 CONCLUSIONS The experimental evidence presented in this paper confirms AE and EME signals as collapse precursors in brittle materials such as rocks in compression. However, the main advantage using EME in the fracture monitoring is that the signal propagation is not affected by refraction and reflection phenomena as in the case of AE. The antennas used for the acquisition of EME do not require to be applied on the external surface of the monitored specimen or structure as in the case of AE transducers. Moreover, the experimental results show that the typical EME detected during the tests are included in the frequency range from 160 kHz to 5MHz. The specimens presented herein have also been employed for the detection of neutron emissions during piezonuclear tests [26]. Considering that the employed helium-3 detector is insensitive to electromagnetic charges in the frequency range from 150 kHz to 230 MHz, the results presented in this paper show that it is possible to exclude the electromagnetic noise as signal disturbances in the neutron detection. ACKNOWLEDGEMENTS The financial support provided by Regione Piemonte (RE-FRESCOS project) is gratefully acknowledged. The acquisition data support provided by the Department of Mechanical Engineering (Università degli studi di Cagliari) is also acknowledged. We are grateful to Dr. Paolo Roccato and Dr. Luca Martino of the National Research Institute of Metrology – INRIM for their valuable assistance in the EME signals elaboration process. Special thanks are due to Dr. Oscar Borla (Istituto Nazionale di Fisica Nucleare – INFN Torino) for his suggestions and active collaboration in the mechanical compressive tests. REFERENCES [1] Warwick, J.W., Stoker, C. and Meyer, T.R. (1982) Radio emission associated with rock fracture: Possible application to the great Chilean earthquake of May 22, 1960, J. Geophys. Res., 87, 2851-2859. [2] Ogawa, T., Oike, K., and Miura, T. (1985) Electromagnetic radiation from rocks. J. Geophys. Res., 90, 62456249. [3] O’Keefe, S. G. and Thiel, D. V. (1995) A mechanism for the production of electromagnetic radiation during fracture of brittle materials, Phys. Earth Plant. Inter., 89, 127-135. [4] Lolajicek, T. and Sikula, J. (1996) Acoustic emission and electromagnetic effects in rocks, Progress in Acoustic Emission VIII, 311-314. [5] Frid ,V., Rabinovitch, A. and Bahat, D. (2003) Fracture induced electromagnetic radiation, J. Phys. D., 36, 1620-1628. [6] Hadjicontis, V., Mavromatou, C. and Nonos, D. (2004) Stress induced polarization currents and electromagnetic emission from rocks and ionic crystals, accompaying their deformations, Nat. Hazards and Earth System Science, 4, 633-639. [7] Fukui, K., Okubo, S. and Terashima, T. (2005) Electromagnetic radiation from rock during uniaxial compression testing: The effects of rock characteristics and test conditions, Rock Mech. Rock Eng., 38, 411423. [8] Rabinovitch, A., Frid, V. and Bahat, D. (2007) Surface oscillations. A possible source of fracture induced electromagnetic oscillations, Tectonophysics, 431, 15-21. [9] Gokhberg, M.B., Morgunov, V.A., Yoshino, T. and Tozawa, I. (1982) Experimental measurement of electromagnetic emissions possibly related to earthquakes in Japan, J. Geophys. Res., 87, 7824-7828. [10] Nagao, T., Enomoto, Y., Fujinawa, Y. et al. (2002) Electromagnetic anomalies associated with 1995 Kobe earthquake, Journal of Geodynamics, 33, 401-411. [11] Karamanos, K., Dakopoulos, D., Aloupis, K. et al. (2006) Preseismic electromagnetic signals in terms of complexity, Physical Review E, 74, 016104. [12] Kaiser, J. (1950) An investigation into the occurrence of noises in tensile tests, or a study of acoustic phenomena in tensile tests. Ph. D. dissertation, Munich (FRG), Technische Hochschule München. [13] Pollock, A.A. (1973) Acoustic emission-2: Acoustic emission amplitudes, Non-Destructive Testing, 6, 264269. [14] Ohtsu, M. (1996) The history and development of acoustic emission in concrete engineering, Magazine of Concrete Research, 48, 321-330. [15] Carpinteri, A., Lacidogna, G. and Pugno, N. (2007) Structural damage diagnosis and life-time assessment by acoustic emission monitoring, Engineering Fracture Mechanics, 74, 273-289. [16] Carpinteri, A., Lacidogna, G. and Manuello, A. (2007) Damage mechanisms interpreted by acoustic emission signal analysis, Key Engineering Materials, 347, 577-582. [17] Carpinteri, A., Lacidogna, G., Niccolini, G. and Puzzi, S. (2008) Critical defect size distributions in concrete structures detected by the acoustic emission technique, Meccanica, 43, 349-363.

64 [18] Misra, A. (1977) Theoretical study of the fracture-induced magnetic effect in ferromagnetic materials, Physics Letters, 62A, 234-236. [19] Misra, A. (1978) A physical model for the stress-induced electromagnetic effect in metals, Applied Physics, 16, 195-199. [20] Jagasivamani, V. and Iyer, K.J.L. (1988) Electromagnetic emission during the fracture of heat-treated spring steel, Mater. Lett., 6, 418-422. [21] Miroshnichenko, M. and Kuksenko, V. (1980) Study of electromagnetic pulses in initiation of cracks in solid dielectrics, Soviet Physics-Solid State, 22, 895-896. [22] Carpinteri, A., Lacidogna G. and Manuello, A. (2009) The b-value analysis for the stability investigation of the ancient Athena Temple in Syracuse, Strain, doi: 10.1111/j.1475-1305.2008.00602.x. [23] Lacidogna, G., Manuello, A., Durin, G., Niccolini, G., Agosto, A. and Carpinteri, A. (2009). Acoustic and magnetic emissions as precursor phenomena in failure processes, Proc. of SEM Annual Conference & Exposition on Experimental and Applied Mechanics, Albuquerque, 1-4 June 2009, Paper No. 540. [24] Schiavi, A., Niccolini, G., Tarizzo, P., Lacidogna, G., Manuello, A. and Carpinteri, A (2009) High and low frequency elastic wave propagation in brittle materials under compression, Proc. of SEM Annual Conference & Exposition on Experimental and Applied Mechanics, Albuquerque, 1-4 June 2009, Paper No. 539. [25] Lacidogna, G., Carpinteri, A., Manuello, A., Durin, G., Schiavi, A., Niccolini, G. Agosto, A (2010) Acoustic and electromagnetic emissions as precursor phenomena in failure processes, Strain, in print. [26] Carpinteri, A., Lacidogna G., Manuello, A., Borla, O (2010) Evidence of piezonuclear reactions: From geological and tectonic transformations to neutron detection and measurements, Proc. of SEM Annual Conference & Exposition on Experimental and Applied Mechanics, Indianapolis, 7-10 June 2010, Paper No. 458.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Gear with Asymmetric Teeth for use in Wind Turbines S. Ekwaro-Osire 1, T.-H. Jang, A. Stroud, I. Durukan, F.M. Alemayehu Mechanical Engineering Department Texas Tech University Lubbock, TX 79409-1021 A. Swift, J. Chapman Wind Science and Engineering Research Center Texas Tech University Lubbock, TX 79409-1021 ABSTRACT In the US, wind energy is one of the electrical energy sources that are growing fastest. The growth has been linear at a rate of about 20% to 30% per year over the last decade. For the next two decades, the wind industry has set its goal to more than 20%. However, there still remains concern on the reliability of wind turbines. This concern is often directed to the fact that gearbox failure has been a major problem in the wind industry. Compared to the other wind turbine components, gearbox failures result in the second highest down time per failure. Currently, there are several initiatives underway to improve gearbox reliability in wind turbines. Additionally, due to the increasing performance requirements, there has also been need of new gear designs. Recently, theoretical analyses have shown that asymmetric gears may offer a potential to reduce the costs associated with the gear failures, while at the same time maintaining the fatigue life. Also, for wind turbine gearboxes, the gears experience only uni-directional loading. In these instances, the geometry of the drive side does not have to be symmetric to the coast side. This allows for the designing of gears with asymmetric teeth. The objective of this research was to design and construct a testbed for testing the performance of asymmetric gears. The tests to be performed on the testbed include gear dynamics, tip relief modification, high-contact-ratio, and wear. BACKGROUND In the US, wind energy is one of the fastest growing electrical energy sources [1]. Its growth has been linear at a rate of about 20% to 30% per year over the last decade [2]. For the next two decades, the wind industry has set its goal to more than 20%. Despite the projected increase in this sector, concern still remains about the reliability of wind turbines. Since gearbox replacement is expensive and gearbox reliability is a key issue, research in gearbox technology has increased in importance. With the recognition of the impact of gearbox failures, there is wide spread agreement in the wind sector community that in the next several years, the turbine drive-train technology will be evolved significantly to improve reliability and reduce cost and weight [3]. Recently, due to the increased interest in accounting for the uncertainty involved in wind turbines, there has been an interest in utilizing probabilistic techniques in the design of wind turbines. Veldkamp [4] presented a probabilistic approach for the design of wind turbines. In his study, he noted that the important stochastic parameters influencing fatigue loads include wind characteristics (e.g., average wind speed, cut out wind speed, turbulence intensity, and wind field shear), material properties (e.g., fatigue strength), geometry (e.g., dimensions), and aerodynamics (e.g., edge moment blade). Musial et al. [5] discussed a multi-year research and development initiative to improve gearbox reliability in wind turbines. This initiative comprised both testing and analysis efforts. This initiative seeks to address the problem within the design process by developing solutions in order to speed up improvements in wind turbine gearbox. The approach proposed involves drive-train analysis, dynamometer testing, and field testing. Another comprehensive initiative is the “UpWind” project underway in Europe [6]. This is the largest long-term wind energy research project that the European Union has ever funded. 1

Corresponding author: [email protected]

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_9, © The Society for Experimental Mechanics, Inc. 2011

65

66 In this initiative, a multi flexible body dynamics-based wind turbine model with a detailed model of the gearbox was created. The gears inside the gearbox and their tooth contact were modeled with high detail. Experimental data was used to verify the model. Lately, there has been a lot of research activity on spur gears with asymmetric teeth. New gear designs are needed because of the increasing performance requirements placed on wind turbines, such as high load capacity, high endurance, low cost, long life, and high speed. In wind turbine gearboxes, the gears experience only uni-directional loading. In this instance, the geometry of the drive side does not have to be symmetric to the coast side. This allows for the design of gears with asymmetric teeth. The theoretical work conducted thus far [7-15] have shown that asymmetric gears may offer a potential to reduce the costs associated with the gear failures, while at the same time maintaining fatigue life. Other than the theoretical studies of asymmetric gears, there have not been experimental studies published. Hence, the overall objectives of the research was to (1) design and construct a gear testbed, (2) conduct performance tests on asymmetric gears, and (3) conduct reliability analysis and dynamic finite element analysis on asymmetric gears. This paper focuses only on the design and construction of a gear test bed. The tests the testbed can perform include gear dynamics, tip relief modification, high-contact-ratio, and wear. ASSYMETRIC GEARS In previous studies, related to bending stress and load capacity, high performance has been achieved for gears with asymmetric teeth. These gears provide flexibility to designers due to their non-standard design. If they are correctly designed, they can make important contributions to the improvement of designs of gears in wind turbine industry [9, 12, 13, 16]. An asymmetric tooth is shaped using different pressure angles on the drive side and coast F F side of the tooth (see Figure 1). In the design of asymmetric teeth, the choice of pressure angles on drive and coast side Coast Coast Drive Drive is very important to obtain satisfactory performances.Most side side side side of the recent research on the benefits of involute spur gears with asymmetric teeth has focused on geometrical design and stress analysis [16-20]. It has been shown that as the (a) (b) pressure angle increases, the root fillet stress and contact Figure 1: Asymmetric gear teeth, (a) pressure stress decrease significantly. In few studies [21], the effects angle larger on drive side than on coast side, (b) of various parameters, such as pressure angle and tooth pressure angle smaller on drive side than on height on the dynamic load and the static transmission coast side. errors of spur gears with asymmetric teeth, were investigated. On comparing the spur gears with asymmetric and symmetric teeth, it was shown that for asymmetric teeth, increasing the addendum leads to a significant decrease in the dynamic factor, static transmission error, and root fillet stress. Karpat and Ekwaro-Osire [7] studied the wear of involute spur gears with asymmetric teeth under dynamic loading. They observed an interaction between wear and dynamic loads for spur gears with asymmetric teeth. It was shown that, as the pressure angle on the drive side increases, wear depth decreases considerably. GEAR TESTING TESTBEDS A mechanical system exhibits combined parametric excitation and clearance type non-linearity. For the validation of the analytical solution, Blankenship and Kahraman [22] developed a gear dynamics test rig in order to demonstrate non-linear behaviors in mechanical oscillator and compared with numerical integration results and experimental measurements. Petry-Johnson et al. [23] developed a spur gear efficiency test machine for the experimental investigation of high-speed spur gear efficiency for both jet-lubricated, dry sump conditions and diplubricated conditions. The experimental results showed that the influence of rotational speed, oil viscosity, oil bath level, and rotational direction on load independent power loss was quantified. Based on their experiments, the authors proposed an efficiency model to predict the instantaneous mechanical efficiency of a gear pair under typical operating, surface, and lubrication conditions. The predictions from the model were shown to be within 0.1% of the measured values [24]. Seetharaman and Kahraman [25] performed experiments over a wide range of operating speed, temperature, oil levels, and key gear design parameters. Among others, their studies included the spin power losses of spur gear pairs operating under dip-lubricated conditions. Their measurements indicated that the static oil level, rotational speed, and face width of gears have a significant impact on spin power losses. Kahraman and Blankenship [26] designed a gear test rig to investigate the influence of involute contact ratio and linear involute tip relief on the torsional vibration behavior of a spur gear pair. The influence of involute contact ratio on dynamic transmission error is quantified and a set of generalized, experimentally validated design

67 guidelines for the proper selection of involute contact ratio to achieve quite gear systems is presented. Using the same rig, the authors were also able to study the dynamic transmission error (DTE) values [27]. Begg et al. [28, 29] constructed a mechanical diagnostics test bed (MDTB). It was constructed as a multi-sensor instrumented gear/transmission test stand with a primary interest of studying the mechanical fault evolution in damaged rotating components that involve mechanical power transmission, such as a gearbox. The system speed and torque set points were produced by analog input signals supplied by a data acquisition computer. Shafts were connected with tandem flexible and rigid couplings. Torque-limiting shear couples were used on both sides of the gearbox to prevent transmission of excessive torque as could occur with gear jam or bearing seizure [28-30]. Begg et al. [29] developed a dynamic model of the system for response simulation, for example, vibratory responses. These responses allowed the authors to get a physical insight on locating areas of placements and specification of vibratory measurement sensors. Their gearbox was instrumented with accelerometers, thermocouples, acoustic emission sensors, and oil debris sensors. Their MDTB has the capability of testing single and double reduction industrial gearboxes with ratios from about 1.2:1 to 6:1 and with ratings that can range from 5 to 20 HP. Lin et al. [31] have also used condition vibration data collected from a series of test runs of single reduction helical gearboxes by the test bed. This data contained vibration signatures captured by accelerometers mounted at different positions on the gearbox. The data was used to calculate the fault growth parameter which is used in the formulation of a condition based maintenance policy. Despite the fact that currently, the cost of wind energy continues to decrease, the costs associated with turbine construction have increased. This has made prototype field-testing of wind turbines expensive, and has created a demand for component or assembly dynamometer testbeds. The National Renewable Energy Laboratory has a dynamometer testbed for wind turbine drive-train components and systems from 100 kW to 2.5 MW [32]. It can be used to carry out two types of tests, namely, (1) wind turbine system tests (which includes, drive-train endurance testing, turbulent wind simulation testing, and load event testing), and (2) component testing (which includes bearings, controllers, couplings, gears, gearboxes, generators, lubricant systems, power conversion systems, and shafts) [33]. TESTING OF ASSYMETRIC GEARS The experimental research in this paper will use field data to demonstrate the efficacy of using gears with asymmetric teeth in the gear trains of wind turbines. It will also demonstrate how these gears would improve the reliability and cost of wind turbines. The gear testbed introduced in this paper consists of six main units, namely frame, control and data acquisition system, dynamometer (variable speed), gear test unit, power electronics, and sensors. A schematic of the proposed testbed is shown in Figure 2. The frame and all the components of the testbed is designed using the solid modeler ProENGINEER. Structural analysis is carried out using finite element analysis package ANSYS. Both static and dynamic analyses were conducted. This was done to assure that during the running of the motors, vibrations do not occur near the resonance of the frame, thus interfering with the tests.

Figure 2: Gear testbed

68 The layout of the elements of current testbed is depicted in Figure 3 [13]. The current layout has a frame, dynamometer, generator, gear test unit, control system, and power electronics box. The test bed is designed in such a way that both sides of the gearbox are connected to the drive and load side using flexible torque limiting shear couples. These couples, in addition to the flexibility provision, will be used as safety device to prevent excessive torque that could occur when there is gear jam and bearing seizure. The detailed design of an innovative wind power testbed for testing the applicability of asymmetric gears in wind turbine application is presented. A variable frequency drive control system of the testbed allows the use of realistic wind profiles to drive a variable-speed motor. The testbed will be used to conduct gearbox experiments with a focus on: efficiency, dynamic loading, gear rattle and vibrations, fatigue life, acoustic emissions, tooth wear, and bearing wear.

Gear Test Unit

Slip Ring

Torque Sensor Speed Reducer Dynamometer

Generator Coupling Bearing

Frame

Figure 3: Constructed testbed The data acquisition system was configured in such a manner to maximize the data collection from the experimental testbed. The main elements of the data acquisition system, include: two connector blocks, one motion controller, one data acquisition card, one signal conditioner for the torque sensors, one signal conditioner for the thermocouples, and one PXI controller. Apart from the connector blocks, all these elements are mounted into a single chassis. The details of the data acquisition system are shown in Figure 4. The essential elements of this system were acquired from National Instruments Corporation.

69 Figure 4: Data acquisition system In this second objective of the proposed research, the results obtained in the published manuscripts [7, 9, 10, 12] will be the main focus of the testing. The tests to be conducted will include: dynamic loading, tip relief modification, high-contact-ratio, and wear. The dynamic loading will be measured using the strain gages mounted on the surface of gears. Load variation will be measured using the two torque sensors which are mounted on the either side of the gear test unit. The involute contact ratio will be defined as the change in the total number of tooth pairs in contact as gears rotate. The influence of involute contact ratio will be investigated by measuring the dynamic transmission error (DTE) of the gear pairs. DTE is defined as a relative displacement at the gear mesh interface. DTE is measured by attaching piezoelectric accelerometers tangentially to each gear wheel. The testbed will also allow for surface wear measurements. In the test, the tip relief modification will be defined as the effect of linear involute tip on spur gear performance. For this, the measured DTE will be used as the response variable. For these tests a prescribed dynamometer speed time-history, based on field data, will be applied. The wind field data, for use generating the time histories to drive the variable-speed motor, will be acquired from the Wind Science and Engineering Research Center’s 200 meter tower. The results of the tests will compare symmetric spur gears and asymmetric spur gears, with the application in wind turbines. The symmetric spur gears and asymmetric spur gears used in the testing were purchased from a gear manufacturer. The designed testbed has the following advantages over the previously used testbeds: (1) adaptability to test different asymmetric gear size, and (2) smooth drive speed variability that uses wind profiles from field data. Additional to the tests that will be performed on the testbed, two types of analyses will be conducted, reliability analysis (see Figure 5) and dynamic finite element analysis. The dimensions of the models used in the analyses will have the same dimensions as those used in the experimental tests. The solid modeler ProENGINEER will be used to create the 3-D models of the gears. These models will then be transferred to the finite element analysis package ANSYS for analysis. A dynamic finite element will be conducted. The results obtained from the finite element analysis will be used to validate the experiments results.

Figure 5 Reliability Analysis SUMMARY Since gearbox replacement is expensive and gearbox reliability is a key issue, research in gearbox technology has increased in importance. From previous studies, it has been noted that if gears with asymmetric teeth are correctly designed, they can make important contributions to the improvement of designs of gears in wind turbine industry. The objective of this research was to design and construct a testbed for testing the performance of asymmetric gears. The detailed design of an innovative wind power testbed for testing the applicability of asymmetric gears in wind turbine application is presented. A variable frequency drive control system of the testbed allows the use of realistic wind profiles to drive a variable-speed motor. The testbed will be used to

70 conduct gearbox experiments with a focus on: efficiency, dynamic loading, gear rattle and vibrations, fatigue life, acoustic emissions, tooth wear, and bearing wear. The designed testbed has the following advantages over the previously used testbeds: (1) adaptability to test different asymmetric gear size, and (2) smooth drive speed variability that uses wind profiles from field data.

REFERENCES 1. Thresher R, Robinson M, Veers P, "Wind energy technology: Current status and r&d future," in Physics of Sustainable Energy, Berkeley, California, 2008. 2. Thresher R, Robinson M, Veers P, "The future of wind energy technology in the united states," in World Renewable Energy Congress, Glasgow, Scotland, UK, 2008. 3. U.S. Department of Energy, "20% wind energy by 2030 -- increasing wind energy’s contribution to u.S. Electricity supply," DOE/GO-102008-2567, 2008. 4. Veldkamp D, "A probabilistic approach to wind turbine fatigue design," in European Wind Energy, Milan, Italy, 2007. 5. Musial W, Butterfield S, McNiff B, "Improving wind turbine gearbox reliability," in European Wind Energy, Milan, Italy, 2007. 6. Röthlingshöfer T, Brecher C, Schröder T, Defourny M, Granville D, Hemmelmann J, Pape M, Quell A, Paulsen US, "Investigation of the dynamic behaviour of a wind turbine drive train for the use in a design platform," in European Wind Energy, Milan, Italy, 2007. 7. Karpat F, Ekwaro-Osire S, "Wear of involute spur gears with asymmetric teeth under dynamic loading," in ASME International Mechanical Engineering Congress and Exposition, Chicago, 2006. 8. Karpat F, Ekwaro-Osire S, "Influence of tip relief modification on the wear of spur gears with asymmetric teeth," in STLE Annual Meeting & Exhibition, Philadelphia, Pennsylvania, 2007. 9. Karpat F, Ekwaro-Osire S (2008) Influence of tip relief modification on the wear of spur gears with asymmetric teeth. Tribology Transactions, 51:581-588. 10. Karpat F, Ekwaro-Osire S, "Dynamic analysis of high-contact-ratio spur gears with asymmetric teeth," in 2008 ASME International Mechanical Engineering Congress & Exposition, Boston, Massachusetts, 2008. 11. Karpat F, Ekwaro-Osire S, Cárdenas-García JF, "Photoelastic analysis of an asymmetric gear tooth," in SEM Annual Conference & Exposition, Springfield, Massachusetts, 2007. 12. Karpat F, Ekwaro-Osire S, Cavdar K, Babalik FC (2008) Dynamic analysis of involute spur gears with asymmetric teeth. International Journal of Mechanical Sciences, 50:1598-1610. 13. Karpat F, Ekwaro-Osire S, Chapman J, Swift A, "Wind power test bed," in WindPower 2007, Los Angeles, California, 2007. 14. Karpat F, Ekwaro-Osire S, Khandaker MPH, "Probabilistic analysis of mems asymmetric gear tooth," in ASME International Mechanical Engineering Congress and Exposition, Chicago, IL, 2006. 15. Karpat F, Ekwaro-Osire S, Khandaker MPH (2008) Probabilistic analysis of mems asymmetric gear tooth. Journal of Mechanical Design, 130:042306-1–042306-6. 16. Litvin FL, Lian Q, Kapelevich AL (2000) Asymmetric modified spur gear drives: Reduction of noise, localization of contact, simulation of meshing and stress analysis. Computer Methods in Applied Mechanics and Engineering, 188:363-390. 17. Kapelevich AL, Shekhtman YV (2003) Direct gear design: Bending stress minimization. Gear Technology:4449. 18. Kapelevich A (2000) Geometry and design of involute spur gears with asymmetric teeth. Mechanism and Machine Theory, 35:117-130. 19. Di Francesco G, Marini S (2005) Asymmetrical gear wheels: Automatized procedure for the design. VDI Berichte, 1904 II:1735-1742. 20. Brecher C, Schafer J (2005) Potentials of asymmetric tooth geometries for the optimisation of involute cylindrical gears. VDI Berichte, 1904 I:705-720. 21. Karpat F, Cavdar K, Babalik FC, "An investigation on dynamic analysis of involute spur gears with asymmetric teeth: Dynamic load and transmission errors," in Power Transmissions 2006, Novi Sad, 2006, pp. 69-74. 22. Blankenship GW, Kahraman A (1995) Steady state forced response of a mechanical oscillator with combined parametric excitation and clearance type non-linearity Journal of Sound and Vibration, 185:743-765 23. Petry-Johnson TT, Kahraman A, Anderson NE, Chase DR (2008) An experimental investigation of spur gear efficiency. Journal of Mechanical Design, 130:062601-062610.

71 24. Xu H, Kahraman A, Anderson NE, Maddock DG (2007) Prediction of mechanical efficiency of parallel-axis gear pairs. Journal of Mechanical Design, 129:58-68. 25. Seetharaman S, Kahraman A (2009) Load-independent spin power looses of a spur gear pair: Model formulation. Journal of Tribology, 131. 26. Kahraman A, Blankenship GW (1999) Effect of involute contact ratio on spur gear dynamics. Journal of Mechanical Design, 121:112-118. 27. Kahraman A, Blankenship GW (1999) Effect of involute tip relief on dynamic response of spur gear pairs. Journal of mechanical design, 121:313-315. 28. Begg CD, Byington CS, Maynard KP, "Dynamics modeling for mechanical fault transition," in 54th Meeting of the Society for Machinery Failure Prevention Technology, Virginia Beach, VA, 2000, pp. 203-212. 29. Begg CD, Merdes T, Byington C, Maynard K, "Dynamics modeling for mechanical fault diagnostics and prognostics," in Maintenance and Reliability Conference (MARCON 99), Gatlinburg, Tennessee, 1999. 30. Merdes TA, Lang DC, Kozlowski JD, Meister K, "Wear particle analysis results for variably loaded single reduction helical gearboxes," Pennsylvania State University 1998. 31. Lin D, Wiseman M, Banjevic D, Jardine AKS (2004) An approach to signal processing and condition-based maintenance for gearboxes subject to tooth failure. Mechanical Systems and Signal Processing, 18:9931007. 32. Musial W, McNiff B, "Wind turbine testing in the nrel dynamometer test bed," in WindPower 2000, Palm Springs, California, 2000. 33. Green J, "225-kw dynamometer for testing small wind turbine components," in WindPower 2006, Pittsburgh, Pennsylvania, 2006.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Material Brittleness and the Energetics of Acoustic Emission Dr. Adrian A. Pollock, Principal Scientist, MISTRAS Group Inc. 195 Clarksville Road, Princeton Junction, NJ 08550, [email protected] ABSTRACT This paper will review energy aspects of the acoustic emission (AE) phenomenon and its relationship to material properties especially brittleness. The spectral energy density of the AE wave at low frequencies is related to the moment tensor, but this is only a fraction of the total energy converted in the deformation or damage process. The “conversion efficiency” from static elastic energy to dynamic AE energy is governed by the source speed, and this in turn is related to the brittleness of the material. Meanwhile, the spectral bandwidth of the AE near the source is governed by the duration of the source event. The resulting relationships between brittleness and acoustic emissivity will be discussed. Examples will be drawn from metals, fiber reinforced composites and geological materials. A further factor that has a strong influence on a material’s damage tolerance is its heterogeneity. This also has a strong influence on its acoustic emissivity, specifically on the amplitude distribution. In a recent development in the practical application of AE to industrial plant monitoring, these factors and others are integrated in a model of the Probability of Detection (POD) for fatigue cracks growing in a mixed mode comprising both ductile and brittle deformation mechanisms. INTRODUCTION Acoustic emissions are transient elastic waves generated by the rapid release of energy from localized sources within a material [1]. The phenomenon has been observed over a wide range of magnitudes, from the atomic scale (as in dislocation movement) to the planetary scale (as in earthquakes and moonquakes) and even the stellar scale. It is of scientific interest in the study of deformation and fracture, and of practical interest for nondestructive testing of industrial products and structures. In general, elastic waves (acoustic waves) are generated and propagate when unbalanced forces are present in elastic media. In the case of acoustic emission (AE), the unbalancing of forces characteristically takes place when there is sudden plastic strain or sudden formation of new surface in stressed material. Brittle fracture is a prime example. In brittle fracture, new surface is suddenly created and the material near the new surfaces accelerates in the direction of the now-unbalanced tractive forces. The resulting movement propagates as an elastic wave, following the wave equation ü

= F(x,t) + ( + 2μ) (u) - µ x ( x u)

(1)

In Equation 1,  is the density, u is the particle displacement at position x and time t, F is the external driving force per unit volume (often zero) and the terms in  represent the force on unit volume resulting from the stress field gradients in all three directions x1, x2, x3.  and μ are elastic constants known as the Lamé constants. µ is the well-known shear modulus, and ( + 2μ) is the elastic modulus applicable to unidirectional tensile/compressive strain. The movement communicated to the material depends on how quickly the new boundary is formed, as will be discussed below. When a crack grows, energy is drawn from the elastic stress field and converted into several other forms. Some of the energy goes into the acoustic emission (AE) wave. This wave propagates outwards from the source, dispersing through the structure and eventually becoming absorbed through the degradation of the wave energy into heat and local microstructural change. The AE wave serves to redistribute the stresses within the structure on its way to a new equilibrium. The boundary conditions on the total system play an important role in this redistribution of the stress field. The AE wave is broadband in nature, carrying energy at all frequencies from zero up to a high limit that is essentially the inverse of the duration of the source event [2].

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_10, © The Society for Experimental Mechanics, Inc. 2011

73

74

The subject of this paper is the energy transformations that take place during this process. Attention is directed mostly to the source event and to the difference between initial and final states, rather than to the dissipative processes that occur during wave propagation. Several applicable approaches and models have been developed over the years. An attempt is made to discuss the most pertinent models from several schools of thought, and to see what they have in common or in contrast. Fracture mechanics, micromechanics, seismology and moment tensor analysis, and some previous models by the author can all bring insights to this area. OVERVIEW OF ENERGY CONVERSIONS DURING CRACK GROWTH Energy conversions during crack growth (or other AE processes) are shown and terms are defined in Figure 1. When the AE event takes place in a closed system, there is a reduction in the elastic energy stored in the system. We will call this energy drawn from the stress-strain field the “source energy”. The source energy is divided among several forms [2] of which the ones of interest here are surface energy, energy of plastic deformation, and the energy of the acoustic emission wave.

Stored Elastic Energy

Source Energy

Event Energy

Stored Elastic Energy

Plastic Deformation Surface Energy Acoustic Emission

Figure 1. Energy Conversions: Source Energy and Event Energy

In referring to “surface energy”, we mean the energy of disbonding the single layer of atoms that forms the newly created surface. Figure 1 shows this surface energy component as greater in ductile fracture than in brittle fracture, on the grounds that (in metals) the ductile-fracture surface is microscopically rougher and its actual area is larger. The formation of this surface is qualitatively different from the disordering of the atomic lattice in the layer of material adjacent to the surface. In the layer adjacent to the surface, energy is expended on dislocation

75

movement and slip, creating the plastic zone that surrounds the crack tip. This is what we mean by the caption “Plastic Deformation” in Figure 1. When a crack in a metal is running fast, this plastic zone is formed dynamically and continuously and is very important, even though it may not maintain the full size that it had at the moment the crack became unstable. RELATIVE MAGNITUDES OF THE ENERGY TERMS In Figure 1 the magnitude of the source energy depends only on the initial and final geometrical configuration and loads, but its partitioning among the resulting forms depends on other factors such as the brittleness of the material and the speed with which the source event takes place. Figure 1 illustrates this concept by depicting two different material processes labeled “ductile” and “brittle”. In the “ductile” process the great majority of the source energy is shown as being expended in plastic deformation, an intermediate amount is as surface energy and very little as AE. In the “brittle” process the source energy is the same, but much less energy is shown as expended in plastic deformation, considerably less as surface energy and much more as AE. The energy released as AE is termed the “event energy”. The double depiction of Figure 1 is useful for emphasizing the variability of the partitioning, but a graphic representation like this cannot adequately portray the huge range in the relative magnitudes of the energy terms that is actually found in real materials. Consider for example the relative magnitudes of the surface energy and plastic energy terms. Interatomic forces are much more heavily disrupted on the surface layer than in the plastic zone so one might think that this surface energy would be important. However, plastic zones in metals are often macroscopic in size and because so much more material is involved, their energy of formation can be orders of magnitude larger than the surface energy. To take an example: typical values of surface energy for solids are on the order of 2J/m2. Compare this with a measured 10,000J/m2 energy absorption rate for plasticity around a fastrunning crack, in a representative low-temperature test on a double cantilever beam specimen of a manganese molybdenum steel [3]. An order of magnitude estimate for the AE energy release in these same large crack jumps was 300J/m2. Thus in this particular example of brittle crack growth in steel, the plastic deformation accounted for virtually all the source energy, the AE energy accounted for just a few percent and the surface energy accounted for a few hundredths of a percent. Surface energy can be much more important in nonmetallic materials such as minerals and glass, as in the work of Griffith that laid the foundation of fracture theory. It is widely stated that Griffith’s energy criterion does indeed explain the observed fracture behavior of glass, but is inadequate for metals. To demonstrate the magnitudes involved in this, Table 1 shows values of the critical strain energy release rate, GIc, for several metals and for a glass, along with a value for the typical surface energy of solids. It is clear that surface energy is insignificant to the propagation of cracks in metals, but will be significant in glass. Table 1 Surface energy is insignificant in metals but significant in glass Stainless Steel

4340 Steel

7075 Aluminum

Soda Glass

Critical stress intensity factor KIc (MPa-m1/2)

200

50

25

0.7

Critical strain energy release rate GIc (J/m2)

200,000

12,000

3,000

7

Surface energy (J/m2)

~2

~2

~2

~2

CRACK INITIATION AND ARREST; MATERIAL HOMOGENEITY AND AE AMPLITUDE DISTRIBUTIONS In classic fracture mechanics theory, a pre-existent crack in a brittle material will start to run when the elastic stress field can supply enough energy to meet its growth needs. The energy released from the remote stress field is conceived as flowing into the region around the crack tip. The strain energy release rate is denoted by G.

76

Differently stated, the crack will start to run when the stress intensity factor (the magnitude of the stress field at the crack tip) reaches a critical value. Once the crack starts running, the next question is when and where it will stop. This will determine the energy release and hence, the amplitude of the AE. Reference [3] describes experiments at reduced temperature, with double cantilever beam steel specimens which exhibited large crack jumps. A close proportionality was observed between the distance jumped and the elastic strain energy released (determined from the load drops that accompanied the crack jumps). This close proportionality indicated the mechanisms governing crack arrest: the crack comes to rest when it runs out of energy. In the early part of the crack jump, the rate at which elastic energy is being released is faster than the rate at which it is being absorbed by plasticity along the new crack surfaces. The energy surplus is present, transiently, in the form of dynamic elastic/kinetic energy or AE in the specimen. In the late part of the crack jump, the elastic energy release rate is slower than the absorption rate. In this stage, the energy surplus helps to sustain crack growth a little longer. Along the spectrum of possible crack growth processes, brittle fracture by large crack jumps is at one end and ductile tearing is at the other. Ductile tearing comes about when the test body is under high enough load and the loading system, in delivering additional deflection, is delivering just enough energy to meet the demands of plastic deformation around the growing crack. This can be a quasi-static situation in which the crack advances slowly and steadily, with no sudden movement and no AE. If the material is not completely homogeneous, there may be local regions where small increments of crack growth can take place with less than the usual amount of energy absorption. To that extent, the crack growth can be emissive and the emissivity will be a function of the homogeneity of the material and stress field. Often in metals, a fracture surface will show both brittle and ductile indications. Cleavage may be mixed with ductile tearing. A mixed mode of this kind has been assumed as the starting point for a probability of detection (POD) model for AE that can assist the design of monitoring installations (4). This model addresses the probability of detecting a growing fatigue crack, and it assumes that a certain percentage of the fracture surface is formed by emissive cleavage, while the rest is formed by non-emissive ductile tearing. Two of the input parameters to the model are the percentage of cleavage, and the slope (b-value) of the AE amplitude distribution. The correlation between the b-value and material brittleness is well-established (5). A causal correlation between the b-value and material non-homogeneity was proposed by Mogi (6,8) and by Scholz (7,8), pioneers in the exploration of AE in geological materials as a small-scale model of earthquake processes. SEISMIC SOURCES AND THEIR THEORETICAL REPRESENTATIONS A systematic description of the energetics of acoustic emission requires an adequate, quantitative, physics model of the source event. After many years of work by many researchers, it emerged in the 1980’s that the best mathematical-physics description of the acoustic emission source was the “moment tensor”. This is a concept developed in the field of seismology to describe the magnitude and directionality of earthquake motion. Applied to acoustic emission, it has been used successfully in work on concrete, dams and building foundations [9]. The most widely cited text in the broad field of quantitative seismology is the work of Aki and Richards [10], which includes definitive discussion of the moment tensor and the double-couples that are its components. The seismic moment tensor is a mathematical representation of the movement on a fault during an earthquake. Written as a 3 x 3 matrix, it comprises nine generalized couple, or nine sets of two vectors. The tensor describes the source strength and fault orientation. The matrix is symmetrical so its independent components are six in number, rather than nine. If the movement on the fault is distributed over a substantial area it is hard to represent it. But if we limit our interest to great distances and long wavelengths, the source can be reduced to a point. The displacement at a distant point, due to a localized displacement such as a tension crack or a slippage event, can be written as un(x,t) = Mpq * Gnp,q

(2)

Here u is the displacement of the material at position x and time t. Mpq is the source moment tensor and Gnp,q is the spatial derivative of the Green’s function. (The Green’s function gives the displacement field at point x and time t resulting from a unidirectional unit impulse at a precisely localized place and time). So, equation (2) shows

77

how the movement in three dimensions, at a great distance from a seismic source, is compounded from the directional elements that make up the source (the moment tensor) together with the time- and distance- response characteristics of the propagating medium (the Green’s function). The Green’s function is a matter of acoustic wave propagation. Only a few cases are simple enough for an analytic solution to be found. To show an example, Figure 2 shows the case of the epicenter source for a halfspace, after Wadley et al. [11]. A point impulsive force is applied at a depth x below the surface at time zero, parallel to the surface (upper response) or perpendicular to the surface (lower response). The response curves show the surface displacements in directions parallel to the applied forces. In each case there is a wave arrival at the longitudinal wave speed and another at the shear wave speed. Figure 2 Epicenter Response to Impulsive Force In reference [11], Wadley et al. start with a discussion of the Green’s function, and proceed to derive the following expression for the maximum displacement u at a distance x3 from a newly forming, penny-shaped crack that opens with velocity V to radius a in an infinite solid under stress σ: u = 8(1-ν2)σa2V / 3EcLx3

(3)

Here ν is Poisson’s ratio, E is Young’s modulus and c1 is the longitudinal wavespeed. To give a physically intuitive interpretation of this equation, it can be arranged as a series of ratios: u = α .(V/cL).(σ/E).(a2/x3)

(4)

2

where α = 8(1-ν )/3, or if we take ν = 0.29 (the value for steel), u = 0.777 x3 Ω.(V/cL).(σ/E)

(5)

where Ω is the solid angle subtended by the crack area at the point of detection. These equations reveal that the displacement in the AE wave is directly proportional to the crack area, to the source velocity (normalized by the speed of sound), and to the applied stress (normalized by the elastic modulus). This is for an initiating Mode I crack; the extension of an existing crack brings up some further questions. In more complex shapes such as plates, the Green’s functions can be calculated by numerical methods. They are more elaborate in form, and we can see how the AE signal is shaped by the structure in which it propagates. Figure 3 shows the out-of-plane surface movement in response to pencil lead break at a distance of 100mm on a 60mm thick steel plate. This response is calculated by a simulation program called PlotRLQ. A 100-400kHz frequency filter has been applied. The wave motion looks much like a natural AE, although it has not yet passed through a sensor. Since this is a step force loading rather than an impulsive loading, this is actually the time integral of a Green’s function rather than the Green’s function itself.

Figure 3 Response to Step Force on a Plate

78

INTER-RELATIONSHIPS BETWEEN MODELS, ENERGY TERMS AND BRITTLENESS The moment tensor must be convolved with the spatial derivative of the Green’s function to find the displacement at a distant point: this is Equation (2). “Spatial derivative” is essential because the AE source is by nature not an unbalanced force, but a dipole - actually, a system of dipoles (couples) in x, y, z directions. The strength of the couple has the dimensions of moment per unit area. “Convolution” is essential because the moment tensor develops over time. Dimensionally, the moment tensor has the same units as energy. Aki and Richards [loc.cit. p.55] give the following equation for the seismic energy radiated away from the source region in the case of slip on a fault plane: Es = η M0 σ’13 / µ

(6)

where σ’13 is the mean of the initial and final values of the static stress component σ13. In this expression, η is called the “seismic efficiency”. Let us bring to this equation the nomenclature of the AE model shown in Figure 1. Es is evidently what we have called the “event energy”. “Seismic efficiency” must be interpreted as the ratio of event energy to source energy. It can be inferred that the “source energy” is equal to M0 σ13’/ µ. It is interesting to compare the ratio σ13’/ µ in this expression with the σ/E term in Equations (4) and (5). What determines the seismic efficiency? All indications are that it is closely related to the source speed. This is implicit in the V/c term in Equations (4) and (5). An earlier discussion of this broad relationship was already given in Reference [2] and the practical importance of source speed for the detectability of different kinds of crack growth was described in detail in Reference [12]. If two displacement steps have the same height but one takes place in a shorter time, they have the same spectral energy density at low frequencies but the shorter event has a spectrum that extends to higher frequencies, and thus carries more total energy. Figure 4 is a representation of brittle crack growth steps, on axes of stored elastic energy vs. crack length, as developed in Reference [3]. The process begins at point A where a pre-existing crack is subjected to increasing stress. Blue curves BCD, EF (and others not shown) are constraining curves on which crack movement must begin and end, such as lines of equal displacement when a compliant test specimen is loaded in a very stiff test machine. At point B the critical strain energy release rate GIc for rapid crack movement is attained. This is represented by the slope of the curve BCD at point B. As the crack starts to run, its dynamic energy requirement Gdyn is represented by the slope of the straight line from B. The crack comes to rest when the straight line again Figure 4

Stored Elastic Energy

Energy Representation of Brittle Crack Growth Steps

B

C E F

D A Crack Length

79

intersects the constraining curve on which movement started. The more brittle condition is when there is a big difference between GIc and Gdyn. In the illustration, this is the crack coming to rest at D. A less brittle condition is represented by the three crack jumps that take place on the path from B to C. Gdyn is closer to GIc. The crack jumps are shorter and more numerous. In the case Gdyn = GIc there would not be any rapid crack jump and the line on Figure 4 would be following a smooth curve instead of pursuing a series of vertical rises and diagonal drops. This would be a case of ductile tearing. The gap that opens up between the straight lines and the constraining curves in Figure 4, as the cracks propagate, represents excess energy made available during the first half of the crack propagation process. Since this is a case of unbalanced forces, much or all of this excess will take the form of acoustic energy, i.e. as acoustic emission. In the more brittle condition, represented by the path BD, the gap between the straight line and the curve is proportionately greater than in the less brittle condition, represented by the path BC. This study area is not limited to AE researchers. Indeed it is part of the much larger field of fracture research, typified by reference [13]. There are far too many models and insights to be summarized in this short paper. CONCLUSION Brittleness is closely related to the difference between GIc (the strain energy release rate at which a crack starts to move) and Gdyn (the energy required to deform material in the plastic zone as it advances during crack movement conditions). One aspect of acoustic emission energy is its appearance in the specimen during crack propagation, to the extent that GIc > Gdyn. In another aspect, acoustic emission energy can be related to the moment tensor and in particular, to the seismic efficiency. A third aspect is the way the propagation of the AE wave serves to redistribute stress (and thus elastic energy) through the stressed body as it finds its way to a new equilibrium. Brittleness has to do with mechanical stresses, stability and dynamic response to unbalanced forces. Acoustic emission deals with exactly the same things, therefore the connection is very intimate. REFERENCES 1. Standard Terminology for Nondestructive Testing, E 1316, ASTM Book of Standards, Volume 3.03, published annually. 2. Pollock, A. A., Acoustic Emission from Solids Undergoing Deformation, Ph.D. thesis, University of London, p. 202, April 1970. 3. Radon, J. C. and Pollock, A. A., Acoustic Emissions and Energy Transfer during Crack Propagation, Engineering Fracture Mechanics, Vol. 4, pp. 295-310, 1972. 4. Pollock, A. A., A POD Model for Acoustic Emission – Discussion and Status, Review of Progress in Quantitative Nondestructive Evaluation, Vol. 29B, AIP Conf. Proc. Volume 1211, pp. 1927-1933, 2009. 5. Pollock, A. A., Acoustic Emission Amplitude Distributions, International Advances in Nondestructive Testing, Ed. Warren J. McGonnagle, Vol 7, pp. 215-239, 1981. 6. Mogi, K., Study of Elastic Shocks Caused by the Fractureof Heterogeneous Materials and its Relations to Earthquake Phenomena, Bull. Earthqu. Res. Inst. Vol. 40, pp. 125-173, 1962. 7. Scholz, C. H., The Frequency-Magnitude Relation of Microfracturing in Rock and its Relation to Earthquakes, Bull. Seis. Soc. Am. Vol, 58, No. 1, pp. 399, 415, February 1968. 8. Pollock, A. A., Physical Interpretation of AE/MA Signal Processing, Proceedings, Second Conference on Acoustic Emission / Microseismic Activity in Geologic Structures and Materials, Pennsylvania State University, Edited H. R. Hardy and F. W. Leighton, Trans Tech Publications, pp. 399-422, 1980. 9. Yuyama, S., Acoustic Emission for Fracture Studies Using Moment Tensor Analysis, J. Strain Analysis, Vol. 39, No. 6, Special Issue Paper S08103, 2004. 10. Aki, K. and Richards, P.G., Quantitative Seismology, Second Edition, University Science Books, 2002. 11. Wadley, H. N. G., Scruby, C. B. et al, Acoustic Emission During Deformation and Fracture of Aluminium Alloys: Uniaxial Tests, AERE - R 10362, United Kingdom Atomic Energy Authority, 1981. 12. Acoustic Emission Testing, Volume Five, Nondestructive Testing Handbook, Second Edition, American Society for Nondestructive Testing, pp. 77-83, 1987. 13. Freund, L. B., Hutchinson, J. W. and Lam, P.S., Analysis of High-Strain-Rate Elastic-Plastic Crack Growth, Engineering Fracture Mechanics, Vol. 23, No. 1, pp. 119-129, 1986.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

b-value of plain concrete beams based on AE Quanta S. Muralidhara a*, Hamid Eskandari b, B.K.Raghu Prasad c, R. K Singh d *

Faculty, Department of Civil Engineering. BMS College of Engineering, Bangalore, Presently

research scholar, Department of Civil Engineering, Indian Institute of Science, Bangalore, India. b

Assistant Professor, Department of Civil Engineering. Sabzevar Tarbiat Moallem University,

Iran. c

Professor, Department of Civil Engineering, Indian Institute of Science, Bangalore, India.

d

Outstanding Scientist, Reactor Safety Division, BARC, Mumbai, India.

Key Words: AE Energy, Quanta, b -value. * Corresponding author. Tel.: +91 - 080 -2293 2435; Fax: +91 - 080 - 2360 0404 E-mail address: [email protected], [email protected]

Abstract In seismology, Gutenberg-Richter relationship log10 N = a − bM is an empirical relationship between the magnitude of earthquake and its recurrence frequency. The constant ‘ b ’, is called the b-value and is the log linear slope of frequency-magnitude distribution. An analogy is drawn between earthquake and failure process in concrete. During the failure process of concrete, Acoustic emission (AE) energy is released in the form of energy waves having certain peak amplitudes. While estimating the b-value during fracture in concrete, peak amplitudes of the AE signals are used. Right from the onset of micro-cracks till failure, the AE events are recorded with their peak amplitudes and AE energies. Interestingly, the AE energy release has been observed to be in “packets” or bursts. These energy packets have been called by the authors as AE quanta. In the estimation of b-value, peak amplitudes of events of groups having a definite number are used. Instead of using amplitudes of arbitrary group of events, quanta are utilized as groups for obtaining the b-value,. Unlike in seismology, wherein the b-value could be nearly unity, it is found that the b-value from quanta is much less than that obtained from the amplitudes. 1 INTRODUCTION 1.1 Guttenberg and Richter empirical relationship In seismology, seismic activity is designated by magnitude and is correlated to the energy released during the event. The relationship between seismic energy, E released during an earthquake and the magnitude M is given by the expression LogE ≈ d ⋅ M T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_11, © The Society for Experimental Mechanics, Inc. 2011

(1) 81

82

The value of d is 1 and 1.5 for small and large earthquakes respectively (Ekstrom et al.1988). Hence, there is a power-law function between N and E in the form: N ≈ E − B , where N is the incremental frequency and B = b d . The frequency of occurrence of earthquakes with larger magnitudes is less when compared with those with smaller magnitudes. Based on the frequency of occurrence, Gutenberg and Richter proposed an empirical formula which relates magnitude and frequency.

log10 N = a − bM

(2)

N = number of earthquakes with magnitude greater than M, a = seismic activity and the constant ‘b’ a damage parameter, is the log linear slope of frequency-magnitude distribution (Gutenberg et al.1944). 1.2 Acoustic emission (AE) In concrete, the fracture process zone ahead of a crack tip is the consequence of the formation of micro-cracks, a few of which later link up to form a macro-crack. An analogy is drawn between earthquake and failure process in concrete. Although the scales of damage in concrete and earthquake are different, there is a similarity in the damage process wherein elastic energy is released in the form of waves from sources located inside the medium (Carpinteri. 2006). These waves are the acoustic emission (AE) waves, which are captured by piezo-electric sensors. The captured AE data is a source of information about the damage process in concrete. The technique has been utilized to assess the damage in important concrete structures like bridges (Ohtsu et al.2002, Shigeishi et al.2001, & Colombo et. al.2005).During the failure process of concrete, stress energy is released in the form of energy waves having certain peak amplitudes. The peak amplitudes AdB (analogous to the magnitude of earthquake in seismology) are used in place of magnitudes of earthquakes in Equation 1 while estimating the b-value of concrete failure. However the peak amplitude is to be divided by a factor of 20, because of the fact that the AE peak amplitude is measured in dB, whereas the Richter magnitude of earthquake is defined in terms of the logarithm of maximum amplitude (Cox et al.1993, Hatton et al.1993, & MVMS Rao et al. 2005). The modified expression is given in Equation 3. logN = a − b( AdB / 20)

(3)

83

The b-value analysis of acoustic emissions is in general obtained by grouping the events based on either time or number to groups of events, each containing about 50 events. Researchers have (Shiotani et al 2001 & Colombo et al.2003) showed that the number of events in each group can influence the b-value. Although studies to determine b-value of concrete fracture using maximum amplitudes of AE waveforms are reported, very few are reported with AE absolute energy. In fact a preliminary study has been made on beams cast with self consolidating concrete and loaded under three point bend condition, to determine b-value using AE absolute energy instead of the usual peak amplitude (Hamid 2008). The energies and the magnitudes of earthquakes are related by the expression LogE ≈ d ⋅ M . Incorporating the above into Equation 2, the relationship between frequencies of recurrence and energy the following expression is obtained. b log10 N = a −   log10 E d 

(4)

In which the value of d is 1 for small earthquakes and 1.5 for large earthquakes. Equation 3 is made suitable in AE technique by assuming d equal to 1, since the extent of energy release in concrete fracture is very small when compared with that in earthquakes. The modified expression is

log10 N = a − b log10 E

(5)

In which b = b-value based on AE energy and E = AE energy. 1.3 Absolute AE energy Generally, micro-cracks emit waves with smaller amplitudes and waves from macrocracks have larger amplitudes (Landis 1999). However, total counts and period of the wave along with amplitudes give a clear description of AE energy. A plot of amplitude versus absolute AE energy from the AE data of a beam is shown in Fig.1. It is seen that at lower energies and higher amplitudes, the latter may not be directly proportional to the energies. It clarifies the fact that the same absolute AE energy level may show different amplitudes and vice versa.

Hence using AE energy instead of amplitude

improves the accuracy of results from the analysis. Hence the maximum amplitude of the wave signal recorded by the sensor, which is generally used in the AE analysis, cannot be considered to characterize fully the fracture in concrete. Instead, the absolute AE energy seems to be an appropriate replacement to represent an event.

84

Fig. 1: Load-time-cumulative absolute AE energy plot from D2T20UB02 beam

1.4 Peak absolute energy An event in AE studies corresponds to an internal activity due to deformations and dislocations. The location of the event is computed by AE software (AE Win SAMOS) [ p]. The sensors are located at different points on the surface of the structure. The energy recorded by each sensor varies since it depends on the distance of the sensor from the event location; closer the sensor to the event, higher the energy recorded. It is evident that, only one energy level is associated with an event, although different sensors will record different energy levels depending on their proximity to the event. Making use of this event energy for analysis, it is possible to get a more authentic scenario of the activities inside the body of the structure. There is no direct way to measure this single event energy. However an indirect way of measuring this energy is by choosing the maximum of the energies recorded by the sensors called the peak energy (Muralidhara et.al.2010), corresponding to that event, since that sensor is closest to the event location. This could also be explained by the wave attenuation as observed by researchers (Berthelot et.al. 1993). A more accurate result could be obtained if the attenuation of the energy is also considered in the analysis. It is evident from the AE event shown in Figure 2 that sensor-1 would register the maximum energy, since it is closest to the event. The relationship between the single event energy, the energy captured by the sensor and the distance of sensor from the event is given in the following equation

85

Ei ∝

E ri 2

(6)

Figure 2: Location of event from sensors

Where i =1, 2, 3 and ri the distance between the i th sensor and the event. E = energy released during that event while the Ei = energy recorded (captured) in the i th sensor. It is evident that the energies E1, E2, and E3 captured by sensors 1, 2 and 3 would be such that E1 > E2 > E3. It is obvious that E1 is the peak value of energy among the energies recorded by the sensors because it is the closest to the event and is assumed to be equivalent to the actual event energy. Hence it is sensible to make use of this peak recorded energy amongst the sensors, in the analysis. Further justification can be made by observing the plot (Fig.3) showing absolute energy of individual channel corresponding to each event, superposed with peak absolute energy. Matching to each event, only one of the channels shows maximum AE energy which is the peak energy. It is also observed that the sensor, recording the maximum energy, is not the same for all events. In literature, channel wise results are often discussed and plots from the data recorded by each channel are considered in the analysis. However it is more appropriate to discuss the AE analysis from event point of view rather than channel wise. This is the contribution from the study on AE applications to concrete fracture.

86

Figure 3: A partial plot showing absolute energy of individual channel corresponding to each event, superposed with peak energy from AE data of D2T20UB02 beam of event from sensors

1.5 AE Quanta Right from the onset of micro-cracks till failure, the AE events are recorded with their peak amplitudes and AE energies. Interestingly, the peak absolute AE energy recorded has been observed to be in “packets” or bursts. These energy packets have been named as AE quanta. It is observed that the quantum of AE energies occurs in a definite pattern and appear to be periodic. The energies start with a low value and rise to a peak in a typical quantum and that pattern repeats. One can interpret the physics of the pattern saying that the low energies are due to the formation of micro-cracks at the interface (could be even at nano level) while the high energy could be associated with complete de-bonding. From Fig.4, a plot of AE energy-time over a small time interval is shown. It could be observed that the value of the absolute AE energy rises over a time interval. Micro-crack formation records less AE energy than macro-crack formation. In other words a waveform with less energy is captured. A macro-crack is formed after coalescence of several micro-cracks. The same is seen as a record of large AE energy value after several smaller values in AE data. The smaller values are there due to disturbances at micro level. A cumulative value of these energies is seen as quanta. Each quantum of energy represents a stage in the damage process. Instead of using arbitrary group of fixed number of events with their amplitudes, AE quanta are used to determine b-value. The number of events in each quantum varies from quantum to quantum.

87

Figure 4: Plot of a typical quantum of AE energy

2 RESEARCH SIGNIFICANCE The b-value of AE data from concrete specimens is used to describe the damage process. However it is beset with the variations due to sampling size of event group. As already been pointed out the number of events in each sampling group is likely to influence the b-value. Standardization of event group size seems a requirement to bring consistency into analytical study on b-value. In this study an attempt has been made to determine the b-value from AE energy and the event group is chosen based on bursts of energy or quanta. Quanta conform to the damage stages and justify well for their use in the determination of the b-value, apparently a damage parameter.

3 EXPERIMENTAL SETUP Plain concrete single edged notched beam specimens of characteristic strength 45 MPa and with the geometrical proportion as given in Table.1 were tested under three point bending and monotonic loading conditions. The notch to depth ratio varied from 0.25 D to 0.33 D. A 500kN capacity servo controlled DARTEC machine under crack mouth opening displacement (CMOD) control was employed. The central deflection of the beam was recorded by a LVDT which could measure up to 0.1 micron. The clip gauge was used for the measurement of CMOD having a resolution of 0.1 micron. The test was performed keeping the CMOD rate at 0.0005 mm/sec. The acquisition of

88

loading and displacement parameters along with the acoustic emission data were simultaneous. Table 1: Dimensions of the beams

Type

Length, mm

Depth, mm

Width, mm

Span, mm

D1

375

95

47.5

282

D2

750

190

95

564

The AE equipment used was from Physical Acoustic Corporation, Princeton, New Jersey, USA. The AE instrument was an 8 channel with AEwin for SAMOS (sensor based Acoustic Multi-channel Operating System) E2.0 system. The AE instrument has sensors to receive the AE signals, pre-amplifiers and data acquisition system to acquire and analyze AE data. A typical AE sensor is 19 mm in diameter and 22 mm in height with a resonant frequency of 60 kHz. The threshold value was kept at 45 dB to minimize the effect of noise. Sensors were attached to the specimen surface by using vacuum grease (High vacuum silicone grease).Before applying the vacuum grease to the specimen surface at the sensor locations, the surface was gently rubbed and cleaned using acetone solution to remove dust and to ensure better bonding between sensor and the specimen. Four sensors used for the AE acquisition were arranged on one face of the specimen as shown in Figure 5. The locations of events have the origin of reference at the bottom left corner of the specimen. The sensors were initially tested for their sensitivity by pencil lead-breaking test. Further automatic sensor testing (AST) available in the AE software was employed to check the proper fixity of the sensors to the concrete surface.

Figure 5: Profile of the D2 type beam showing the position of sensors

89

All the beams of type D1 and D2 were tested in servo controlled Dartec machine under three point bend condition and CMOD control. Dartec machine was set to acquire data such as time, load, CMOD and LVDT (central deflection). The actuator of servo controlled machine was made to just touch the specimen at top. The displacement rates under CMOD control for notched and un-notched beams were chosen at 0.0005mm/sec and 0.0001mm/sec respectively. After everything is set, acquisition in both Dartec and AE instrument were started simultaneously. Acquisition was stopped when the specimen fractured fully and the load value had reduced to about 0.05kN.

4 RESULTS AND DISCUSSION At the initial and middle stages of loading history, there is continuous micro and macro-crack formations. In fact this represents the formation of fracture process zone in concrete. Correspondingly lower and higher AE energies are recorded during micro and macro-crack formations. From a thorough inspection of AE data, it is possible to identify several quanta during the entire loading history. Each quantum has smaller AE energy recordings at the initial stages and significantly large AE energies at later stages as already seen from Figure.3. The pattern could be attributed to the formation of microcracks and coalescence of these into a few macro-cracks of different sizes. The b-values were calculated selecting amplitudes from 100 events group and using Equation 2. For b-values based on peak absolute AE energy, the same set of events groups was selected and Equation 4 was adopted. The b-value results obtained from amplitudes and energies from 100 events group and quanta are tabulated in Table 2, and plotted against time as shown in Fig.6.

90

Figure 6: A combined plot of load-time- b-value from quanta, energy from a group of 100 events and amplitude from a group of 100 events.

5 CONCLUSIONS The b-values calculated based on event energy are less than those calculated with event amplitudes. Quanta based b-values are found to be less than those from amplitudes and energy from an arbitrarily chosen (100events in the present study) event group. They also show more fluctuations during the initial and middle portion of loading stages. a decrease in b-value is seen due to material damage (micro-cracking and macro-cracking) while b-value show a rising trend due to toughening mechanisms like aggregate interlocking, tortuosity of crack path etc. however amplitude based bvalues calculated using groups of 100 events show not much of activity during the same period. In other words they portray fewer activities which are not true. The least b-value is as low as 0.15 while the maximum is as high as 0.47. Although there is an analogy of seismic activity and concrete fracture, the range of b-values are different.

91

Table.2. Details of b-value based on AE amplitudes from 100 event group and AE energy from 100 event group and AE quanta. b-value Time

b-value Time

AE amplitudes interval

interval from group of

(Sec)

Time

b-value

interval

Based on

(Sec)

Quanta

AE energies from group of (Sec)

100 events

100 events

0-209

0.89

0-209

0.41

0-167

0.3

209-290

0.72

209-290

0.54

167-189

0.33

290-407

0.85

290-407

0.48

189-229

0.47

407-522

0.76

407-522

0.57

229-263

0.36

522-779

0.49

522-779

0.41

263-282

0.15

779-1333

0.44

779-1333

0.55

282-290

0.23

1333-2720

0.72

1333-2720

0.52

290-325

0.33

325-386

0.27

386-403

0.2

403-428

0.32

428-461

0.38

461-503

0.32

503-601

0.2

601-646

0.13

646-745

0.15

745-837

0.31

837-1010

0.15

1010-1370

0.3

1370-1668

0.19

1668-2720

0.3

92

6 REFERENCES

Berthelot, J. M. Souda, M. B. and Robert, J. (1993). “Study of wave attenuation in concrete”, J. Mater. Res., vol. 8, no. 9, pp. 2344–2353. Carpinteri, A. (2006), “Critical Behavior in Concrete Structures and Damage Localization by Acoustic Emission” Key Engineering Materials, Vol. 312, pp. 305310. Colombo, S. Main, I. G. and Forde, M. C. (2003). “Assessing damage of reinforced concrete beam using ’b-value’ analysis of acoustic emission signals”, ASCE, J Mater Civil Eng, ASCE, vol. 15, no. 3,pp. 280–286. Colombo,S., Forde, M.C., Main, I.G., and Shigeishi,M. (2005), “Predicting the ultimate bending capacity of concrete beams from the “relaxation ratio” analysis of AE signals”, Construction and Building Materials, vol. 19, pp.746-754. Cox, S. J. D. and Meredith, P. G. (1993), “Microcrack formation and material softening in rock measured by monitoring acoustic emission”. Int. J. Rock Mech. Min. Sci. Geomech. Abstr., vol. 30, pp.11–21. Ekstrom, G. and Dziewonski, A. M. (1988). “Evidence of bais in estimations of earthquake size”, Nature, vol. 332, pp. 319–323. Gutenberg, B. and Richter, C. F. (1944). “Frequency of earthquakes in california”, Bull. Seism. Soc.Am., vol. 34, pp. 185–188. Hamid Eskandari naddaf. (2008). “Fracture characteristics of self consolidating concrete (Prediction of strength and monitoring of fracture)” PhD thesis, Indian institute of science, Bangalore,India, p. 161. Hatton, C. G., Main, I. G. and Meredith P. G. (1993). “A comparison of seismic and structural measurements of scaling exponents during tensile sub-critical crack growth”. J. Struct. Geol., vol.15, pp. 1485–1495. Landis, E. N. (1999). “Micro-macro fracture relationships and acoustic emissions in concrete”, Cons Build. mater., vol. 13, no. 1-2, pp. 65–72. Muralidhara, S., Raghu Prasad, B. K., Hamid Eskandari, Karihaloo, B. L. (2010). “Fracture process zone size and true fracture energy of concrete using acoustic emission”, Cons Build. Mater. Vol. 24, no. 4, pp. 479-486.

93

Ohtsu,M., Uchida,M., Okamoto,T., and Yuyama,S. (2002). “Damage Assessment of Reinforced Concrete Beams Qualified by Acoustic Emission”, ACI Struct Journal, pp. 411-417. Rao, M. V. M. S., and Prasanna Lakshmi, K. J. (2005). “Analysis of b-value and improved b-value of acoustic emissions accompanying rock fracture”, Current Science, vol. 89, no. 9, pp. 1577-1582. Shigeishi, M., Colombo, S., Broughton, K.J., Rutledge, H., Batchelor, A.J.,and Forde, M.C. (2001). “Acoustic emission to assess and monitor the integrity of bridges”, Construction and Building Materials, vol.15, pp. 35-49. Shiotani, T., Yuyuma, S., Li, Z. and Ohtsu, M. (2001). “Application of AE improved bvalue to quantitative evaluation of fracture process in concrete materials”. Journal of Acoustic Emission, vol.18, pp.118–133.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA 95 ©2010 Society for Experimental Mechanics Inc.

AE monitoring of the Syracuse Athena Temple: Scale invariance in the timing of ruptures G. Niccolini1, G. Durin1, G. Lacidogna2, A. Manuello2, A. Carpinteri2 1

National Institute of Metrological Research - INRiM Strada delle Cacce 91 – 10135 Torino, Italy E-mail: [email protected]

2

Politecnico di Torino, Department of Structural Engineering & Geotechnics, Corso Duca degli Abruzzi 24 – 10129 Torino, Italy

ABSTRACT We present a comparative statistical analysis between the time series of the acoustic emission (AE) events detected from the ancient Greek Athena temple in Syracuse (Eastern Sicily, UNESCO World Heritage List since 2005), and the time series of small and intermediate earthquakes occurred in this part of Sicily during the AE monitoring period. The waiting-time distributions for both time series and different magnitude thresholds are described by a unique scaling function indicating self-similarity over a wide range of magnitude scales. The similarity between the waiting times in AEs and earthquakes suggests a correlation between the microfracturing process accompanying ageing and deterioration of the monument and the regional seismicity. Our results reveal that the structure of the Athena temple is particularly sensitive to the normal seismic activity, suggesting structural AE monitoring as a useful tool for seismic hazard assessment. INTRODUCTION Fracture in heterogeneous materials occurs as the culmination of progressive damage due to loading conditions or harsh environments. In particular, the macroscopic failure of rocks and concrete specimens is preceded by the spontaneous release of elastic energy due to microcrack growth in the form of transient elastic waves (acoustic emission), which thus play the role of fracture precursors [1-4]. Therefore, as it provides information on the internal state of a material, the AE monitoring successfully finds applications in civil engineering, where it is used to evaluate the integrity of large-sized structures (buildings, bridges, etc) before they become safety hazards [5-8]. Thus, the predictive power of AE monitoring can be exploited in the seismic areas of Italian territory, where a number of ancient buildings and monuments are exposed to harsh environmental conditions. In particular, many structures may undergo accelerated ageing and deterioration due to small and intermediate earthquakes, rather numerous in that areas. This damage, often inaccessible for visual inspection, results in increased vulnerability to stronger earthquakes. In this framework, we present the results of an AE monitoring of the Athena temple in Syracuse, incorporated in the walls of the Cathedral (on the UNESCO World Heritage List since 2005). The statistical properties of the AE time series from one pillar of the temple and the Eastern Sicily earthquake time series are studied, in order to investigate a possible correlation between AE signals and local earthquakes.

AE MONITORING OF THE SYRACUSE CATHEDRAL In the seventh century the Cathedral of Syracuse (Fig. 1(a) and (b)) was built onto the ruins of the ancient and famed Athena temple (5th century B.C.), and afterward repeatedly modified in consequence of damages caused by earthquakes. The columns of the temple remained visible, incorporated in the walls of the Cathedral. At present, the structure exhibits an extended damage pattern, especially in four pillars at the end of the nave, which show several repaired areas, replacements, and also several cracks [9]. The peculiarity of the Cathedral pillars is that they had been obtained cutting out the stonework walls of the internal cell of the Athena temple. T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_12, © The Society for Experimental Mechanics, Inc. 2011

95

96

(a)

(b)

Figure 1: Facade (a) and lateral view (b) of the Cathedral of Syracuse.

(b)

(c)

(a)

(d)

Figure 2: Layout of the Cathedral of Syracuse showing the location of the pillar subjected to AE monitoring (a); axonometric view of the monitored pillar (b); AE transducers applied to the pillar (c); four lateral sides of the pillar with cracking pattern and located AE sources (black points). The AE activity in one of the nave pillars, shown in Fig. 2(a) and (b), was monitored over a four-month period T (September 2006-January 2007). This element (save for a few strengthening works performed during a restoration process in 1926) is entirely made of calcareous stone blocks, probably installed during the initial construction of the temple. The AE waves were detected by attaching to the pillar surface six PZT transducers (Fig. 2(c)) working in the range of 50 to 500 kHz, that converted the surface vibrations into an electric signal (AE signal, expressed in volts). Since the frequency response of the transducers in the considered range is sufficiently flat, the AE signal amplitudes can be assumed to be proportional to the acceleration at the specimen surface over the considered frequency range, and thus be used to quantify the magnitude of AE events, which is defined as Log(A / 1µV), where A is the AE signal amplitude [5-7]. After setting an appropriate detection threshold (fixed at 100 µ V) to filter out the acoustic noise due to human activities and spurious signals, we started the AE data acquisition storing a set of parameters, i.e., the arrival time and the amplitude of each detected signal [10]. Extracting few parameters from the signals is a less storage-consuming technique than recording the whole waveforms and allows at the same time to follow the evolution of damage. The AE sources were located by the use of multiple transducers, exploiting the arrival time information of received signals. The AE sources and the cracking pattern of the monitored pillar are illustrated in Fig. 2(d). It can be noted that the AE sources accumulated near the most visible crack paths, revealing that the pillar undergoes slow but progressive damage.

97 The accumulated number of AE events as a function of time is displayed in Fig. 3(a) together with the earthquakes (magnitude ≥ 1.2) whose epicenter was within 50 km far away from Syracuse (Fig. 3(b)), occurred during the monitoring period of the Cathedral. An interesting correlation between the AE activity detected on the Cathedral and the seismic activity appears, where the main peaks of AE released energy, measured by means of the AE count rate, are observed in connection with seismic events.

3000

3

2000

2

1000

1

0

0

200 150 100 50

AE count rate (events/hour)

4

4000

Magnitude

Accumulated number of AE events

5000 7/12/06 5/12/06 14/01/07 14/10/06

22/09/06

50 km

14/10/06 9/01/07

11/12/06

19/09/06

16/10/06

1000 2000 3000 Monitoring time (hours)

(a)

(b)

Figure 3: Cumulative and differential AE event count in the pillar as a function of time and seismic events (plotted by triangles) (a), occurred during the AE monitoring period within the encircled area in the map (b). SCALING LAW FOR AE AND EARTHQUAKE WAITING-TIME DISTRIBUTIONS The space-time organization of AE and earthquake source process is ruled by different power laws, among them, the Gutenberg-Richter (GR) law for the magnitude distribution [11]; the Omori law for the rate of aftershocks as a function of time from the main shock [12]; and the fractal distribution law for the epicentres [13,14]. Such power-law scaling led statistical physicists to view damage phenomena and seismicity as an intermittent flow of energy within a critical system [15], where fracture energy and ultimate strength play the role of critical parameters. However, power laws and critical exponents are not the only scaling predictions for fracture systems. Equations of state in the form of scaling laws can be established as well, where the fulfilment of scaling laws is illustrated by data collapse [16]. From an analysis of the probability density functions (PDF) of waiting times between earthquakes in a hierarchy of spatial domain sizes and magnitudes in southern California, Bak et al. [17] discussed a unified scaling law in which the PDF of waiting times, the GR law, the Omori law, and the fractal distribution law were combined in a single framework. The point of view of Bak et al. was later refined and extended to many different regions of the Earth by Corral [18], and further applied to laboratory fracture [19-21]. Following this approach, we choose a minimum magnitude Mth for a given time series of events, in such a way that only events with magnitude M ≥ Mth are taken into account. All the N(Mth) surviving events define a point process in time, or time series {ti}, where we name waiting times the time intervals between consecutive events, defined as τi ≡ ti – ti –1. For several values of Mth, the waiting-time PDFs for a given time series of events are defined as pMth(τ) ≡ Prob(τ ≤ waiting time < τ + dτ )/ dτ. All the PDFs can be described by a unique distribution if they are rescaled by their rates; in other words, performing the transformation τ → τ / < τ >Mth, where < τ >Mth = T / (N(Mth) – 1) is the mean waiting time, all the rescaled PDFs collapse onto a single curve f, thus illustrating the existence of a scaling law [18-21]: pMth(τ) = f(τ/ < τ >Mth) / < τ >Mth

(1)

The scaling law (1) expresses self-similarity of the waiting time PDFs, since they can be obtained from each other by a similarity transformation. The scaling function f can be well approximated by a generalized gamma function, which is a common parameterization for all fracture systems from microscopic scale (AEs) to geological scale (earthquakes) [18-21]: f(θ ) ∝ θ

– (1 – γ)

exp[(–θ / x) ] n

(2)

98 where θ ≡ τ / < τ >Mth is the rescaled waiting time. The values of fitting parameters γ, x, and n generally depend on the considered fracture system and the window of observation; in particular, the power-law exponent 1– γ indicates the clustering degree in time of events. Here, that is observable during the AE monitoring of a Syracuse limestone cylindrical specimen (base diameter 120 mm, height 120 mm) subjected to uniaxial compression up to failure at constant piston –1 velocity of 10−4 mm s (displacement control). The test was performed in laboratory using a servo-hydraulic press with a maximum capacity of 500 kN shown in Fig. 4(a). Note that the loading condition, realized controlling the displacement, made it possible to observe the descending part of the stress vs. time curve beyond the peak load (softening branch). After fixing at 200 µ V the detection threshold for the AE signals, we verified that no spurious signals were detected before the beginning of the test. 4×103

10

3×10 2×10

3

8

AE

6 4

1×103 0

Stress (MPa)

NAE(t)

Load 3

2 0

1×104

2×104 t (s)

(a)

3×104

4×104

0

(b)

Figure 4: Servo-hydraulic press and Syracuse limestone specimen with applied transducers (a), applied load and accumulated number of AE events as a function of time: note the dense AE sequence announcing the specimen failure, revealed by the load drop subsequent the peak marked by the dashed line (b). 102 f (τ / < τ >Mth) (dimensionless)

f (τ / < τ >Mth) (dimensionless)

103

101

10–1

10–3 10–5

10–3 10–1 τ /< τ >Mth (dimensionless)

(a)

101

100

10–2

10–4 10–3

10–1 101 τ /< τ >Mth (dimensionless)

(b)

Figure 5: Rescaled waiting-time PDFs for the period preceding the peak load (a), and for the whole duration test (b), for several Vth values, equivalent to Mth = Log(Vth / 1µV). The data collapse illustrates the fulfilment of a scaling law, where the solid line is the scaling function (2). In Fig. 4(b) the AE time series is shown, where the high AE count rate during the stage before the peak load suggests coalescence of microcracks to form the main fracture. We calculate the waiting-time PDFs pMth(τ) from the first part of the AE time series, up to the peak load, and also from the complete time series. The corresponding rescaled PDFs in Fig. 5 illustrate data collapse and therefore the validity of the scaling law in (1), although the scaling function f for the sub period up to peak load (the fit of the gamma distribution in (2) yields the parameters γ = 0.23 ± 0.02, x = 2.23 ± 1.00, n = 1.15 ± 0.16) is clearly different from the one

99 for the whole test duration (γ = 0.46 ± 0.04, x = 5.61 ± 3.15, n = 0.99 ± 0.14). In particular, the power law is steeper (1– γ ≈ 0.77 against 0.54) indicating that the clustering degree is higher in this case (compare Figs. 5(a) and (b)), where the dense sequence leading to specimen failure dominates (Fig. 4(b)). Following the same approach, we investigate the existence of a possible correlation between the AE time series from the pillar and the earthquake time series extended to whole Eastern Sicily recorded during the period T, by comparing the relative PDFs of waiting times (Fig 7(a)). Note that now we select a larger region to contain enough seismic data to be statistically significant for this analysis (compare Figs. 3(b) and 6). The remarkable finding is that the scaling function f is found the same for both fracture phenomena (Fig. 7(b)). In other words, we would generally expect that the two families of rescaled distributions collapsed onto two distinct curves, whereas we observe the collapse of the two families of PDFs onto a unique curve (the fit yields γ = 0.35 ± 0.02, x = 1.15 ± 0.42, n = 0.53 ± 0.04), which can be interpreted as a signature of nontrivial correlation between AE activity on the pillar and earthquakes. Therefore, this evidence reinforces the idea previously proposed that the micro cracking process in the pillar may be triggered by earth’s trembles, thus revealing that the structure of the Athena temple is particularly sensitive to the seismic activity of the region.

Figure 6: Map of Sicily showing the AE monitoring site (yellow square) and the strongest earthquakes (red circles marked with their date and magnitude) occurred in Eastern Sicily during the monitoring period.

10–1

104

f (τ / < τ >Mth) (dimensionless)

p Mth (τ) (s–1)

10–3

10–5

10–7

10–9 100

102 100 10–2 10–4 10–6

10

2

τ (s)

(a)

10

4

10

6

10–5

10–3 10–1 101 τ /< τ >Mth (dimensionless)

(b)

Figure 7: Waiting-time PDFs for five seismic magnitude thresholds Mth and four AE signal thresholds Vth (a), and their collapse onto a single curve illustrating the fulfilment of a unique scaling law (b).

100 CONCLUSIONS From a complex-system prospective, the network of micro fractures in the Cathedral and earthquakes in the territory can be regarded as a single fracture system, where peaks in the AE released energy are correlated with earthquake occurrence. The suggested correlation between AEs and earthquakes is reinforced by the existence of a unique scaling law governing the timing of fracture events at all the magnitude scales (selfsimilarity from the AE scale to the seismic scale). In conclusion, the presented study suggests that the AE structural monitoring coupled with the analysis of the regional seismic activity can be a tool of crucial importance in seismic hazard assessment and mitigation. AKNOWLEDGMENTS The financial support provided by the Regione Piemonte (Italy) RE-FRESCOS Project, is gratefully acknowledged. REFERENCES 1. Scholz, C.H. (1968) Microfracturing and the inelastic deformation of rock in compression. Journal of Geophysical Research. 73(4), 1417–1432. 2. Lockner, D. (1993) The role of acoustic emissions in the study of rock fracture. Int. J. Rock Mech. Min. Sci. Geomech. Abs. 7, 883–889. 3. Guarino, A., Garcimartin, A. and Ciliberto, S. (1998) An experimental test of the critical behaviour of fracture precursors. Eur. Phys. J. B. 6, 13–24. 4. Turcotte, D. L., Newman, W. I. and Shcherbakov, R. (2003) Micro and macroscopic models of rock fracture. Geophys. J. Int. 152, 718–728. 5. Colombo, S., Main, I., G. and Forde, M. C. (2003) Assessing damage of Reinforced Concrete Beam using ‘‘b-value’’ Analysis of Acoustic Emission signals. J. Mat. Civil Eng. (ASCE). 15, 280–286. 6. Carpinteri, A., Lacidogna, G. and Niccolini, G. (2007) Acoustic emission monitoring of medieval towers considered as sensitive earthquake receptors. Natural Hazards and Earth System Sciences. 7, 1–11. 7. Carpinteri, A., Lacidogna, G., and Puzzi, S. (2009) From criticality to final collapse: evolution of the “b-value” from 1.5 to 1.0. Chaos, Solitons & Fractals. 41(2), 843–853. 8. Carpinteri, A., Lacidogna, G., Niccolini, G. and Puzzi, S. (2008) Critical defect size distributions in concrete structures detected by the Acoustic Emission technique. Meccanica. 43(3), 349–363. 9. Binda, L., Cantini, L., Condoleo, P., Saisi, A. and Zanzi, L. Investigation on the pillars of the Syracuse Cathedral in Sicily, in 3-Day International Conference on Structural Faults & Repair, M. C. Forde, Eds., Eng. Technics Press, Edinburgh, 2006, CD-ROM, 1-12. 10. Carpinteri, A., Lacidogna, G. and Manuello, A. (2009) The b-value analysis for the stability investigation of the ancient Athena Temple in Syracuse. Strain. Doi: 10.1111/j.14751305.2008.00602.x. 11. Richter, C.F. (1958) Elementary Seismology (W.H. Freeman, San Francisco and London, San Francisco). 12. Utsu, T., Ogata, Y. and Matsu’ura, S. (1995) The centenary of the Omori formula for a decay law of aftershock activity. J. Phys. Earth. 43, 1–33. 13. Weiss, J. (2003) Scaling of fracture and faulting of ice on earth. Surveys in Geophysics. 24(2), 185– 227. 14. Carpinteri, A., Lacidogna and G., and Niccolini, G. (2009) Fractal analysis of damage detected in concrete structural elements under loading. Chaos, Solitons & Fractals. 42, 2047–2056. 15. Sornette, A. and Sornette, D. (1989) Self-organized criticality and earthquakes. Europhys. Lett. 9(3), 197–202. 16. Stanley, H.E. (1999) Scaling, renormalization and universality: three pillars of modern critical phenomena. Reviews of Modern Physics. 71, 358–366.

101 17. Bak, P., Christensen, K., Danon, L. and Scanlon, T. (2002) Unified scaling law for earthquakes. Phys. Rev. Lett. 88 178501. 18. Corral, A. (2006) Statistical Features of Earthquake Temporal Occurrence. Modelling Critical and Catastrophic Phenomena in Geoscience (Springer Lecture Notes in Physics 705) ed P Bhattacharya P et al (Berlin: Springer). 19. Davidsen, J., Stanchits, S. and Dresen, G. (2007) Scaling and universality in rock fracture. Phys. Rev. Lett. 98 125502. 20. Niccolini, G., Durin, G., Carpinteri, A., Lacidogna, G. and Manuello, A. (2009) Crackling Noise and Universality in Fracture Systems. J. Stat. Mech.: Theory and Experiment. Doi: 10.188/17425468/2009/01/P01023. 21. Niccolini, G., Bosia, F., Carpinteri, A., Lacidogna, G. and Manuello, A. (2009) Self-similarity of waiting times in fracture systems. Physical Review E. 80 026101.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Analysis of energy released by elastic emission in brittle materials under compression A. Schiavi1, G. Niccolini1, P. Tarizzo1, G. Lacidogna2, A. Manuello2, A. Carpinteri2 1

National Institute of Metrological Research - INRiM Strada delle Cacce 91 – 10135 Torino, Italy E-mail: [email protected]

2

Politecnico di Torino, Department of Structural Engineering & Geotechnics, Corso Duca degli Abruzzi 24 – 10129 Torino, Italy

ABSTRACT Experimental results are presented for fracture tests carried out on concrete and rock specimens under compression, and an analysis is performed for low-frequency acoustic emissions (elastic emissions, or ELE) due to crack growth. ELEs are vibrations of the specimen surface with relevant amplitudes and low frequencies (between 1 kHz and 20 kHz), appearing at the very last stages of the test and then revealing imminent failure. A spectral analysis of the ELEs is performed by measuring with a calibrated transducer the local acceleration of the specimen surface in the application point of the transducer. Quantitative estimation of ELE released energy is given in terms of kinetic energy using a simple kinematic model. INTRODUCTION Crack growth in a elastic stressed-strained solid causes the release of stored strain energy in the form of transient elastic waves (acoustic emission, or AE) detectable by means of piezoelectric transducers attached on the external surface of the solid. Revealing the presence of evolving damage, AE plays the role of fracture precursor [1-12]. AE can be detected over a wide frequency range, from few Hz to MHz. Recently, the authors proposed a distinction between high-frequency AE and low-frequency AE, the latter called elastic emission or ELE [13]. Such distinction simply emphasizes that these two kinds of emissions are traced back to different stages of damage. Elastic emissions are defined as the acoustic emissions with wavelength λELE greater than the maximum size dMAX of the tested specimen, i.e. with wavelength greater than the medium of propagation. This is a remarkable issue, since an ELE event would imply a displacement of the entire body (rigid mode), while high-frequency AE are oscillations of the medium particles around their equilibrium position due to microcrack growth. Each ELE event is generated from fractures acting as shocks able to displace the solid with low-frequency vibrations. That is confirmed by numerous investigations, which establish that the AE activity varies during the damage process [13-15]; as soon as microcracks coalesce to form cracks of large size, ELE clearly appear in addition to high-frequency AE. Eventually, the frequencies drop off to a single Hz, producing the audible sound of the solid being fractured. The presence of ELE, which accompany actual matter displacements, i.e., formation of large fractures and even plastic deformations, is thus the signature of irreversible damage leading to the final specimen collapse. It is noteworthy that simultaneous detection of ELE and Electromagnetic Emission (EME) is often observed [16]. This study focuses on the quantitative estimation of the ELE released energy by inserting the measurements of surface acceleration, performed with a calibrated transducer, into a simple kinematic model describing the velocity profile of the specimen surface. EXPERIMENTAL SET-UP The solids under investigation are two concrete specimens (cubic and cylindrical) and two Green Luserna granite cylindrical specimens subjected to uniaxial compression up to failure. The tests are performed in T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_13, © The Society for Experimental Mechanics, Inc. 2011

103

104 displacement control at constant piston velocity of 10−3 mm s , using a servo-hydraulic press with a maximum capacity of 500 kN, as shown in Figure 1. The specimens adhere to the press platens without any coupling material (specimen-platen contact with friction). A piezoelectric accelerometer, working in the range of 1 to 20 kHz, is attached to the specimen surface for detection of ELE events, which are characterized by –2 –2 the output response of the calibrated transducer (charge sensitivity 9.20 pC/m s ), expressed in mm s . In order to filter out environmental background noise, we set an appropriate detection threshold for the signal –2 acquisition, fixed at 40 dB (referred to 1 µm s ). In this way, we verify that no spurious signals are detected before the beginning of the test. Furthermore, we proceed to the identification and quantification of the mechanical noise of the press in the low-frequency range during the test. Therefore, we measure the propagation velocity cL of the longitudinal elastic waves in the tested materials in order to verify that the wavelengths λELE of detected signals fulfil the relationship, λELE ≥ dMAX, which defines the condition for the propagation of ELEs in a given specimen. Equivalently, such relationship defines an upper limit for the ELE frequencies defined as: –1

f ELE ≤

cL

(1)

d MAX

As shown in Table 1, the physical quantities of interest (specimen size, elastic wave velocity) indicate that the adopted transducer, working in the range of 1 to 20 kHz, properly detects ELE events. Table 1: Technical features of tested materials Specimen

Material

Density 3 [kg/m ]

Mass [kg]

Thickness [m]

1 2 3 4

Concrete Concrete (cubic) Luserna gran. Luserna gran.

2500 2200 2480 2480

0.146 2.200 0.580 0.245

2.6·10 -2 1·10 -2 10.6·10 -2 5.3·10

-2

Base diameter [m] -2 5.3·10 -2 5.3·10 -2 5.3·10

Elastic waves speed [m/s] 4270 ~ 4500 2950 2950

fELE [kHz] 164.2 45.0 27.8 55.7

Figure 1: Servo-hydraulic press and Luserna granite specimen with applied transducers (a), and damaged specimen after the test (b). EXPEIMENTAL RESULTS The mechanical response of concrete specimen 1 is quasi-brittle [17-19], i.e., with accumulation of damage revealed by the decreasing slope of the curve in the load vs. time (or displacement, since the test is in displacement control) in diagram of Figure 2 (a). The loading condition makes it possible to observe the softening branch (the descending part of the curve beyond the peak load), which is here characterised by a sudden drop in load-carrying capacity suggesting the propagation of the main fracture. The time histories of applied load and ELE signals are represented in Figure 2 (a). The ELE signals are the sharp peaks which stand out against the background noise due to the mechanical vibrations of the press. Excluding the initial transient (0-200s) of the ELE activity occurring when the press platen is brought into contact with the end

105 specimen surface, relevant ELE activity is observed in the stage preceding the specimen failure (600s-peak load time). This section of the ELE activity is crucial, since it plays the role of fracture precursor and is traced back to the growth of macrocracks. Such role of ELE in this stage is confirmed by the resulting accelerated strain and the deviation from linear elasticity before the failure. The ELE activity during the post-failure stage, though intense, is probably less important, as it simply describes opening or friction between the existing fracture surfaces. It is noteworthy a decrease in the background noise, which is due to the reduced load exerted by the press. The mechanical response of concrete specimen 2 during the compression test is linear until failure (perfectly brittle behaviour [17-19]), which thus occurs abruptly without any apparent accumulation of damage, as shown in Figure 2(b); however, the ELE activity reveals the existence of a growing damage process (2000s – failure time), even if it is not revealed by the specimen stiffness loss in the load vs. time diagrams. During the last seconds of the specimen life, an abrupt increase in the ELE rate reveals that the final stage of damage process is dominated by the propagation of large cracks leading to the specimen failure.

500

L o ad [kN ]

400 300 200 100 0 0

500

1000

1500

2000

2500

3000

T ime [s]

(a)

(b)

Figure 2 (a) and (b): Time history of ELE signals (upper) and applied load (lower) for concrete specimen 1 (a), and concrete specimen 2 (b). Note the mechanical response of specimen 1 (quasi-brittle) and specimen 2 (perfectly-brittle). The Luserna granite specimens 3 and 4 behave quite differently, being characterized by some load drops before the final drop which describes a catastrophic failure with negligible residual strength, as shown in Figure 3 (a) and (b). Excluding the initial transient, clusters of ELEs are detected in the stage preceding the first load drop (900-1100 s in specimen 3, and 1600-2100 s in specimen 4), which reveals the onset of serious damage, and during the stage between the first and the final load drop (1200-1400 s in specimen 3), which announces the impending failure.

(a)

(b)

Figure 3: Time history of ELE signals (upper) and applied load (lower) for Luserna granite specimen 3 (a) and 4 (b).

106 The ELE signals are acquired storing the complete waveforms, and then they are subjected to spectral analysis and plotted in time-frequency-amplitude coordinates; Figures 4 (a) and (b) respectively represent two examples of ELE spectrum of concrete (specimen 1) and Luserna granite (specimen 3).

(a)

(b)

Figure 4: Spectral analysis of ELE activity, in terms of vibration acceleration level between 1 kHz and 20 kHz, in two time windows of damage process in concrete (a) and Luserna granite (b). ESTIMATION OF ELE RELEASED ENERGY BY A KINEMATIC MODEL We assume that the fracture mode causing an ELE event is of explosive type, i.e. crack opening; in particular, we use a model of a horizontal penny-shaped growing crack. As the specimen is clamped between the press platens, such an explosive source causes a radial dilatation. We calculate the kinetic energy associated with this rapid volume change in the specimen. The vertical axis of the cylindrical specimen is the axis of symmetry, with the origin at the crack centre, and z is the vertical coordinate. Because of the assumed radial symmetry, the velocity field at the specimen surface depends on z only, v = v(z). We think of the specimen as the superposition of layers with thickness dz and mass dm = M dz/h, being M the specimen mass (Fig. 5(a)). First, we calculate the contribution of a generic layer to the kinetic energy; let us imagine the layer decomposed into sectors, all moving outwards from the source crack with the same velocity v because of the assumed symmetry (Fig. 5(b)). Therefore, the 2 kinetic energy of the layer is ½dmv , obtained by summing the translational kinetic energy of all sectors. The total kinetic energy of the specimen, emitted as ELE wave, is given by summing over all layers:

E kin = ∫M

1 2 M h2 2 v ( z )dm = ∫ v ( z )dz 2 2h −h 2

(1)

Since the wavelength associated with the motion is greater than the height h of the specimen, we can assume a parabolic-like profile for the velocity field v(z), as shown in Fig. 5(c); as boundary conditions, we take zero velocity at the ends of the specimen and maximum (measured) velocity v0 in the middle, where the accelerometric transducer is applied, such that v(z) is determined by:

 v( z ) = − az 2 + bz + c  v( 0 ) = v0  v( −h / 2 ) = v( h / 2 ) = 0 

(2)

Inserting (2) into (1) gives the following expression for the kinetic energy:

E kin =

4 Mv02 15

(3)

107

F

z v dV

v

dV dm

(a)

(b)

(c)

Figure 5: Decomposition of the specimen into layers of volume dV (a), cross-section of the layer containing a horizontal circular growing crack: all sectors (one is highlighted) move outwards from this explosive source (b), causing radial dilatation of the specimen (c). We use (3) to calculate the elastic energy released as ELE waves for fracture events in concrete and Luserna granite specimens at initial and late stages of damage process, exploiting the measured values for –11 –9 surface velocity; the calculated energies for concrete specimens range between 1.0×10 J and 1.2×10 J, –11 –8 while for Luserna granite specimens range between 1.6×10 J and 2.2×10 J. However, since most of the elastic energy released during crack growth is dissipated as heat, the next step, that is establishing an energy balance, remains an open question and a challenging task. CONCLUSIONS In this work the damage process occurring in concrete and Luserna granite specimens under compression is studied by means of low-frequency elastic wave propagation induced by meso and macrocrack growth. In particular, elastic emissions (ELEs) in the frequency range 1-20 kHz are detected, analyzed and quantified in terms of kinetic energy. ELEs appear announcing significant load drops and in particular specimen failure, even if not revealed by the stiffness loss in the load vs. displacement diagrams. This evidence could be of importance in terms of testing materials since irreversible damage (i.e. macrocracks and plastic deformations) preceding structural failure can be clearly detected. ACKNOWLEDGEMENTS The financial support provided by “Regione Piemonte” Re-Frescos project, is gratefully acknowledged. REFERENCES 1. Scholz, C.H. (1968) Microfracturing and the inelastic deformation of rock in compression. Journal of Geophysical Research. 73(4), 1417–1432. 2. Ohnaka, M. and Mogi, K. (1982) Frequency characteristics of acoustic emissions in rocks under uniaxial compression and its relation to the fracturing process to failure. Journal of Geophysical Research. 87 (B5), 3873–3884. 3. Khair, A.W. (1984) Acoustic emission pattern: an indicator of mode of failure in geologic materials as affected by their natural imperfections. In Proceedings of the 3rd Conference on Acoustic Emission/ Microseismic Activity in Geologic Structures and Materials, 1981, University Park, Pa. Edited by H.R. Hardy, Jr., and F.W. Leighton. Trans Tech Publications, Clausthal, Germany, 45–66. 4. Lockner, D. A., Byerlee, J. D., Kuksenko, V., Ponomarev, A. and Sidorin, A. (1991) Quasi-static fault growth and shear fracture energy in granite. Nature. 350, 39–42.

108 5. Lockner, D. (1993) The role of acoustic emissions in the study of rock fracture. Int. J. Rock Mech. Min. Sci. Geomech. Abs. 7, 883–889. 6. Landis, E. N. and Shah, S.P. (1995) Frequency-dependent stress wave attenuation in cement-based materials. Journal of Eng. Mech. 121, 737–743. 7. Ohtsu, M. (1996) The history and development of acoustic emission in concrete engineering. Magazine of Concrete Research. 48, 321–330. 8. Eberhardt, E., Stead, D., Stimpson, B. and Read, R.S. (1998) Identifying crack initiation and propagation thresholds in brittle rock. Journal of Can. Geotech. 35, 222–233. 9. Colombo, S., Main, I., G. and Forde, M. C. (2003) Assessing damage of Reinforced Concrete Beam using ‘‘b-value’’ Analysis of Acoustic Emission signals. J. Mat. Civil Eng. (ASCE). 15, 280–286. 10. Rao, M.V.M.S. and Lakschmi, P.K.J. (2005) Analysis of b-value and improved b-value of acoustic emissions accompanying rock fracture. Current Science-Bangalore. 89, 1577–1582. 11. Anzani, A., Binda, L., Carpinteri, A., Lacidogna, G. and Manuello A. (2008) Evaluation of the repair on multiple leaf stone masonry by acoustic emission. Materials and Structures (RILEM). 41, 1169– 1189. 12. Carpinteri, A., Lacidogna, G., Niccolini, G. and Puzzi, S. (2008) Critical defect size distributions in concrete structures detected by the Acoustic Emission technique. Meccanica. 43(3), 349–363. 13. Schiavi, A., Niccolini, G., Tarizzo, P., Carpinteri, A., Lacidogna, G. and Manuello, A. Acoustic emissions at high and low frequencies during compression tests in brittle materials. Strain, in print. 14. Livshits, L. D. and Gavrilov, B. G. (1987) Sources of acoustic emission and the seat of failure. Dokl. Akad. Nauk SSSR. 292(3). 15. Kukalov, G. I. and Yakovitskaya, G. E. (1993) Acoustic emission and stages of the crack-formation process in rock. Mining Institute, Siberian Branch, Russian Academy of Sciences, Novosibirsk. 2, 111–114. 16. Lacidogna, G., Carpinteri, A., Manuello, A., Durin, G., Niccolini, G. and Agosto, A., Acoustic and electromagnetic emissions as precursor phenomena in failure processes. Strain, in print. 17. Kachanov, L. M. (1986) Introduction to Continuum Damage Mechanics (Martinus Nijhoff). 18. Lemaitre, J. and Chaboche, J. L. (1990) Mechanics of Solid Materials (Cambridge University Press). 19. Krajcinovic, D. (1996) Damage Mechanics (Amsterdam: Elsevier).

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Numerical simulation of AE activity in quasi-brittle materials under compression

Stefano Invernizzi*, Alberto Carpinteri, Giuseppe Lacidogna, Amedeo Manuello Department of Structural Engineering & Geotechnics, Politecnico di Torino, Corso Duca degli Abruzzi 24 – 10129 Torino, Italy. *[email protected] ABSTRACT We present some numerical simulations of AE due to damage propagation in quasi-brittle materials under compression. To this purpose, the AE cumulative number, the time frequency analysis and the statistical properties of AE time series will be numerically simulated adopting the so-called “particle method strategy”. The method provides the velocity of particles in a set simulating the behavior of a granular system and, therefore, is suitable to model the compressive wave propagation and acoustic emission (corresponding to cracking) in a solid body. The numerical simulations correctly describe the compression test in terms of mean stress-strain and crack pattern. The size effects on the peak compressive strength and on the AE count are correctly reproduced. In addition, the amplitude distribution (b-value) and temporal evolution of AE events due to cracking, crucial for the evaluation of damage and remaining lifetime, was simulated according to the experimental evidences.

INTRODUCTION Nondestructive and instrumental investigation methods are currently employed to measure and check the evolution of adverse structural phenomena, such as damage and cracking, and to predict their subsequent developments. The choice of a technique for controlling and monitoring reinforced concrete or masonry structures is strictly correlated with the kind of structure to be analyzed and the data to be extracted [1,2]. This study addresses the pure compression test carried out in the laboratory and performed on drilled concrete cores obtained from two pilasters sustaining a viaduct along an Italian highway [3]. At the same time, the cracking process taking place during the test was monitored using the acoustic emission (AE) technique. A similar approach has been already exploited in [4] attempting to link the amount of AE with the structural deflections. In the assessment of structural integrity, the AE technique has proved particularly effective [4-6], in that it makes it possible to estimate the amount of energy released during the fracture process and to obtain information on the criticality of the process underway. Strictly connected to the energy detected by AE is the energy dissipated by the monitored structure. The energy dissipated during crack formation in structures made of quasi-brittle materials plays a fundamental role in the behavior throughout their entire life. Recently, according to fractal concepts, an ad hoc method has been employed to monitor structures by means of the AE technique [3-6]. The fractal theory takes into account the multiscale character of energy dissipation and the strong size effects associated with it. With this energetic approach, it becomes possible to introduce a useful damage parameter for structural assessment based on a correlation between AE activity in the structure and the corresponding activity recorded on specimens of different sizes, tested to failure by means of pure compression tests. The main achievement of the present work consists in showing how the amount of cracking obtained from the numerical simulation and the experimentally detected AE events share the same dimensional and temporal scaling laws.

COMBINED COMPRESSION AND AE TESTS By means of the AE technique, we analyzed the damage evolution in two pilasters sustaining a viaduct along an Italian highway built in the 1950s. From the pilasters we drilled some concrete cylindrical specimens in order to detect the mechanical properties of the material under compression and to evaluate the scale effects in size and time on AE activity [3].

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_14, © The Society for Experimental Mechanics, Inc. 2011

109

110

(a)

(b)

Figure 1: Apparatus adopted for compression tests (a); Stress and cumulated event number versus time (b).

Test specimens and testing equipment The cementitious material, of rather good mechanical characteristics, presents an apparent specific weight of 3 about 2.22 g/cm and a maximum aggregate size of about 15 mm. For each pilaster, three different specimen diameters d are considered in a maximum scale range of 1:3.4. The specimens present three different slendernesses: λ = h/d = 0.5, 1.0 and 2.0, with d chosen equal to 27.7, 59, 94 mm, respectively. For each of these nine geometries, three specimens have been tested, for a total of 54 cases (two pilasters). The average values obtained from the experimental data of pilaster P1 are reported in Table 1. The system adopted in the compression test utilizes rigid steel platens, the lateral deformation of concrete being therefore confined at the specimen ends, which are forced to have the same lateral deformation as the rigid platens (Figure 1a). Table 1: Experimental values obtained on concrete samples from pilaster P1. Specimen Type C11 C12 C13 C21 C22 C23 C31 C32 C33

Diameter d [mm] 27.7 27.7 27.7 59.0 59.0 59.0 94.0 94.0 94.0

Slenderness λ=h/d 0.5 1.0 2.0 0.5 1.0 2.0 0.5 1.0 2.0

Experimental peak stress [Mpa] 91.9 62.8 48.1 68.1 53.1 47.8 61.3 47.8 44.1

Nmax at 1186 1191 1188 8936 8934 8903 28502 28721 28965

u

Numerical peak stress [Mpa] 46.9 48.0 46.2 45.8 47.8 47.5 46.4 46.1 45.9

The stress and cumulated event number versus time for a specimen of intermediate size is represented in Fig. 1b. In the figure the critical number of AE cumulative events Nmax is represented in correspondence of the peak stress  u . Similar results can be observed in the other cases.

PARTICLE SIMULATIONS The simulations have been carried out with ESyS-Particle, an open source implementation of the Distinct Element Method [7]. ESyS-Particle has been developed in-house within the Earth Systems Science Computational Centre (ESSCC) at the University of Queensland [8] since 1994. The simulation is based on direct integration of the Newton’s motion equations with the Verlet algorithm. The normal interaction between colliding particles is linear and proportional to the particle small overlapping, whereas Coulomb friction, with both static and dynamic friction coefficients, rules the tangential interaction. In addition, bonded link are established between neighbor particles, according to the scheme in Figure 2a. The bonded link is

111 elastic perfectly brittle. The rupture of the bond is based on a fracture criterion that accounts for the axial, shear, torsion and bending behavior. The particles are filled together with the random packing algorithm LSMGenGeo [9], on the base of the maximum and minimum particle radius, which in our case correspond respectively to the maximum and minimum concrete aggregate radius (i.e. rmax=7.5 mm, rmin=1.5 mm). An exclusion method provides that once the bonded link is broken, the frictional interaction takes place between the neighbor particles. When the maximum and minimum radius of particles are quite different, experience shows that a power-law size distribution is obtained, providing a good approximation of the actual concrete aggregate size distribution.

(a)

(b)

(c)

Figure 2: Scheme of the bonded interaction between two particles (a); view of the specimen particle size distribution (colors indicate different diameters) (b); detail of the particles (in red) bonded to the loading platen (c). The load is applied to the specimen by means of moving planes. In order to simulate perfect friction between the loading platen and the specimens, the particles closer to the platens were bonded to the moving planes. In Figure 2c, bonded particles of the specimen are shown in red. The larger specimen was made of almost 9000 particles. Table 2: Mechanical parameters adopted for the simulation. Normal Stiffness Shear Stiffness Torsion Stiffness Bending Stiffness Normal Strength Shear Strength Torsion Strength Bending Strength

Bonds 0.5 0.15 0.04 0.017 0.0025 0.0125 0.00125 0.00125

Friction Normal Stiffness Shear Stiffness Dynamic Friction coefficient Static Friction coefficient

1.0 1.0 0.4 0.6

In the present study, special attention was paid to the simulation of the scaling of the mechanical response, rather than to provide a detailed interpretation of a specific set of tests. Therefore, the mechanical parameters adopted in the analysis (shown in Table 2) are to be considered nominal quantities, and do not correspond directly to the experimental ones. -4 -1 The simulations are carried out at a fixed strain velocity equal to 10 s , in this way the duration of each simulation is independent of the specimen size, and it is possible to adopt a unique time integration step equal to 0.04 s. A viscous type damping, proportional to the particle velocity is introduced in the simulation. The choice of the damping coefficient (in our case equal to 0.02) is based on the minimum value that regularizes the simulation without affecting the stress-strain behavior. During the simulation the position and velocity of each particle can be recorded at a certain integration time. The crack pattern and crushing of the specimen is shown in Figure 3 for specimen C23. During the first stage of loading, the specimen barrels. Soon after, the bonds of the central region are broken, diagonal cracks propagate and splitting mechanisms or even the explosion of the sample can take place.

112

Figure 3: Evolution of specimen crushing during the simulated compression test (sample C23).

2!10

4

3!10

4

2!10

4

2!10

4

1!10

4

2!10

4

3

2!10

4

1!10 5 1!10

4

5!10

0

(a)

0

2!10

4

4!10

4

6!10

Time

4

8!10

4

Intact Bonds

Broken Bonds

The stress-strain diagram is shown in Figure 4a for specimen C33. The load is obtained from the integration of the contact forces exchanged with the loading platens, while the position of the platens allows for the calculation of the strain. Figure 4b shows the evolution of the broken and intact bonded links number during the simulation (C33). In the very initial stage the number of intact bonds is constant, corresponding to the linear elastic regime. The damage starts quite before the maximum peak load and propagates slowly at the beginning, with an increasing number of broken bonds. At the peak load the damage spreads across the sample very rapidly, corresponding to a quite brittle behavior and a sudden drop in the number of intact bonds.

(b)

Figure 4: Stress strain diagram (sample C33) (a); Cumulative distribution of broken (continuous line) and intact bonds (dashed line) for increasing time (b).

113

y = 183.97  x 0.10

y = 0.0359 x 0.15 (a) (b) Figure 5: Strain energy evolution (a); comparison between experimental size effect on compression strength (diamonds and dotted line) and numerical simulations (stars and continuous line). In Figure 5a the strain energy of the sample is shown, calculated as the sum of the strain energy of each intact bond at a certain time. In the initial elastic regime, the strain energy is parabolic, but soon deviates as the damage starts to propagate. Finally, Figure 5b shows the bilogarithmic diagram of the peak load versus the specimen volume. The experimental values (diamonds) are aligned on a straight dashed line with slope equal to -0.10. The simulation results (stars) are aligned on a straight continuous line with slope equal to -0.15. The particle numerical simulation is able to describe correctly the trend and magnitude of the size effect on the compression strength, being in good agreement with experimental results.

COMPARISON BETWEEN NUMERICAL AND EXPERIMENTAL AE SCALING In previous works [4-6], a statistical and fractal analysis of data from laboratory experiments was performed, considering the multiscale aspect of cracking phenomena. The fractal criterion takes into account the multiscale character of energy dissipation and the strong size effects associated with it. This makes it possible to introduce a useful energy-related parameter for the damage determination of full-size structures, by comparing the AE monitoring results with the values obtained on a reference specimen sampled from the structure and tested up to failure. This approach has been exploited by the authors also for the interpretation of double flat-jack tests performed in historical masonry walls [10]. Fragmentation theories have shown that, during microcrack propagation, energy dissipation occurs in a fractal domain comprised between a surface and the specimen volume V [11,12]. This implies that a fractal energy density (having anomalous physical dimensions):

=

W max , V D /3

(1)

can be considered as the size-independent parameter. In the fractal criterion of Eq. (1), Wmax = total dissipated energy; = fractal energy density; and D = fractal exponent comprised between 2 and 3. On the other hand, during microcrack propagation, acoustic emission events can be clearly detected. Since the energy dissipated, W, is proportional to the number of the oscillations counts N, related to the AE events, AE can be considered as a size-independent parameter:

AE =

N max , V D /3

(2)

where AE = fractal acoustic emission event density; and Nmax is evaluated at the peak stress,  u . Eq. (2) predicts a volume effect on the maximum number of AE events for a specimen tested up to the peak stress. The extent of structural damage in a full-size structure can be worked out from the AE data recorded on a reference specimen (subscript r) obtained from the structure. From Eq. (2) we get:

114

V  N max = N max,r    Vr 

D /3

,

(3)

from which we can obtain the structure critical number of AE events Nmax. In order to provide a numerical interpretation to the AE phenomenon, the AE counting is put in direct comparison with the number of broken bonds at a certain time. More in detail, each event is composed of the number of bonds broken in a time interval equal to four time integration steps (i.e. equal to 0.16 s), while the magnitude is proportional to the number of broken bonds. This simple assumption implies that each bond releases the same energy at rupture, each bond being assigned the same strength. More detailed analogies have been presented in the literature (e.g. [13]), which require for the definition of a localization and clustering criterion for the events.

y = 0.8814  x 0.76

y = 0.01175  x 0.9 0 (a)

(b)

Figure 6: Frequency of broken bonds during the compression test C33 (a); comparison between experimental size effect on AE counts (diamonds and dotted line) and numerical simulations (stars and continuous line) (b). The frequency and magnitude of simulated events are shown in Figure 6a, compared with the evolution of the mean stress in the sample. In the elastic initial stage no emissions are recorded, since the specimen is undamaged. Afterwards, the frequency of high magnitude events gradually increases. After the peak load, a quite rapid decrease of events is recorded, in good agreement with the experimental evidence of the AE events. Figure 6b shows the bilogarithmic diagram of the experimental AE count (diamonds) and of the broken bonds (stars) at the peak load versus the specimen volume. This diagram emphasizes the size effect on the AE (dashed straight line) with an exponent equal to 0.76, which corresponds well with the numerical size effect on broken bonds (continuous straight line) with an exponent equal to 0.90.

AE frequency–magnitude statistics Since the studies of Mogi and Scholz [14,15] on AE, we know that the Gutenberg-Richter empirical law can be observed at the laboratory sample scale. They showed that a significant overlap exists between the definition of AE and earthquake. This is further reinforced by the evidence that brittle fracture obeys similar statistics from tectonic earthquakes to the dislocation movements smaller than micron size [16]. Moreover, in recent years, experiments employing acoustic emission have established remarkable results concerning the model of process zone and the quasistatic fault growth [17]. Such experiment-based knowledge is expected to be useful for studying the fundamental behavior of natural earthquakes, because it is widely accepted that fault systems are scale-invariant [18, 19] and there exist universal similarities between faulting behaviors, from small-scale microcracking to large-scale seismic faulting. For example, AE events caused by microcracking activity [14,17-20] and stick-slip along a crack plane [21, 22] are similar to those generated by natural earthquakes. By analogy with seismic phenomena, in the AE technique the magnitude may be defined as follows [12,23-24]:

m = Log Amax + f (r) , where Amax is the amplitude of the signal espressed in µV and signal amplitude is taken to be a decreasing function of the distance seismology, the Gutenberg-Richter empirical law [25]:

(4)

f (r) is a correction coefficient whereby the r between the source and the AE sensor. In

115

Log N ( m) = a  bm ,

(5)

expresses the relationship between magnitude and total number of earthquakes in any given region and time period, and is the most widely used statistical relation to describe the scaling properties of seismicity. In Eq. (5), N is the cumulative number of earthquakes with magnitude  m in a given area and within a specific time range, while a and b are positive constants varying from a region to another and from a time interval to another. Eq. (5) has been used successfully in the AE field to study the scaling laws of AE wave amplitude distribution. This approach evidences the similarity between structural damage phenomena and seismic activities in a given region of the earth, extending the applicability of the Gutenberg-Richter law to structural engineering. According to Eq. (5), the b-value changes systematically at different times in the course of the damage process and therefore can be used to estimate damage evolution modalities [12,23].

Figure 7: Evolution of the simulated b-value (circles) with time compared with the stress level (sample C33). For sample C33, the analysis of the b-value during the compression test was carried out. The b-values, computed for non-overlapping subsequent time windows of 120 s (i.e. 3000 time steps), are reported together with the stress curve. In other words, the b-values are determined during the tests by taking into account only current values. With this method, already adopted in other works on the damage analysis in structural concrete elements [23], the simulation time was subdivided into 6 intervals up to 720 s (i.e. 18000 time steps). The first time window has no events, since the specimen is in the linear regime, thus the b-value is not calculated. In Figure 7, the final b-value approaches to one for sample C33 [24]. Figure 7 also shows that AE generated during the early stages of loading implies an high b-value > 1.5. In particular, the b-value obtained for sample C33 results to be greater then 3 immediately after the beginning of the test, it reaches the value 1.5 just before the peak stress and finally tends to 1.0 at the peak stress [26].

CONCLUSIONS A numerical simulation of an innovative concrete compression test combined with acoustic emission monitoring has been proposed. The numerical results agree rather satisfactorily with the experimental evidences, and the crack patterns are simulated correctly. The model is able to describe the decrease in the overall strength with increasing size. In addition, the number of acoustic emissions is put into relation with the number of broken bonds at the peak stress. A good correlation is found between the amount of cracking simulated numerically and the experimental acoustic emission counting for different specimen sizes. Finally, the amplitude distribution (b-value) and temporal evolution of AE events due to cracking, crucial for the evaluation of damage and remaining lifetime, have been simulated in good agreement with the experimental data.

ACKNOWLEDGEMENTS The financial support provided by the Piedmont Region (Italy) to the Project ‘‘Preservation, Safeguard and Valorisation of Masonry Decorations in the Architectural Historical Heritage of Piedmont” (RE-FRESCOS) is gratefully acknowledged.

116

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26]

A. Carpinteri, P. Bocca, Damage and Diagnosis of Materials and Structures, Pitagora Editrice, Bologna, Italy. (1991). A. Anzani, L. Binda, G. Mirabella Roberti, “The effect of heavy persistent actions into the behavior of ancient masonry”, Materials & Structures, 33, 251-261 (2000). A. Carpinteri, G. Lacidogna, N. Pugno, “Structural damage diagnosis and life-time assessment by acoustic emission monitoring”, Engineering Fractures Mechanics, 74, 273-289 (2007). A. Carpinteri, S. Invernizzi, G. Lacidogna, “Structural assessment of a XVIIth century masonry vault with AE and numerical techniques”, International Journal of Architectural Heritage, 1(2), 214-226 (2007). A. Carpinteri, G. Lacidogna, “Structural monitoring and integrity assessment of medieval towers”, Journal of Structural Engineering (ASCE), 132, 1681-1690 (2006). A. Carpinteri, G. Lacidogna, “Damage monitoring of a masonry building by the acoustic emission technique”, Materials & Structures, 39, 161-167 (2006). P.A. Cundall “A computer model for simulating progressive large scale movements in blocky rock systems”, in Proceedings ISRM Symp., Nancy, France, 2, 129-136 (1971). https://twiki.esscc.uq.edu.au/bin/view/ESSCC/ParticleSimulation https://launchpad.net/esys-particle/ A. Carpinteri, S. Invernizzi, G. Lacidogna, “Historical brick-masonry subjected to double flat-jack test: Acoustic Emissions and scale effects on cracking density”, Construction and Building Materials, 23(8), 28132820, (2009). A. Carpinteri, N. Pugno, “A multifractal comminution approach for drilling scaling laws”, Powder Technology, 131(1), 93-98, (2003). Carpinteri, A., Lacidogna, G., Niccolini, G., Puzzi, S. “Critical defect size distributions in concrete structures detected by the acoustic emission technique”, Meccanica, 43, 349–363, (2008). J.F. Hazzard, R.P. Young, “Simulating acoustic emissions in bonded-particle models of rock”, International Journal of Rock Mechanics in Mining Science, 37, 867-872, (2000). Mogi, K. “Magnitude frequency relations for elastic shocks accompanying fractures of various materials and some related problems in earthquakes”, Bull. Earthquake Res. Inst. Univ. Tokyo, 40, 831-853, (1962). Scholz, C. H. “The frequency-magnitude relation of microfracturing in rock and its relation to earthquakes”, Bull. Seismol. Soc. Am., 58, 399-415, (1968). M.C. Miguel, A. Vespignani, S. Zapperi, J. Weiss, J.R. Grasso, “Intermittent dislocation flow in viscoplastic deformation”, Nature, 410, 667-670 (2001). D.A. Lockner, J.D. Byerlee, V. Kuksenko, A. Ponomarev, A. Sidorin, “Quasi static fault growth and shear fracture energy in granite”, Nature 350, 39-42 (1991). T. Hirata, “Fractal dimension of fault system in Japan: fracture structure in rock fracture geometry at various scales”, Pure Appl. Geophys. 131, 157-170 (1989). E. Bonnet, O. Bour, N.E. Odling, P. Davy, I. Main, P. Cowie, B. Berkowitz, “Scaling of fracture systems in geological media”, Rev. Geophys. 39, 347-383 (2001). X.Lei, “How do asperities fracture? An experimental study of unbroken asperities”, Earth Planet. Sci. Lett. 213, 347-359 (2003). N. Kato, K. Yamamoto, T. Hirasawa, “Microfracture processes i the break down zone during dynamic shear rupture inferred from laboratory observation of near-fault high-frequency strong motion”, Pure Appl. Geophys. 142, 713-734 (1994). X.L. Lei, O. Nishizawa, K. Kusunose, A. Cho, T. Satoh, “On the compressive failure of shale samples containing quartz-healed joints using rapid AE monitoring: the role of asperities”, Tectonophysics 328, 329340 (2000). S. Colombo, I.G. Main, M.C. Forde, “Assessing damage of reinforced concrete beam using ‘‘b-value’’ analysis of acoustic emission signals”, J. Mat. Civil Eng. (ASCE) 15, 280-286 (2003). M.V.M.S. Rao, P.K.J. Lakschmi, “Analysis of b-value and improved b-value of acoustic emissions accompanying rock fracture”, Curr. Sci. – Bangalore 89, 1577–1582 (2005). C.F. Richter, Elementary Seismology. W.H. Freeman, San Francisco, CA; and London. (1958) A. Carpinteri, G. Lacidogna, A. Manuello, “The b-Value Analysis for the Stability Investigation of the Ancient Athena Temple in Syracuse”, Strain, doi:10.1111/j.1475-1305.2008.00602.x (2009).

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Mechanical characterization and AE of translucent self-compacting concrete plates in bending A. Manuello1, A. Carpinteri1, S. Invernizzi1, G. Lacidogna1, S. Pagliolico2, A. Torta2 1

Politecnico di Torino, Department of Structural Engineering & Geotechnics, Corso Duca degli Abruzzi 24 – 10129 Torino, Italy

2

Politecnico di Torino, Department of Materials Science and Chemical Engineering, Corso Duca degli Abruzzi 24 – 10129 Torino, Italy E-mail: [email protected]

ABSTRACT An experimental and numerical study on the mechanical behaviour of an innovative composite material based on the combination of a self-compacting concrete (SCC) matrix with transparent glass inclusions is proposed. The experimental tests have been monitored by an acoustic emission (AE) device. The results are interpreted by a FEM model accounting for the fracture of the two different materials and the interface between them. The AE monitoring is used for the definition of the crack pattern, and to determine the fracture energy dissipation domain. INTRODUCTION An experimental and numerical study on the mechanical properties and behaviour of an innovative composite material based on the combination of a self-compacting concrete (SCC) matrix with transparent glass inclusions is proposed. The resulting composite is partially transparent, and can be effectively adopted to realize light separation walls and covering of façades. In addition, the glass inclusions are obtained from resulting of the float glass production, the new material providing a special appeal from the sustainability point of view. The innovative composite material can be compared to the traditional architectural components like the “glass concrete walls”. The panel is obtained by introducing recycled glass into the concrete matrix, with the aim at reclaiming and rendering inert glass waste and industrial residue inside a new sustainable material for design applications. As the geometry of the waste glass is not predefined, each single panel differs from the others and every piece can be considered a unique while it can be produced in an industrial series. The material is also adaptive to the specific design, it can be different in the light transmission as in the sound and heath transmission by variations in the quantity of glass or the mix design of the opaque matrix of concrete. An example of the translucent concrete were presented in the international “Concrete Design Competition Plastic-Opacity 2006” [1,2]. In the present paper, a campaign of experimental tests of fibre reinforced and unreinforced SCC plates in bending is presented. The tests have been also monitored by an acoustic emission (AE) device with six sensors, which allowed for the localization of the damage zones. The AE monitoring is particularly effective for the definition of the actual crack pattern, and thus to determine the true domain where the fracture energy dissipation takes place. The experimental results are interpreted by means of a three-dimensional finite element model accounting for the fracture of the two different materials, glass and SCC, and of the interface between them. EXPERIMENTAL SETUP The experimental tests were conducted using a servo-controlled machine (MTS) with a closed-loop control (Fig. 1a). The flexural tests were realized by a linear actuator (hydraulic jack) with passing stem acting in the middle point of the upper side of the composite plate. In particular, two kinds of composite plates have been tested. In the first case steel fibres were added into the self-compacting concrete (SCC) mix (P1). In the second case no reinforcement was used (P2). The equipment adopted by the authors consisted of six USAM® units for AE measurements, and six pre-amplified piezoelectric (PZT) AE sensors applied to the external surfaces of the specimens (Fig. 1a). Such sensors were calibrated on frequencies from 50 to 800 kHz. PZT sensors were used, thereby exploiting the capacity of certain crystals to produce electric signals whenever they are subjected to a mechanical stress. T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_15, © The Society for Experimental Mechanics, Inc. 2011

117

118 The USAM® units were synchronized for multi-channel data processing [1,2]. The most relevant parameters acquired from the signals were stored in the memory of the USAM units and then downloaded to a PC for multichannel data processing and then micro-crack localisation. The signal parameters recorded by each USAM unit include: the first up-threshold crossing time (t0) for P-waves, the last down-threshold time (t1), and the number of threshold crossings (see Fig. 1b).

S4

(a)

(b)

Figure 1: Experimental setup and positions of AE sensors (a). Typical AE signal identified by the transducer (b). AE MONITORING Acoustic emission is defined as the spontaneous release of localized strain energy in a stressed material resulting, for example, from micro-cracking and can be recorded by sensors (piezoelectric transducers) applied on the surface of the specimen [3,5-7]. To obtain information on the damage evolution in materials and structures, the cumulative AE events are considered [7]. Nevertheless, to investigate the failure mechanisms of AE sources it is also possible to apply more sophisticated procedures, such as the moment tensor analysis. The use of five or more transducers is required to determine the radiation pattern of a general AE source. Fracture type can be deduced using at least six sensors. The basis for these quantitative methods are the localization techniques to be used in order to extract the source coordinates of the AE events as accurately as possible [3,8,9]. With the localization of AE sources, damage or material flaws can be detected. If AE source characteristics are to be analysed quantitatively in order to investigate damage mechanisms, the knowledge of the source locations is a requirement [8-10]. The AE technique was applied to evaluate the damage evolution of the two plates (P1 and P2) in flexural loading. The AE signal received by the transducers is processed by an analyser which counts the oscillations exceeding a certain voltage threshold. This makes it possible to plot the cumulative curves of AE measured continuously throughout the monitoring period. This method, referred to as Ring-Down Counting, is widely used for defect detection purposes. As a first approximation, in fact, the cumulative number N can represent with the quantity of energy released during the monitoring process, and can be assumed to increase proportionately with the damage evolution. In Fig. 2a and 2b, the Load vs. Time and the AE cumulated counts during the tests are reported for specimens P1 and P2. It can be noted that specimen P1, reinforced by steel fibres added into the SCC matrix, shows an higher flexural strength (Pmax=3.5 kN) and a more brittle behaviour if compared to specimen P2 (Pmax=1.7 kN). These differences in the failure mode for specimens P1 and P2 can be also recognized using the AE device and analysing the cumulated AE counts. In fig. 2a (specimen P1) the cumulated AE curve shows a rapid increase very close to the collapse. Between 3200 s and 4200 s, a very significant increase in the cumulated AE can be directly related to the ultimate load and to the catastrophic failure of the reinforced composite plate (Fig. 2a). After the peak load the AE rate starts to decrease and a very low number of AE per unit time is observed (4200-10200 s). A different behaviour can be observed for specimen P2 where no fibre-reinforcement is adopted into the SCC matrix. In this case the cumulated AE number shows an appreciable increase just during the first stages of the test. After 1/5 of the testing time (∼1000 s) approximately 500 AE events have been recorded (Fig. 2b). In this case the AE activity is not concentrated close to the peak load but it involves a progressive damage evolution. In particular, in this case the cumulated AE number systematically increases with the different stages of damage, so it could be used to estimate the development of the fracture process. The cumulative AE count exhibits a strong

119 correlation with the decay of the mechanical properties of the specimens. Sudden increases in the cumulated AE are related to appreciable drops of the Young’s modulus during the damage process in the specimens (see Fig. 2b). These results lead to consider the cumulated AE number as a valid indicator in order to evaluate the failure modes in composite SCC plates and to discriminate between brittle failure (P1) and a more ductile behaviour (P2).

Figure 2: Load vs. time curve and cumulated number of AE for specimens P1 (a) and P2 (b). LOCALIZATION AND CHARACTERIZATION OF AE SOURCES Using the AE device it is also possible to localize the AE source within the monitored specimen or structure. Source localization concepts originated in seismology, where the aim is to localize the hypocenter or epicenter of an earthquake from seismograms obtained at stations distributed over the Earth’s surface [11]. Similar problems are posed when attempting to localize sources of acoustic emission. Again, knowledge of the wave propagation characteristics between source and receiver is necessary. For the most general case, these characteristics depend on the mode of propagation, the elastic moduli, the existence of attenuation due to heterogeneities and anisotropy of the material. The equations used for source location are based on the assumptions of homogeneous and isotropic medium and point-like sources, implying spherical wave propagation. Generally speaking, for single point-like sources in geometries having continuous straight line paths between the source and each receiver, the location technique is called “triangulation procedure” [3,6]. The observed wavelengths are usually larger than the maximum aggregate size in concrete, so that a more or less homogeneous and isotropic distribution of the P-wave velocity can be assumed. During the first stage, the groups of signals, recorded by the various sensors, that fall into time intervals compatible with the formation of microcracks in the volume analysed, are identified. These time intervals, of the order of micro-seconds, are defined on the basis of the presumed speed (vp) of transmission of the waves (P) and of the mutual distances of the sensors applied to the external surface of the material. In the second stage, when the formation of microcraks in a three-dimensional space is analysed, the triangulation technique can be applied if signals recorded by at least five sensors fall into the time intervals. Thus, with this

120 procedure it is possible to define both the position of the microcracks in the volume and the speed of transmission of P-waves. Having denoted with ti the time of arrival at a sensor Si of an AE event generated at point S and at 2 2 2 1/2 time t0, |S − Si| = [(x − xi) + (y − yi) + (z − zi) ] , the distance between Si and source S, in Cartesian coordinates, and assuming the material to be homogenous, the path of the signal is given by: |S − Si| = vp(ti − t0). If the same event is observed from another sensor Sj at time tj: S − S j − S − Si = v p (t j − ti ) ≡ v p ∆ t ji

.

(1)

Assuming the arrival times of the signals and the positions of the two sensors to be known, eq. (1) is an equation with four unknowns, x, y, z and v. Hence, the localisation of S is a problem that can be solved in an exact manner if it is possible to write a sufficient number of equations such as eq. (1), i.e., when an AE event is identified by at least five sensors. The localisation procedure can also be performed through numerical techniques using optimisation methods such as the Least Squares Method (LSM) [3,6]. LSM seeks to determine the source location-estimate which minimizes the sum of all squared time-residuals. The nonlinear system involved by the LSM method is usually solved with iterative techniques and algorithms based on Gauss-Newton’s method as in Seismology [3,6,11] MOMENT TENSOR ANALYSIS From the theoretical standpoint, the moment tensor anlysis relies on the procedure defined by Shigeishi and Ohtsu [8]. Such procedure characterizes the AE signal by taking into account only the first arrival time of the Pwaves. The elastic crack displacements u(x,t), at points x, which are the sources of the AE signals, are given by:

ui (x, t ) = Gip, q (x, y, t ) m pq * S (t ) ,

(2)

where Gip,q(x,y,t) is the space derivative of Green’s function, mpq are the moment tensor components, S(t) is the function describing the displacement time-dependence and the asterisk denotes the convolution operator [8,12]. Green’s functions describe the elastic displacements, u(x,t), due to a unit displacement applied at y at time t. In the SiGMA procedure, proposed by Shigeishi and Ohtsu [6], the magnitude of the elastic displacements, proportional to the amplitude, A(x), of the first P-waves reaching the transducers, is given by a modified version of Eq. (2):

⎛ m11 m12 ⎜ C s REF (t, r ) A(x ) = (r1 r2 r3 )⎜ m21 m22 R ⎜m ⎝ 31 m32

m13 ⎞ ⎛ r1 ⎞ ⎟⎜ ⎟ m23 ⎟ ⎜ r2 ⎟ , m33 ⎟⎠ ⎜⎝ r3 ⎟⎠

(3)

where Cs is a calibration coefficient of the acoustic emission sensors, and R is the distance between the AE source at point y and the sensor located at point x; vector r represents the components of the distance between the source and the sensor; REF(t,r) is the reflection coefficient of the sensitivity of the sensor depending on the angle between the directions of the two unit vectors r and t; these vectors are the unit vector along the R distance and the unit vector along the sensitivity sensor direction, respectively. To represent the moment tensor, it is necessary to determine the six independent unknowns, mpq. The amplitude of the signal, A(x), must be received by at least six AE sensors. Through an eigenvalue analysis of the moment tensor, it is possible to determine the type of crack localised. Using the ratios between the individual eigenvalues and the maximum one, we can write:

λ Y λ λ1 Y = X + Y + Z , 2 = 0 − + Z , 3 = −X − + Z , λ1 2 λ1 λ1 2

(4)

where, λ1, λ2, λ3,are the maximum, intermediate and minimum eigenvalues, respectively, X is the component due to shear, Y is the deviatoric tensile component, Z is the isotropic tensile component. Ohtsu classified an AE source with X > 60% as a shear crack (Mode II), a source with X < 40% and Y + Z > 60% as a tensile crack (Mode I), and a source with 40% < X < 60% as a Mixed Mode crack [8]. Moreover, from an eigenvector analysis, it is possible to determine the unit vectors, l and n, which determine the directions of the displacement and the orientations of the crack surface. Using the USAM® equipment the authors have fine-tuned a computer-based procedure including the AE source location and the moment tensor analysis [10]. The 3D AE source positions obtained are overlapped to the 3D crack patterns obtained from the observation of the cracking map (Fig.3). Damage localization and typology of

121 the crack for specimens P1 and P2 are shown in Figs. 3a,b,c and d. The position of each emission source, the typologies for Mode I, Mode II and Mixed Mode are represented according to the notation listed in Table 1. Table 1: Markers of AE sources and labels identifying crack typology. source Crack mode Mode I Mode II Mixed Mode

Figure 3: Localization and moment tensor analysis performed for the two plates show the crack positions and the crack types (Mode I, Mode II, and Mixed Mode) according to the notation listed in Table 1. The Plate P1, which is characterized by a more brittle behaviour, shows a single main crack crossing the entire specimen. In Fig.3a the AE localized points are shown together with the cracking pattern. In Fig. 3b the crack typologies (Mode I, Mode II and Mixed Mode), for each localized point, are represented according to the notation listed in Table 1 for specimen P1. In Fig. 3c and 3d, analogous results are shown in the case of specimen P2. From the crack typologies shown in Fig. 3b and 3d for the two plates, it can be noted that plate P1 presents predominantly cracks of Mode I, whereas plate P2 presents all the three different types of crack. NUMERICAL SIMULATIONS The numerical model of plate P2 in bending test was built accounting for the distribution of aggregates and cementitious matrix. Quadratic elements were used to represent both the glass inclusions and the matrix, while special interface elements were placed in between the two phases. The adopted mesoscale modeling directly accounts for the inclusions, being each inclusion explicitly represented. The failure of the two phases was assumed as driven by pure plasticity in compression, with limit yielding stress, and by linear softening in tension. A fixed smeared crack model based on total deformation was used. In order to avoid mesh dependency, the constitutive law for continuum elements was regularized providing the area beneath the stress vs. strain diagram in tension to be equal to the tensile fracture energy divided by the crack bandwidth (i.e. the element size). The constitutive relation for interfaces was assumed elastic in compression and linear softening in tension. Since the constitutive law for interface elements is directly formulated in terms of stress vs. crack opening displacement (i.e. a cohesive law is assigned), no further regularization is necessary. All the analyses were performed with the Finite Element Software DIANA 9.4.

122 The distribution and positions of glass inclusions exactly account for the actual geometry of each sample. The mechanical properties of the materials are summarized in Table 2. Figure 4 shows the mesh used for the numerical model.

Figure 4: Mesh used for the numerical model. Table 2: Mechanical properties adopted in the numerical analysis. Glass inclusions 9

Matrix 9

Interface

Young’s modulus

E

7.5 10 Pa

3.9 10 Pa

----

Poisson ratio

ν

0.23

0.24

----

Tensile strength

ft

6

40.0 10 Pa

6

6.0 10 Pa

5.0 10 Pa

Fracture energy

Gf

1.0 N/m

6.0 N/m

3.0 N/m

Compressive strength Density

fc ρ

9

1.0 10 Pa 2500 kg/m

3

8

6

1.0 10 Pa

----

3

----

2900 kg/m

Figure 5: Deformed mesh of the unreinforced plate (P2) under loading.

123

Figure 6: Damage crack localization on plate P2. The numerical results can be put in comparison with the results of AE localization shown in Fig. 3b. In Fig. 5 the deformed mesh of the numerical model for plate P2 is shown. In Fig. 6 the numerical results for the damage localization in plate P2 are reported. The numerical results can be put in comparison with the cracking pattern in the real plate and the results obtained by the AE triangulation procedure (Fig 3c). CONCLUSIONS An experimental and numerical study on the mechanical properties and behaviour of an innovative composite material based on the combination of a self-compacting concrete (SCC) matrix with transparent glass inclusions is proposed. The cumulative AE count exhibits a strong correlation with the decay of the mechanical properties of the two plates (P1 and P2). The cumulated AE number can be considered as a valid indicator in order to evaluate the failure modes in composite SCC plates and to discriminate between brittle failure of the reinforced plate P1 and more ductile behaviour of the unreinforced plate P2. In addition the AE localization, performed by the triangulation procedure, and the moment tensor analysis lead to interpret the cracking pattern and to identify the most predominant crack typology (Mode I). Finally, The numerical results were put in comparison with the cracking pattern for plate P2 and the results obtained by the AE triangulation procedure. ACKNOWLEDGEMENTS The financial support provided by Regione Piemonte (RE-FRESCOS project), Buzzi Unicem Group and ISI Foundation is gratefully acknowledged. We wish to thank Buzzi Unicem Group for the concrete mix design and elaboration of prototypes, and AGC Flat Glass Europe for waste glass supply. Special thanks go to Mr. Vincenzo Di Vasto of the Politecnico di Torino. We also gratefully acknowledge Architect Dario Parigi for the helpful discussion and suggestions on the translucent concrete obtainable by glass inclusions. REFERENCES [1] M.T. Briotti, “Concrete Design Competition 2005-2006 “Plastic – Opacity. Concrete Master Class, Fondazione Bauhaus Dessau 21-26 agosto 2006, L’industria italiana del Cemento, n.826, 940-941, ( 2006). [2] M. Giraud, Concrete: plasticity, strength and transparence, Portland, 40, 32-34, (2007). [3] Carpinteri, A. Lacidogna, G. and Niccolini, G. Critical Behaviour in Concrete Structures and Damage Localization by Acoustic Emission. Key Engineering Materials, 312, 305-310 (2006). [4] Anzani, A. Binda, L. Carpinteri, A. Lacidogna, G. and Manuello, A. Evaluation of the Repair on Multiple Leaf Stone Masonry by Acoustic Emission, Materials and Structures (RILEM), 41, 1169-1189 (2008).

124

[5]

Köppel S. and Grosse, C.U. Advanced Acoustic Emission Techniques for Failure Analysis in Concrete, Proceedings of 15th World Conference on Nondestructive Testing, Rome, (2000). [6] Shah S. P. and Li, Z. Localisation of Microcracking in Concrete Under Uniaxial Tension, ACI Materials Journal, 91, 372-381 (1994). [7] Carpinteri, A. and Lacidogna, G. Damage Monitoring of an Historical Masonry Building by the Acoustic Emission Technique, Materials and Structures, 20, 143-149 (2006). [8] Shigeishi, M. and Ohtsu, M. Acoustic Emission Moment Tensor Analysis: Development for Crack Identification in Concrete Materials, Construction and Buildings Materials, 15: 5-6, 311-319 (2001). [9] Grosse, C.U. Reinhardt, H. W. and Finck, F. Signal Based Acoustic Emission Techniques in Civil Engineering, ASCE Journal of Materials in Civil Engineering, 15:3, 274-279 (2003). [10] Carpinteri, A., Lacidogna, G. and Manuello, A., “An Experimental Study on Retrofitted Fiber-Reinforced Concrete Beams using Acoustic Emission”, in Fracture Mechanics of Concrete Structures, Proceedings of the 6th International FraMCoS Conference, Catania, Italy, 2, 1061-1068 (2007). [11] Geiger, L. Probability Method for the Determination of Earthquake Epicentres from the Arrival Time Only, Nachrichten von der Koniglichen Gesellschaft der Wissenschaften zu Gottingen, 4, 331–349 (1910). [12] Aki, K. Richards P.G. Quantitative Seismology, Theory and Method, 2nd ed. University Science Books, Sausalito CA, (1980).

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Reconfiguration of Large Leaf under Wind Load

Liou*, N.-S., Tsai, S.-S. and Yen, H.-H. Department of Mechanical Engineering, Southern Taiwan University Yuang-Kang City, Tainan Hsien, Taiwan 710 R.O.C email: [email protected]

ABSTRACT In this study, the reconfiguration of Heliconia rostrata leaf was investigated by using fluid structure interaction (FSI) analysis at different wind speeds. In order to perform FSI analysis, the time dependent mechanical properties of petiole and lamina were modeled by using linear viscoelastic constitutive laws and the parameters were fitted from 3 point bending tests and tensile tests experimental data. Besides, in fluid domain, air is considered as incompressible fluid in this study. The drag force and banding moment of Heliconia rostrata leaf at different wind speeds (4 and 6 m/sec) under headwind condition were examined. The FSI analyses show that, compared with drag forces of rigid leaf under headwind, the drag forces of Heliconia rostrata leaf reduce 35% and 63% under 4 and 6 m/sec headwind conditions respectively. The findings of this study can help us design better flexible structures such as wind turbine blades. INTRODUCTION Terrestrial plants are frequently subjected to winds and their response to winds may be critical to their survival. For plants with few large leave, the response of any single leaf under wind plays important role to the total dynamic response of plant under wind. Taking leaf of Heliconia rostrata (Hanging lobsster-claws) as an example, the deformation of petiole and highly deformable leaf laminate may induce a reduction in the effective cross-flow area, and the deformed shape of leaf may become more streamlined. These combination effects, or reconfiguration effects, reduce the drag force to minimize the chance of breakage of stem or uprooting of plant. From mechanical point of view, factors such as the torsional stiffness, bending stiffness of petiole, stiffness of leaf lamina and material damping may affect the reconfiguration of leaf under winds and then affect the reduction of drag force. In this study, how the mechanical properties of petiole and lamina of Heliconia rostrata leaf affect the reconfiguration of leaf and the reduction of drag force under winds was investigated. Fluid structure interaction (FSI) analysis was used to study the response of Heliconia rostrata leaf at different wind speeds. Mechanical tests were used to acquire the viscoelastic properties of leaf and petiole used in this study. The findings of this study can help us design better flexible structures such as wind turbine blades. MATERIAL AND METHOD In order to simulate the influence of the wind on the reconfiguration behavior of Heliconia rostrata leaf, a loosely coupling approach using commercial finite element and computational fluid dynamic software was employed in this work. The mechanical behavior of leaf was calculated using the finite element code ABAQUS and the wind dynamics in the fluid domain was computed by using finite volume code FLUENT. The exchange of force and displacement information between solid and fluid domain is performed by using mesh based parallel code coupling interface MpCCI. The bidirectional coupling algorithm used for FSI analysis in this study is shown in Fig. 1.

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_16, © The Society for Experimental Mechanics, Inc. 2011

125

126

Figure 1: Couple algorithm used for FSI analysis Based on the typical leaf of Heliconia rostrata, the simplified 3D geometry model was built. This simplified geometry model was used to construct the FEA and CFD models (Fig. 2).

(a)

(b)

Figure 2: (a) Simplified geometry model and (b) FEA model of Heliconia rostrata leaf Linear viscoelastic constitutive laws were used to model the mechanical properties of petiole and lamina of Heliconia rostrata. The time dependent mechanical properties of petiole and lamina were acquired by using 3 point bending tests and tensile tests respectively. The experimental setups for 3 point bending tests and tensile tests are shown in Fig. 3. Beam theory was used to convert the experimental force-time and displacement-time data of petiole specimen into stress-time and strain-time data. The parameters of linear viscoelastic constitutive models of petiole and lamina for finite element model were fitted from the aforementioned experimental data. In fluid domain, air is considered as incompressible fluid and turbulence model was used for FSI analysis. By using FSI analysis, the drag force and banding moment of Heliconia rostrata leaf at different wind speeds (4 and 6 m/sec) under headwind condition were examined.

(a)

(b)

Figure 3: (a) Petiole under 3 point bending test (b) leaf under tensile test

127 RESULT AND CONCLUSION The typical reconfiguration shape of Heliconia rostrata leaf under headwind is shown in Fig. 4. It can be seen that the cross-flow area is reduced and the shape becomes streamline.

Figure 4: Typical reconfiguration shape under headwind condition (wind speed 4m/sec) The FSI analyses of this study show that, compared with drag forces of rigid leaf under headwind, the drag forces of Heliconia rostrata leaf reduce 35% and 63% under 4 and 6 m/sec headwind conditions respectively. The FSI analyses also show that the bending moments of Heliconia rostrata leaf reduce 35% and 39% under 4 and 6 m/sec headwind conditions respectively due to reconfiguration. The reduction of drag force and bending moment decreases the chance of uprooting of plant or broken of main stem. The findings of this study can help us design better flexible structures such as wind turbine blades. ACKNOWLEDGEMENT This work was supported by NSC 96-2221-E-218-048-MY3 from the R.O.C government.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Manufacturing Challenges for the Modern Wind Turbine Rotor Gary Kanaby Vice President, Sales Knight & Carver Wind Group® 2423 Hoover Ave National City, CA 91950 [email protected]

OVERVIEW Manufacturing Challenges for the Modern Wind Turbine Rotor As blades have gotten larger, challenges must be met to assure delivery of the power producing element of the turbine. The industry does not allow for years of testing and development; the investment is enormous. Not all of the assumptions about how blades could be scaled have proven out. Margins of the blade design have been reduced due to the fact that weight must be kept at a minimum not only to save cost but to reduce the moment of the blade to a tolerable level. Some questions and a path forward will be presented in this paper. Material Handling and Placement

How do we place 10,000kg of materials into the molds and cycle the tooling in 24hrs? The movement of large quantities of materials has to be done quickly and correctly. Resin Infusion

How do we prevent fiber movement during infusion? Once the glass is placed in the mold, a vacuum compresses the fibers prior to infusion of the resin. After the compression, the fibers must remain in the proper place. With large stacks of materials, especially in the root area forming the circumference of the blade, there is a tendency of the glass to move causing wrinkles or waves. When the fibers are not laid straight, much of the strength is lost.

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_17, © The Society for Experimental Mechanics, Inc. 2011

129

130

How do we assure that the dry glass fiber is properly wetted? The resin needs to have a clear path in order to wet out all the glass fibers. Although this is carefully calculated, variations may mean sometimes not all the fibers wet out. Bond assembly

How do we assemble 50 meter long components with consistent bond lines? As blades get longer the tolerances of bond lines have remained nearly the same. The difficulty of bonding the parts together properly is a big challenge. No longer are test fits made on every blade. Keeping the amount of bonding paste down will not only save money but weight.

Areas of Improvement and Research Opportunities There are many areas of improvement and research opportunities that will be addressed later.

131

MEGAWATT CLASS TURBINES A feeling for the size Just how big are turbines? Compared to a passenger jet airplane, they are large. The blades become longer than the wingspan.

Figure 1: Comparison of a 1 and 3 MW wind turbine to a passenger jet airplane.

132

25 Years of Blade Scaling In the last 25 years the size of standard blade installed has gone from 9 meters in length (weighing 170kg) to 45 meters in length (weighing 10,000kg). The models, design parameters and fabrications methods have not always scaled up as we would have wished.

Figure 2: 9 meter blade on a 100kw turbine.

Figure 3: 60 meter blade for a 5 mw turbine.

Source: Euros

133

BLADE CONSTRUCTION There are two main techniques of blade construction. Although both are used in the industry, this paper deals mainly with the most prevalent one which is stressed shell infused blade.

Spar-shell Spar shell blade construction has both advantages and disadvantages. The spar in this case takes all the loads and is premade as a component. This spar is sometimes built robotically reducing labor. The shells are usually made of a high quality high cost pre-impregnated fiberglass. Although the properties of the materials are very good, building the spar in this way can add weight. An advantage of this construction method is that the bonding area is large and detection of voids is easy though the thin shell laminate in that area.

Figure 4: Spar-shell blade construction.

Stressed-shell The most common way to build blades is to place all the dry materials into a mold half and infuse the resin after a vacuum is applied. The structural “spar cap” is built into the skin. Both skins are bonded to the shear web(s). This method allows fast turnaround of the tooling but the bond lines are more numerous and hard to control. Detection of spar bond deficiencies now is more difficult looking through the thick laminates of the spar cap.

Figure 5 : Stressed-shell blade construction.

134

Dry Material Placement Tons of fiberglass as well as coring must be placed in the mold. Each laminate must be oriented in the proper alignment and begin and end at the proper location. Laser images can be projected onto the mold surface showing each laminate’s location. The fiberglass must be placed in the root area of the blade as well. This is where many problems occur. The dry glass must remain in a vertical position and stay there while vacuum is applied. Adhesive sprays are used to temporarily hold the fabrics in place. Calculations must be made to allow for the movement anticipated during compression. The final adjustments in the factory usually are done through experience and trial and error.

Figure 6: Placement of fiberglass in a large mold.

Source: Siemens

135

Resin Infusion Resin infusion is accomplished by creating a vacuum throughout the laminate and with several ports have the atmosphere push the resin into the fiberglass and core. Careful planning will allow all of the fibers to be wetted. Not only are the entry ports and the vacuum lines placement important, the fiberglass and/or coring must be engineered to allow the resin to flow. This method of construction has slightly higher resin content than pre-preg but the cost savings are considered well worth it. The challenges are to hold the vacuum over a large area and to have the ability to get all the glass infused in a short time. Holding the vacuum can be a problem if there are the slightest leaks. If a leak occurs, a dry spot will result.

Figure 7: Technician listening for a leak on a blade shell during infusion.

Source: Sandia National Laboratories

136

BLADE CONSTRUCTION ISSUES Fiber Straightness and Flatness Fibers can shift during the compression of the fabrics. Wrinkles occur usually in the root area where laminates are thick. The fabric folds over on itself. Most of the strength is gone across the wrinkle. Most wrinkles cannot be detected until after the vacuum bag is removed and the resin is hard. Then there is no other option than to grind out the wrinkle and replace the fiberglass in that region. In other areas of the blade the fabric can move but over a longer distance. These “waves” also cause lose of strength and fatigue life. In a blade usually the compression strength is very important. When the fibers are not straight some compressive strength is lost. Sometimes wrinkles can be seen when inspecting the interior of the blade after infusion. To better look inside the blade, ultra sound is used. Ultra sound is able to penetrate thick glass and as shown below, show a wrinkle.

Figure 8: A wrinkle detected with ultra sound.

Source: Knight & Carver Wind Group®

137

Dry Fiber Fiberglass without resin is useless in a blade and therefore has to be eliminated. As stated previously, fiberglass in the blade receives the resin through an infusion process. Every effort is made to design the infusion so that resin has a path to every part of the laminate. When that doesn’t happen, it is important that the defect be found and dealt with. When the blade has gel coat or paint applied in the mold to the outer skin of the blade, the dry spot is not always evident when the blade leaves the mold. By inspecting the blade at an angle, the naked eye can be trained to pick up these defects. Another way to find dry spots is to use thermographic imaging which will be discussed in a future section of this paper. New fiberglass stitching methods, slits in the core and flow mediums are all ways to improve the ability of the resin to flow.

Figure 9: Dry fiber at the surface directly under the gel coat.

Source: Knight & Carver Wind Group®

138

Bondline Control One of the most difficult steps of the construction process is controlling the bond lines. These glue joints are what holds the entire blade together. As the bond gets thicker, it loses its strength. If the bond line is too thick, there is also the possibility for the bonding paste to be absent, thus no strength at all. Controlling the bond line is accomplished with sophisticated hinge systems and careful quality control. With smaller blades it was not so hard because the parts were small and there were relatively short bond lines. Now with the larger blades the total bond line around the perimeter of the blade and both sides of the double shear web, can reach a total of 250 meters or more. Not only does the bonding paste have to be applied quickly, it must be applied in the right quantities to fill the calculated gap but not so much that the paste is falling down into the blade adding parasitic weight and causing unnecessary material costs.

Figure 10: Mold closing showing the length of the bond lines.

Source: LM Glasfiber

139

Identifying Defects Thermography

A number of techniques have been developed to identify defects in the blades. The first is just a good look with the skilled naked eye. There are a number of things one can spot without specialized equipment. Thermography has become an important means of finding defects in the blades. It is useful finding voids near the surface and dry spots that may not be seen with the naked eye. This method is fairly easy to do after training. A heat source is passed over the blade warming the surface. If a void or dry glass is present, the thermographic camera will detect a “hot spot” and that area will be identified. Later the technician can go back and repair the problem. In the case of a void, adhesive can be injected and the feed hole painted. For a dry spot, the dry material is removed and replaced with the same laminates, then painted.

Figure 11: Hand held thermographic camera.

140

Figure 12: Thermographic image showing a possible void or dry spot.

Source: Knight & Carver Wind Group®

Ultrasound

Ultrasound is a good method to find defects deeper in the laminate. It can detect bond voids or other problems though the thick spar cap or voids in the root. It can be used quite easily after set up. The disadvantage is that the surface must be wet or have a layer of Vaseline on it so the sound waves can penetrate. Water can be fed slowly though a small hose, conveniently applying the right amount of water.

141

Figure 13: Hand operated ultrasound sensor.

Figure 14: Automated ultra sound equipment checking the spar cap.

142

SOLUTIONS TO PROBLEMS Built as Deigned-Avoid open Loop Process Changes It is important that manufacturing follows the specifications. Many times the factory will find ways to “improve” the blade. This input is welcomed; however, there must be a process that assures the idea has gone to engineering and any other departments necessary before the change is released to the floor. This closed the loop process will allow for changes done properly.

Pre-production Testing Many of the problems associated with modern blades could have been determined before production began if stringent tests were performed. Only recently in some cases, have the standards been amended to include fatigue testing. Blades should first be static tested. This verifies the design by checking the frequencies, weight distribution and stiffness. The blade can then be broken to determine the ultimate loads at failure. If required, changes are made and another test can be run. The testing cannot stop here. The blade must be put on a fatigue test stand and tested to the equivalency of at least 20 years. The blades must also be tested in the field collecting load and performance data. Only after these tests confirm the design and manufacturing methods, should the blade be put into production. Many times the schedule is such that some of the testing is cut short. This leaves the blade open for potential problems or failure at the site which are very expensive to remedy.

Trial Fit Before Bonding If the mold was closed and the bond line measured each time, most bonding problems would be eliminated. Not only would the correct bond line be assured, the amount of adhesive would be controlled with a reduction in usage. Since this process is time consuming and adds a step, a compromise might be that the bond line thickness is verified by trial closing at periodic intervals. Thus if the process is wandering, corrections are made before the blade is out of specification.

Bladder Molding/One Piece Infusion Smaller blades are sometimes bladder molded thus eliminating the bonding step. In this process the mold halves are loaded with wet laminates, bladders inserted and the mold closed. When the bladders are inflated, the laminates are pressed in place and the blade comes out of the mold in one piece. If infused, the mold(s) are loaded with dry laminates. When the bladders are inflated, resin is infused. Again, the blade comes out in one piece. Although this system is presently used, engineering and tooling costs are high which prevents easy startup of this process.

Pre-preg Laminates at the Factory Pre-preg fiberglass is fiberglass saturated with the correct amount of resin. This process typically takes place in a controlled environment at a factory removed from the blade factory.

143

The refrigerated material is used as needed. Although the laminate has the best resin to glass ratio, the pre-preg is expensive. This process can be accomplished at the factory where the materials are shipped directly to the factory where the blades are made. The pre-preg is made on a demand basis.

Build blades with more Sub-Components As the blades get larger it becomes more difficult to build the entire blade in one process. Smaller subcomponents can be built separately more reliably and then assembled in the mold later. An example of this would be a root insert that has all of the root attachment inserts molded in. This part is then inserted in the mold during closure and takes away problems of root insert being molded into the large blade half during infusion. Other examples would be tip parts, multiple shear webs and shear webs composed of multiple parts.

Two piece blade The two piece blade is being researched as a solution to the shipping problems of the larger blades. Another advantage is that it is easier to build each of the two smaller parts than the single blade. This is an engineering challenge as the loads must be transferred through the joint.

144

RESEARCH POSSIBILITIES Rapid Blade Design Tool Design tools are needed that can take into consideration dynamic modeling, blade structure, geometry and aerodynamics all at once giving the designer a fast way to optimize the design.

Multi-Piece Blade As discussed, the multi piece blade is an engineering challenge.

Advanced Fabrics The fiber manufacturers need to research fibers that will wet out easier during infusion and have better properties.

Low Cost S-Glass or Carbon Fiber Both S-glass and carbon fiber could prove useful in blade fabrication. Present costs vs. the payback in properties have prevented their widespread use. Lowering the costs could enable more designers to consider these fibers. If the rotor weight could be lowered with carbon fiber, the entire turbine could be made lighter, lowering the cost of the machine,

Component Construction Methods As stated, component manufacturing methods may provide a way to improve the quality of the blades by making it easier to build smaller parts reliably.

Automation of Fiber Placement Laying tons of fiberglass in a mold quickly is a challenge. Automating the process will not only speed up the process but allow repeatability and less chance for variation.

Automation of Manufacturing Processes The remainder of the manufacturing process could be automated again reducing the cost and controlling variability.

Automated Inspection Inspection processes can be automated thus assuring quality. Automation may allow more stringent testing as well.

145

Condition Monitoring Condition monitoring devices can be installed in the blades during manufacture thus lowering the initial cost of utilization. Acceptance by the owners of the turbines and a willingness to install these systems will force the issue.

Smart Blades Smart blades may have active devises installed in them which could deploy during windy conditions making sure the loads are mitigated. Although the machine may pitch the blades, the pitch system is slow compared to wind gusts that load the blades instantly.

Sweep Twist Adaptive Rotor® Although the STAR Blade® has been developed, more research and testing needs to happen before these blades are installed in mega-watt class blades. These blades are allowed to twist to shed loads thus allowing a larger rotor to be utilized. This technology can also be used to lower loads.

146

Figure 15: STAR® rotor on Zond-750 in Tehachapi, CA. Source: Knight & Carver Wind Group®

147

CONCLUSION Blade designs and material selection along with process improvements must keep up with the industry’s demand for larger blades for Mega-watt wind turbines. The industry has come a long way since the 1980s, but there are further improvements awaiting the blades of the future.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Structural Monitoring of Wind Turbine Blades Using Fiber Optic Bragg Grating Strain Sensors

Alan Turner, Tom W. Graver, Micron Optics, Inc., 1852 Century Place, Atlanta, GA 30345 [email protected] Alexis Mendez, MCH Engineering LLC, Alameda, CA 94501

ABSTRACT Over the last few years, fiber optic sensors (FOS) have seen increased acceptance and widespread use in civil engineering, aerospace, marine, oil & gas, composites and smart structure applications. More and more, different research groups and blade manufacturers worldwide have started adopting fiber sensors and fiber Bragg gratings (FBGs) in particular, as practical sensing technology for wind blades. FOS are an attractive technology and reliable sensing solution due to the fact that are completely immune to electromagnetic interference, lightning and electric noise, unlike more conventional electronic sensors that are prone to failure given the harsh and exposed environmental conditions under which wind turbines normally operate. Typically, FBG sensor arrays—either surface-mounted or embedded—have been used to monitor the mechanical behavior of composite rotor blades during the design and qualification stages, as well as in service, to help monitor, on-line, the blades’ condition under rotating, stationary and different wind load conditions. In this paper, will present test field results on the mechanical measurements from an experimental composite blade developed under Sandia Lab’s S-Blade experimental wind turbine program, instrumented with FBG temperature and strain sensors. A discussion of the methodology, on-line monitoring electronic system, and results obtained will be presented. INTRODUCTION All over the world there is growing utilization of wind energy as a clean, environmentally-friendly, alternative energy source. In the U.S. alone, for instance, electricity generation from wind energy is growing rapidly, averaging ~32% per year for the last five years – 2004-2008 [1]; with wind power providing 42% of the entire new U.S. generating capacity added in 2008—up from 35% in 2007—been second only to new natural gas for four years in a row. Furthermore, the increased demand for wind energy is driving the trend toward the construction of bigger turbines, with blade lengths in excess of 60 meters and the ability to produce >7MW of power. As electric utility wind turbines continue to increase in size and initial capital cost, there is an increasing need to ensure their safety, mechanical integrity and durability. This has lead to the adoption and incorporation of structural health monitoring (SHM) system on wind turbines to help optimize their design, operation and maintenance. A key building block on wind turbine SHM systems, are the strain and temperature sensors. Conventional electronic sensors are prone to failure given the harsh and exposed environmental conditions under which wind turbines normally operate. Because of their immune to electromagnetic interference, lightning and electric noise,

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_18, © The Society for Experimental Mechanics, Inc. 2011

149

150

optical fiber sensors have been steadily gaining popularity among wind turbine manufacturers and end-users as a practical, reliable and cost effective on-line structural safety and fatigue monitoring tool, as well as a component pitch control systems. Several studies on their use have already been carried out [2,3] and there are several in service wind turbines around the world that are presently being monitored by such optical systems. Over the last few years, optical fiber sensors have seen an increased acceptance as well as a widespread use for structural sensing and monitoring in civil engineering, aerospace, marine, oil & gas, composites and smart structure applications. Optical fiber sensor operation and instrumentation have become well understood and developed and a variety of commercial sensors and sensing systems are now readily available. In particular, Fiber Bragg Grating (FBG) sensing technology has emerged to be an enabling and practical optical solution for SHM and environmental sensing. FBGs offer many distinct advantages over conventional sensors: (1) immunity to electromagnetic interference, (2) integration into composite materials or concrete with minimal degradation to the host material, (3) significant advantages in cabling compared with electrical sensors due to their smaller dimensions, low weight and cost, (4) capacity of multiplexing many sensors in a long single fiber lead, allowing remote and quasi-distributed measurements, (5) capacity of mass-production with good repeatability, and (6) resistance to corrosion. An initial background review on FBG sensor technology is available in [4].

OPTICAL FIBER BRAGG GRATING SENSING An optical fiber Bragg grating sensing system consists of two key components: sensors and interrogators. A FBG is a segment (~10mm in length) of optical fiber with a periodic variation in the refractive index in the core (see Fig. 1). The Bragg grating acts as light reflector with maximum reflection at a certain wavelength, and its peak wavelength shift is directly proportional to any external induced strain, temperature variation, or pressure effect. Typically, FBG devices have an intrinsic temperature sensitivity of ~10pm/°C, strain sensitivity of ~1.2pm/ε , at a wavelength of 1550nm. Sensitivity of these measurands can be enhanced by proper mechanical package design. Other measurands can be monitored by introducing a proper transduction mechanism or other sensing network designs. FBG sensors are superior in their signal processing simplicity as the information is directly obtained by detecting Bragg wavelength shifts induced by the measurands.

Figure 1. Transmission and reflection spectra of a Fiber Bragg Grating A sensor interrogator measures the FBG sensor response, specifically detecting FBG wavelength shifts with high resolution, accuracy, and speed. Nowadays, FBG interrogation systems have been developed and deployed in many fields of applications. The versatile instruments can interrogate simultaneously hundreds of sensors kHz speeds, with 2pm wavelength stability and sub-pm resolution and repeatability over the 0 to 50°C operating temperature range.

151

OPTICAL MONITORING OF WIND TURBINE BLADES FBG strain sensors make for attractive and suitable devices for monitoring the integrity of wind turbine blades, on line, and detect mechanical damage and asses fatigue; as well as for use as feedback sensors for active blade pitch control. FBG sensor arrays—either surface-mounted or embedded—have been used in the past to monitor the mechanical behavior of rotor blades in different types of wind turbines. These optical fiber sensors provide useful information on the stress and strain levels experienced by the blades under different passive and active load conditions. The sensors help provide valuable information on the structural strength as well as the service fatigue of the blades. Typically, sensors can be used during the design and qualification stages to corroborate the measured strain values against design models. In service, pre-mounted FBG sensors help monitor on-line the condition of the blades under rotating, stationary and different wind conditions, as well as their overall structural health. In addition, FBG strain sensors have also been used as sensing monitors for active pitch control of blades, by effective measuring the blades shape and deflection in real time under wind loading. Table 1 below lists some of the various different applications of FBG sensors in wind turbines. Table 1. Applications and uses of FBG sensors on wind turbines. •

Design Verification & Testing – Ultimate strength & fatigue – Design verification & improvement – Qualification & compliance testing

•

Blade Monitoring – Static & dynamic loads (blade root & full length, wind speed, turbulence) – Surface stresses – Blade deflection (determine shape profile and avoid tower strikes) – Pitch control – Vibrations (frequencies & modal response) – Surface condition (ice, dust, fouling)

•

Tower & Foundation – Loads, bending & torque – Vibration modes

EXPERIMENTAL TESTING—S-BLADE PROJECT Micron Optics was selected by Sandia National Laboratories to participate in a blade instrumentation project— denominated the S Blade, for sensor blade—with the objective to implement sensing technology in a wind turbine blade at the time of blade manufacturing, and to assess the performance of the sensing systems throughout the life of the wind turbine blade [5]. The project involves monitoring of a composite blade with three different commercial systems. For this project, a swept laser FBG sensing interrogator (model sm130) was deployed to monitor strain in the composite blades with 1εand 0.05εrepeatability at 1KHz rate and 1000 averages, respectively. The sm130 can simultaneously monitor 4 fiber-optic channels with >80 sensors per channel, with a power consumption of ~25W. The sm130 is managed by a controller (model sp130) which is an industrial grade, high performance industrial computer that facilitates communication and power control to the sm130 module, and its data processing environment helps engineers to directly extract relevant strain and temperature information on the turbine blades. Particularly important to remote operation, the sp130 also provides power management through wake-on-LAN and wake-on-clock functions, and support remote data transmission through most PC compatible protocols. The range of FBG sensors incorporated in the turbine blades includes non-metallic optical strain gauges (os3200), non-metallic temperature sensors (os4350), and temperature compensation sensors (os4200). The

152

sensor network is configured to incorporate nine os3200, four os4350, and one os4200 on the low-pressure skin side of the blade, and ten os3200, three os4350, and one os4200 on the high-pressure skin side of the blade. The test wind turbine is located in Amarillo, Texas at the US Department of Agriculture, Agriculture Research Service and Wind Energy Technology test site. The first system is conventional foil gages and RTDs (Resistive Temperature Devices) to monitor strain and temperature in the root and along the length of the blade. The second system consists of FBG strain gages, placed beside each of the foil gages (See Figure 2) and FBG temperature sensors installed for temperature sensing and temperature compensation for the FBG strain gages. The third system consists of accelerometers placed on the high and low pressure side of the blade at different locations to monitor the blade’s vibration modes (See Figure 3).

Figure 2. Photograph of a wind turbine blade with two surface-mounted, non-metallic, dielectric FBG strain sensor (circled in white). Note the foil strain sensors located next FBG sensors.

Figure 3. Three accelerometers, one RTD and a FBG based temperature sensor (circled in white). Notice in Figures 2 and 3 that the FBGs are connected in series. There are a total of 14 sensors (9 strain and 5 temperature sensors) on the low pressure side (see figure 4) and 14 sensors (10 strain and 4 temperature sensors) on the high pressure side (see figure 5). The only fiber optic cables that egress from the root of the blade are two 3 mm armored fibers compared with the 30 cables that egress from the foil gages and RTDs. The FBG

153

strain and temperature monitoring system was controlled and monitored from a remote location. All equipment was housed in an environmentally controlled, weather proof enclosure (See Figure 6). SUMMARY High-speed, high-resolution fiber Bragg grating sensing systems have been deployed to optimize the design, operation and maintenance of wind turbines from manufacturing to in-service operation. The multi-channel sensing network can monitor the composite blades with 1εrepeatability at 1kHz rate. All sensors have performed well under load test through the manufacturing cycle, and are operational in field service.

Figure 4. Low Pressure Skin - All sensors are located on one fiber cable and connected to Ch1.

Figure 5. High Pressure Skin - All sensors are located on one fiber cable and connected to Ch2

a)

b)

Figure 6. a) Fiber optic data acquisition and control system. b) The data system mounted on the wind turbine hub (circled in white).

154

ACKNOWLEDGEMENTS The authors wish to thank Mark A. Rumsey and coworkers at Sandia National Lab (Wind Energy Technology Deparment), for their helpful discussions and collaboration. REFERENCES 1. American Wind Energy Association (http://www.awea.org/pubs/factsheets.html), “Wind Energy Basics”, (02/2009). 2. L.W. Rademakers, et al., “Fiber Optic Blade Monitoring”, European Wind Energy Conference, London, 22-25, (11/2004). 3. T.W. Verbruggen, et al., ”Fiber Optic Blade Monitoring”, Energy Research Centre of the Neatherlands, Publication ECN-E-06-034 (2006). 4. A.D. Kersey, M.A. Davis, H.J. Patrick, M. LeBlanc, K.P. Koo, C.G. Askin, M.A. Putnam, and E.J. Friebele, “Fiber Grating Sensors,” J. Lightwave Technology, 15(8): 1442-1463 (1997). 5. M.A. Rumsey, “Sensor Projects at Sandia National Laboratories,” 2008 Wind Turbine Workshop (2008).

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Degradation of Polymer Electrolyte Membranes

Alan Jones*, Jaya Malladi Advanced Materials Laboratory, Department of Mechanical Engineering, Indiana University – Purdue University Indianapolis, 723 W. Michigan St. Indianapolis, IN 46202 email: [email protected]

ABSTRACT Premature failure of polymer electrolyte membranes used in proton exchange membrane fuel cells result in short life and degradation of performance of the fuel cell stack. Changes in the humidity and temperature cause swelling and shrinking of the membrane which result is stresses in the membrane. Stress relaxation and changes in conductivity will occur in the membrane. A novel experimental facility which allows the control of humidity, temperature, load or strain and the simultaneous measurement of the proton impedance has been developed and used to measure the stress relaxation and associated changes in conductivity of Nafion membranes. It was found that at constant strain, both stress relaxation and a drop in the conductivity of the membrane occurs.

INTRODUCTION Damage processes in polymer electrolyte membranes, which are used in fuel cells to separate the reactant gases and conduct positively charged particles from the anode to the cathode of the fuel cell, result in premature failure and degradation of performance of the fuel cell. Membrane materials are chosen due to their high proton conductivity and, just as importantly, their mechanical and thermal stability. Even so, the operational life of the membrane is still rather short and dramatic life improvement is needed to make fuel cells commercially competitive. A number of different models have been created to help understand the fuel cell performance and changes to the performance. Most of the models, for example [1-10], use basic laws of mass, momentum, species and energy to understand the performance of the fuel cell. Very few of the models incorporate aging effects or any type of degradation to the performance of the fuel cell. In addition, most model

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_19, © The Society for Experimental Mechanics, Inc. 2011

155

156

parameters utilize steady state values and do not account for transient conditions or changes in the performance of the components over time. The operating environment of fuel cells used for transportation applications is very severe. The fuel cell is exposed to many start-stop cycles, changes in the reactant gas humidity and temperature, as well as changing power requirements as the vehicle is accelerated and decelerated. The combination of all of these factors results in the fuel cell being subjected to non-steady state environments for a significant portion of its usable life. Currently the most commonly used type of membrane used in proton exchange membrane fuel cells is a perfluorosulfonate acid proton exchange membrane (PFSA), with Nafion from DuPont being the standard. PFSA proton exchange membranes are utilized due to their relatively high proton conductivity and the high chemical stability. Other membranes have demonstrated higher proton conductivity, but lack the mechanical durability to be commercially competitive. On the other hand, many much more durable, and temperature tolerant membranes exist, but their proton conductivity is too low. The mcirostructure of a PFSA membrane is comprised of a fluorocarbon backbone chain with sulfonic acid groups attached to the backbone. The fluorocarbon chains are hydrophobic, while the sulfonic acid groups are hydrophilic. A number of different researchers have developed conceptual models to help understand these networks and explain what happens to the membrane as it is exposed to water. For example, the ionic-cluster model [11], the polymer aggregate model [12], the core-shell model [13], the Eisenberg-Hird Moore model [14], and the bundle-cluster model [15] all provide explanations of the interactions of the hydrophobic and hydrophilic networks. Regardless of the model chosen, when the membrane is exposed to water, either humidity in the air or liquid water, the hydrophilic networks absorbs the water which results in significant swelling of the membrane. During dehydration of the membrane, the free water is removed which contracts the hydrophilic network resulting in shrinking of the membrane. Therefore, as the membrane is exposed to reactant gasses with non-constant humidity, the water content of the membrane will continuously change as the water content changes which will result in significant swelling and shrinking of the membrane. In a typical fuel cell, the proton exchange membrane is sandwiched between the electrodes and bipolar plates which constrain the membrane. This constraint, along with the swelling and shrinking behavior of the membrane will result in stresses in the membrane and, in effect, will result in fatigue loading of the membrane. In addition, the membrane-electrode assembly is fabricated using a hot-pressing technique with the membrane relatively dehydrated. The membrane is then placed in the humidified environment of the operating fuel cell, resulting in swelling of the membrane, a constant strain (if the humidity and temperature is not changed) and stress relaxation The mechanical performance of perfluorosulfonate membranes has been an active area of investigation. The stress-strain behavior at different strain rates, the stability of the membrane under static loading, fatigue loading and hygrothermal loading have all been investigated recently [16-18]. While these studies have shown the mechanical capabilities of the membrane, very few of the studies have evaluated the associated changes to the conductivity of the membrane due to mechanical loading and aging effects. Liu et al. [19] was the first to evaluate the change in conductivity of the membrane after straining the membrane. In this case, the conductivity changes were measured while the membrane was held at constant strain while being submerged in water. The stress relaxation of the membrane was evaluated separately in air at various temperatures. The results of these independent tests were combined to evaluate the effect of strain on the mechanical performance and conductivity of the membrane. The actual conductivity of the membrane being strained in air was not measured.

157

To really understand how the proton conductivity of the membrane is affected by stress and strain, an in situ method of measuring the conductivity of the membrane while it is loaded or exposed to different humidity and temperature profiles is needed. For this purpose, a novel experimental facility has been developed which will allow temperature, humidity, force, or strain to be controlled and the impedance of the membrane to be measured periodically.

RESULTS AND DISCUSSION Figure 1 shows a schematic of the experimental facility used to characterize the mechanical and electrochemical properties of the membrane subjected to different load, temperature and relative humidity profiles. The chamber is installed on a MTS 810 universal loading machine and a 444 Newton load cell is used to measure the force in the membrane. The relative humidity in the chamber is controlled using a humidified purge gas. Dry nitrogen from two different cylinders is used as the feed gasses. The nitrogen from the first cylinder is passed through a temperature controlled water bubbler to humidify the gas. The humidified nitrogen is mixed with dry nitrogen, and the ratio of the wet/dry mix is controlled to achieve the desired humidity. Dry conditions are achieved by circulating only dry nitrogen gas. The humidity in the chamber is measured using a thin-film polymer capacitor probe. The temperature in the chamber is measured with a platinum RTD probe placed near the membrane. A National Instruments PCI-6229 M Series DAQ along with LabView software was used to collect all the data as well as control the tubular heater installed in the chamber. Electrochemical Impedance Spectrometry (EIS) was performed using a Gamry Potentiostat 600 between frequencies of 100kHz to 0.1 Hz.

Figure 1. Schematic of the experimental facility for membrane characterization (MTS control signals, LVTD and computer not shown) The electrodes for the impedance measurement were embedded in polytetrafluoroethlene blocks attached to linear bearings. This allowed the probes to be moved into contact with the membrane during the EIS measurement and removed from the membrane after the measurement was complete. The linear bearings were controlled using non-rotating shaft micrometer heads mounted on the outside of the chamber walls. Figure 2 shows a schematic of the chamber with the embedded probes. This facility allows the impedance testing of the membrane while maintaining the environmental conditions (the chamber door does not need to be opened) and without removing the load from the membrane.

158

Figure 2. Side and front view of the chamber with EIS probes. The large needle bearing (shown on the right side of the front view of Figure 2) is used to move the front probe out of the way of the window. This will allow visual inspection of the membrane during the experiment as well as provide a method for utilizing additional measurement techniques such as laser or optical extensometry or other types of tests. Membranes, 12mm wide and 50mm long, were cut from commercial Nafion 115 sheets using a steel-rule die in the machine direction. Each specimen was acidified by boiling it in 0.5 M H2SO4 for one hour followed by boiling in deionized water for an additional hour. The acidified membranes were stored in deionized water at room temperature before use and allowed to equilibrate with the test atmosphere before loading or impedance measurements. The membrane resistance was obtained by extrapolating the impedance data to the real axis on the high frequency side. Figure 3 shows a typical stress relaxation curve for Nafion at constant humidity of 30% and temperature of 35 oC after it has been stretched to seven percent strain and held at that strain for the duration of the experiment.

159

Figure 3. Stress relaxation of Nafion at seven percent strain, 30% relative humidity at 35 oC. Solid line is the model. The stress relaxation can be modeled using a modified three term Prony. The Prony series is modified to account for the equilibrium stress,

where t1, t2, and t3 are the characteristic times associated with the relaxation processes in the Nafion membrane and σ∞ is the equilibrium stress.

Figure 4. Change in conductivity at seven percent strain and 30% humidity and 35 oC. Solid line is the model The change in conductivity during stress relaxation at seven percent strain, 30% humidity, and 35 oC can be seen in Figure 4. The impedance of the membrane was measured in-situ five different times during the test, and one time before straining the membrane. As can be seen, the conductivity of the membrane decreased to about 25% of its original conductivity over the length of the experiment with a single exponential curve fit indicating that one process is active in the change of the membrane conductivity.

160

CONCLUSION An experimental facility that allows the simultaneous measurement of the load and proton impedance of membranes in prescribed environments was developed and used to characterize the mechanical and electrochemical changes occurring in a membrane at constant strain. It was found that the viscoelastic stress relaxation of the membrane followed a three-term Prony series very well and that the changes in conductivity followed a single exponential decay. Additional experiments on the stress/strain and conduction behavior of the membrane in time varying environments and loading conditions are underway to further characterize the degradation behavior of the membrane. These data will be used to develop a coupled constitutive relationship that will allow prediction of the mechanical and electrochemical degradation of the membrane and guide further development of more robust membranes and membrane-electrode assembly methods.

ACKNOWLEDGEMENTS The authors wish to acknowledge the support of the U.S. Army Research Laboratory (W911NF-07-20036) for support of this work.

REFERNECES 1. Bernardi. Water-balance calculations for solid-polymer-electrolyte fuel cells Journal of the Electrochemical Society, 137:3344-3350, 1990. 2. Bernardi and Verbrugge. Mathematical model of a gas dffusion electrode bonded to a polymer electrolyte. AICHE Journal, 37:1151-1163, 1991 3. Bernardi and Verbrugge. A mathematical model of the solid polymer electrolyte fuel cell, Journal of the Electrochemical Society, 139:2477-2491, 1992. 4. Springer, Wilson, and Gottesfeld. Modeling and experimental diagnostics in polymer electrolyte fuel cells, Journal of the Electrochemical Society, 140:3513-3527, 1993. 5. Nguyen. A gas distributor design for proton-exchange-membrane fuel cells, Journal of the Electrochemical Society, 143:L103-L105, 1996. 6. Bevers, Wohr, Yasuda, and Oguro. Simulation for a polymer electrolyte fuel cell electrode. Journal of Applied Electrochemistry, 27:1254-1264, 1997. 7. Broka and Ekdunge. Modeling the PEM fuel cell cathode. Journal of Applied Electrochemistry, 27:281-289, 1997. 8. Eikerling, Kharkats, Kornyshev, and Volfkovich. Phenomenological theory of electro-osmotic effect and water managment in polymer electrolyte proton conducting membranes. Journal of the Electrochemical Society, 145:2684-2699,1998. 9. Marr and Li. Composition and performance modeling of catalyst layer in a proton exchange membrane fuel cell. Journal of Power Sources, 77:17-27, 1999. 10. Baschuk and Li. Modeling of polymer electrolyte membrane fuel cells with viariable degrees of water flooding. Journal of Power Sources, 86:181-196, 2000.

161

11. Hsu and Gierke. Ion transport and clustering in nafion peruorinated membranes. Journal of Membrane Science, 13:307-326, 1983. 12. Heijden, Rubatat, and Diat. Orientation of drawn nafion at molecular and mesoscopic scales. Macromolecules, 37:5327-5336, 2004. 13. Haubold, Vad, Jungbluth, and Hiller. Nano-structure of nafion: a saxs study. Electrochimica Acta, 46:1559-1563, 2001. 14. Eisenberg, Hird, and Moore. A new multiplet-cluster model for the morphology of fandom ionomers. Macromolecules, 23:4098-4107, 1990. 15. Liu, Kyriakides, Case, Lesko, Li, and McGrath. Tensile behavior of nafion and sulfonated poly(arylene ether sulfone) copolymer membranes and its morphological correlations. Journal of Polymer Science Part B: Polymer Physics, 44:1453-1465, 2006. 16. Tang, Santare, Karlsson, Cleghorn, and Johnson. Stresses in proton exchange membranes due to hygro-thermal loading. Journal of Fuel Cell Science and Technology, 3:119-124, 2006. 17. Tang, Karlsson, Santare, Gilbert, Cleghorn, and Johnson. An experimental investigation of humidity and temperature effects on the mechanical properties of peruorosulfonic acid membrane. Materials Science and Engineering:A, 425:297-304, 2006. 18. Tang, Peikang, Jiang, Wang, and Pan. A degradation study of naffion proton exchange membrane of pem fuel cells. Journal of Power Sources, 170:85-92, 2007. 19. Liu, Hickner, Case, and Lesko. Relaxation of proton conductivity and stress in proton exchange membranes. Journal of Engineering Materials and Technology, 128:503-509, 2006.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

THE NONLINEAR VISCOELASTIC PROPERTIES OF PFSA MEMBRANES IN WATER-IMMERSED AND HUMID AIR CONDITIONS Lei Yan1, Timothy A. Gray1, Kshitish A. Patankar2, Scott W. Case1, Michael W. Ellis3, Robert B. Moore4, David A. Dillard1*, Yeh-Hung Lai5, Yongqiang Li5, Craig S. Gittleman5 1

Engineering Science and Mechanics Department, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061 U.S.A. 2 Macromolecular Science and Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061 U.S.A. 3 Mechanical Engineering Department, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061 U.S.A. 4 Chemistry Department, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061 U.S.A. 5 Electrochemical Energy Research Lab, General Motors Research and Development, Honeoye Falls, NY 14472 U.S.A. ABSTRACT Proton exchange membranes (PEM) in an automotive fuel cell stack can experience significant temperature and hydration changes as the stack responds to the demanding automotive duty cycle. Since mechanical stresses resulting from the hygrothermal cycles are believed to contribute to the loss of mechanical durability that are sometimes experienced in operating PEM fuel cells, it is important to characterize the mechanical behavior of PEMs over a wide range of hygrothermal conditions. In this study, the linear and nonlinear viscoelastic properties of PEMs equilibrated with both humidified air and liquid water are characterized using a custom-built multistation stress relaxation fixture. Specifically, relaxation data of a commercially available, perfluorosulfonic acid PEM was collected over a temperature range of 30-90°C and strain levels from less than 1% to over 20% or more. A comparison of immersed data to dry conditions and a range of humidity levels is presented in this paper. Significant nonlinearity is observed in the membrane, but becomes less pronounced at longer times. Cyclic tests with various strain levels o were carried out on the membranes at 70 C in immersed conditions. The nonlinearity exhibited by the PEM under the larger strain levels was represented quite accurately with a Schapery unaxial hereditary single integral model. For this initial effort, material nonlinear parameters were chosen to simulate the stress output from larger strain levels. Complex loading profiles at various rates were used to validate the model and good agreement was achieved between experimental results and numerical predictions.

* Address correspondence to David A. Dillard, Engineering Science and Mechanics, Virginia Tech, Blacksburg, VA, USA 24061. Email: [email protected]

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_20, © The Society for Experimental Mechanics, Inc. 2011

163

164 1. Introduction and Background Proton exchange membranes (PEM) in an automotive fuel cell can experience significant temperature and hydration changes as the fuel cell responds to the demanding automotive duty cycle. The cyclic mechanical stresses resulting from the hygrothermal cycling of the constrained membrane are believed to contribute to failure of PEMs[1]. The effect of hydration on the mechanical properties of PEMs has received extensive investigation. A relationship between water uptake and strain behavior is also an essential element of a constitutive model for these membranes, as hygrothermal swelling will affect the stress state within the constrained membrane. Bauer et al. [2] measured the mechanical properties of DuPont’s Nafion® 117 using a dynamic mechanical analyzer (DMA) at elevated temperatures and high relative humidity (RH) conditions. They concluded that, although water acts as a plasticizer at lower temperatures, at high temperatures it stiffens the membrane by forming hydrophilic clusters. Tang et al. [3] tested a group of mechanical properties of Gore-Select® PEMs at temperature and humidity combinations, o ranging from 25 to 85 C and 30 to 90% RH. Young’s modulus, the proportional limit stress, and breaking stress were all shown to decrease with increasing temperature and relative humidity. In o a study by Satterfield and Benziger [4], at temperatures greater than 90 C, hydrated membranes were found to be stiffer than dry membranes due to the hydrogen bonding within sulfonic acid clusters. Linear viscoelastic constitutive modeling in the form of relaxation modulus along with hygral shift factors has been successfully used by Patankar et al. [5, 6] to describe the effects of both temperature and humidity on Gore-Select® Series 57 and Nafion® 211. To better understand hydration effects on durability, characterization of PEMs under liquid water saturation as an extreme exposure condition is necessary, as the membrane may experience contact with liquid water during operation. Kyu and Eisenberg [7] compared stress relaxation for Nafion® in an immersed condition and dry state and concluded that faster relaxation occured for immersed samples and the glass transition was profoundly influenced by the presence of water in the structure. Liu et al. [8] conducted immersed stress relaxation on Nafion® membranes in deionized water at 25% and 50% percent strain, and found that the stresses relaxed more rapidly than the proton conductivity at the same strains. A reduction in the dynamic moduli of the material was found by Uan-Zo-Li [9] on Nafion® membranes submerged in water. From the stress-strain o o curves for Nafion®, Kundu et al. [10] reported that at 50 C and 80 C the Young’s modulus and yield strength decreased with complete hydration by one order of magnitude from the dry state because of the plasticization effect of water. Similar significant softening and reduction of load capability in liquid water was also observed by Solasi [11], comparing engineering stress-strain curves over a range of hydration ranging for equilibrium with humidified air to equilibium with liquid water.. Constitutive behavior of Nafion® membranes in liquid water was recently characterized via hyperelastic equations based on uniaxial tension tests[12] by Kusoglu et al. In the current study, the nonlinear viscoelastic properties of a commercial perfluorosulfonic acid (PFSA), Dupont’s Nafion® NRE 211 have been characterized in both humidified air and immersed in liquid water. The Mooney-Rivlin and Ogden models used in [12] did not appear to capture the rate dependent behavior of membranes tested herein in immersed conditions. Exploring alternative nonlinear viscoelastic models, the strain-based Schapery single integral method based on relaxation data appears to model the measured behavior quite well. (We had previously demonstrated that a stress-based Schapery model based on creep-data could be successfully applied to NRE211 membranes tested in humidified air, as reported by Patankar et al[13].) 2. Experimental A double-jacketed environmental chamber with stable temperature control and the ability to be filled for immersion testing was used in conjunction with a stepper motor-driven load frame equipped to test six specimens simultaneously, as shown in Figure 1. Water circulates between

165 two concentric glass cylinders that surround the test chamber by means of a thermostatted recirculating water bath. The air or water temperature in the inner chamber can be controlled by circulating water in the outer jacket. Using the six-station fixture, stress relaxation tests were conducted for samples immersed in liquid water at four temperatures, 30°C, 50°C, 70°C, and 90°C. Displacement-controlled, large deformation uniaxial tensile tests on NRE 211 were also conducted in the six-station fixture in immersed conditions. For both the stress relaxation tests and tensile tests, the specimens were equilibrated for 24 hours prior to testing to allow the membranes to swell in the water prior to loading. This step was important, as swelling strains for the NRE 211 membranes can be on the order of 20%[12], which are significant in comparison to any of the applied mechanical strains. o

Stress relaxation tests at 70 C and 60% relative humidity were also conducted on our six-station fixture. The humid air was introduced from the bottom of the chamber. Tests in humid air and dry o conditions at 90 C were conducted in the humidity chamber of a Q800 dynamic mechanical analyzer (DMA), TA Instruments. Standard tensile clamps were used to load slender specimens.

Figure 1. Six-station relaxation fixture 3. Results and Discussion 3.1.

Environmental effect on relaxation behavior o

Stress relaxation data for samples immersed in liquid water immersion data at 70 C are compared with stress relaxation results at the same temperature obtained in humid air environments in Figure 2. A master curve was constructed with hygral shift factors, a H and the o reference condition is the humid air environment 60%RH. At 70 C, the immersed relaxation modulus E(t) is lower than the modulus for the 60% RH condition. And, the dry membrane is o stiffer than the membrane in either of the hydrated conditions, indicating that at 70 C the water acts as a plasticizer and the plasticizing effect of water is largely affected by the water content of o the Nafion®. At 90 C, plotted in Figure 3, a smooth master curve cannot be built because of the o crossover of relaxation curves at dry and liquid water conditions. At 90 C, in equilibrium with 50%RH air, the membrane became stiffer than in the dry state. This behavior has been reported in the literature [2, 4] and is consistent with the suggestion that water absorption creates hydrophilic clusters that stiffen the membrane. Satterfield and Benziger [4] reported that at

166 o

temperatures above 90 C, the elastic modulus in 100%RH air is greater than in 0%RH air. Our o results indicate that for the immersed condition at 90 C, the relaxation modulus is very close to the dry air condition at high temperature at times less than 100 seconds. At longer periods of time, the modulus in immersed conditions is higher. Interestingly, while the modulus of the hydrated membrane (equilibrated at 50% RH) is greater than the modulus of the dry membrane, a further increase in hydration (to the liquid water equilibrated condition) leads to a lower modulus.

Figure 2. Shifted stress relaxation modulus in immersed, 60% RH, and dry environments shifted o to a reference temperature of 70 C (tests were conducted on six-station unit, and indicated percentages correspond to imposed relaxation strains.)

o

Figure 3. Shifted stress relaxation modulus in immersed, humid, and dry environments at 90 C (results of 50%RH and dry condition were carried out on humidity chamber of DMA)

167 In immersed condition, the tensile stress relaxation modulus, E(t), shown in Figure 4 was obtained over a range of temperatures from 30°C-90°C. A master curve was obtained for NRE o 211 by horizontal shifts of the relaxation curves to reference temperature T ref =70 C. The horizontal shift factor used here was almost identical to the thermal shift factor for a hygral master curve of NRE211 in humid air[6], suggesting a similar temperature effect in humid air and waterimmersed conditions. The shift factors used to construct the relaxation master curves are presented as a function of temperature and were fit to the Arrhenius equation (activation energy o is 225 kJ/K/mol) with the reference temperature of 70 C. No vertical shift was required to obtain the master curves in Figures 4, although the alignment is not perfect, as can be noted by the tails.

Figure 4. TTSP master curve of immersed condition During a stress relaxation test, at a chosen time period, the stress can be plotted against the strain level of the test in an isochronal plot as shown in Figure 5. At each temperature, a stressstrain curve can be created, providing an indication of the linear viscoelastic region of NRE211 in immersed conditions. The nonlinearity of the viscoelastic behavior of NRE211 membranes is strongly temperature and time dependent. Results obtained at shorter times, lower temperatures, or drier states tend to show more pronounced yielding, which is much less evident as the material becomes more elastomeric at longer times, higher temperatures, or greater hydration levels (presumably because water plasticizes the membrane).

(a) (b) Figure 5. Isochronal plots for immersed membrane at two stress relaxation time (a) 100s, (b) 10000s

168 3.2.

Nonlinear viscoelastic characterization and modeling

Displacement-controlled, large deformation uniaxial tensile tests on NRE 211 were conducted in the six-station relaxation fixture in immersed conditions. Figure 6 shows that the rate dependent o stress-strain behavior for Nafion at 70 C does not follow hyperelastic theories that neglect timedependence of the material. A pronounced nonlinearity was observed when an immersed -1 membrane was subjected to large deformations. At the rate of 0.3%∙s , a slope change of the stress-strain curve happened around 10% strain. Cyclic test results were plotted against the ramp test results on a virgin specimen in Figure 7. It is interesting to note that the cyclic curve rejoins the ramp curve quite accurately upon extension beyond the prior maximum showing a memory of the prior maximum strain state, which is often found in filled rubber materials. It may also be observed that a small residual strain remains after unloading the membrane.

o

Figure 6. Rate dependence of water-immersed NRE211 membrane at 70 C

Figure 7. Cyclic stress-strain curves (solid) vs. ramp tension (dashed) To capture the rate dependent nonlinear behavior of NRE211, a general nonlinear viscoelastic model developed by Schapery[14] was implemented in this work. For the case of uniaxial loading under a variable input of strain, the constitutive equation is given by

169 t d (h2ε ) σ (t ) = h∞ E∞ε + h1 ∫ E (ψ −ψ ′) dβ 0 dβ

(3)

E∞ and E (ψ ) are equilibrium and transient components of the linear viscoelastic relaxation modulus, respectively. The reduced time variables, ψ and ψ ′ are defined by where

dξ aε (ξ ) β dξ ψ ′( β , ε ) = ∫ 0 a (ξ ) ε

ψ (t , ε ) = ∫

t

0

(4) (5)

h∞ , h1 , h2 , aε are nonlinear material parameters which are dependent on strain (and temperature and hydration level); aε is also known as strain shift factor. . If the material is linear viscoelastic h∞= h= h= aε= 1 and Eq. 3 is reduced to the Boltzmann superposition integral 1 2 t dε σ (t ) ∫ E (t − β ) dβ . given by, = 0 dβ Here

The steps to implement the Schapery model are given in Figure 8. First, the representative equation of the linear viscoelastic relaxation modulus (corresponding to 2% strain, over the time range 10s-1000s) as shown in Figure 9, was fitted with a Prony series model (Eq. 7). Coefficients of the equation are listed in Table 1. It is worth mentioning that in Figure 9, the dependence of the relaxation modulus on strain level suggests a nonlinearity in the membrane at large strains.

Figure 8. Modeling steps

= E (t )

6

∑Ee i =1

i

− t /τ i

+ E∞

(7)

170

Figure 9. Relaxation modulus for immersed membranes vs. time at 2%, 12% and 27% strain, and Prony series fit of modulus at 2% strain Table 1. Prony series for generalized Maxwell model of linear viscoelastic relaxation modulus for n=6. E i (MPa) 10.14 2.07 7.06 8.27 3.89 14.16 6.93

τ i (s) 10 50 100 500 1000 5000 ∞

To simulate the experimental cyclic tests, a trapezoidal loading and unloading profile was established. The membrane was subjected to a constant rate of strain for 0 25 µm). This type of device is also limited by the packing density of the membrane material, which is a measure of active membrane area per unit volume, and difficulty in obtaining a hermetic seal at the junction of the membrane tubes and tube sheets. Conversely, rolled metal foils are produced by methods that are technologically mature, and can be used in the as-rolled condition. Cross sectional area reductions of up to 90% increase the strength of the material through work hardening, further improving mechanical properties of the material. Rolled material in coil form integrates easily into the fabrication of finished devices using established joining technologies, such as welding, brazing, diffusion bonding, or gaskets to create hermetically sealed components. The design of planar type membrane units is very similar to plate and frame heat exchangers, and offers the same advantages of simplicity and scalability as those devices. Additionally, planar materials and components lend themselves to simplified quality assurance testing using established industry methods and equipment. The metal fabrication industry considers the minimum practical thickness of most rolled alloy foils to be approximately 25.4 µm (0.001 in). There are exceptions to this, most notably the fabrication of consumer grade aluminum foil, which is typically rolled to 20 µm (0.00079 in) in thickness. While the technology exists to roll materials down to 2.5 µm (0.0001 in) and thinner, sufficient economic or market demands have not driven largescale manufacture of alloys in these thickness ranges. Over the past 50 years, Hamilton Precision Metals has established the technological expertise to effectively roll ultra thin (down to 2.0 µm) metal foils in a variety of alloys, including those currently preferred as hydrogen permeable membranes. This technical capability, coupled with sound metallurgical engineering expertise, a strong background in process development, and a continuing commitment to improve its manufacturing advantage relative to the fabrication of these types of materials has helped position Hamilton Precision Metals as a potential industry leader in the fabrication of hydrogen permeable membrane materials. These materials offer developers in the fuel cell industry the most direct path to commercial components that will meet market requirements for the foreseeable future. Stainless steel foil for flexible solar cell manufacturing Photovoltaic technology for producing useable quantities of electric power from sunlight was developed in the 1950’s for powering satellites in low earth orbit. These so called first generation solar cells were made from wafers of electronics grade doped single crystal silicon. Polycrystalline silicon cells soon followed. Although of high efficiency, these first generation cells are rigid and heavy owing to the inflexible silicon used in their manufacture. This then necessitates that the cells be made in a sequential batch type of manufacturing on rigid substrates. Glass is the substrate of choice for first generation photovoltaic cells. About 20 years ago, second generation solar cells were developed. These substitute a thin film of photovoltaic material for the single crystal or polycrystalline silicon. Vapor deposited amorphous silicon, cadmium indium gallium selenide (CIGS) and cadmium telluride (CdTe) thin films have all been developed for photovoltaic generation of electric power. The use of thin films affords several benefits. First, much less active material is needed to make the solar cell, since the thin films are typically only microns thick. Second, since the films themselves are flexible, the second generation cells can be manufactured on a flexible substrate instead of the more common rigid glass substrate used in first generation cells. Use of a flexible substrate makes the cells much lighter and more rugged, expanding applications for their use to military, camping and other recreational activities. More importantly, the ability to make solar cells on long coils of substrate material allows continuous coil-to-coil manufacturing in place of the batch operations required for glass substrates. Continuous coil-to-coil manufacturing is significantly cheaper than batch manufacturing. So, lower amounts of the active materials in the

206

thin film and coil-to-coil manufacturing combine to significantly lower the cost of these photovoltaic cells. Although the second generation cells have somewhat lower conversion efficiency than first generation cells (watts per square meter), their lower manufacturing cost gives them the lowest cost (watts per $ installed cost) for electric power generated of any photovoltaic cell type. To be useful as a photovoltaic cell substrate, several criteria must be met. The material must be available in long coils at thicknesses between 25 and 50 µm at a reasonable price. The substrate must be able to withstand elevated temperatures and mildly corrosive environments during the photovoltaic cell manufacturing operations. It must be stable for 20+ years in outdoor service. The substrate thermal expansion behavior should match that of the active thin film as closely as possible, and it needs to have a glass-like smoothness to minimize the number of inactive areas in the finished photovoltaic cell. Several years of experience has shown that ferritic stainless steel, when properly processed, can meet these requirements. Of the ferritic grades available, type 430 has found the widest application in flexible photovoltaic cell manufacture. Over the past 10 years, Hamilton Precision Metals has been manufacturing type 430 stainless steel substrates for flexible photovoltaic cell manufacture. Substrate material is cold rolled on a 20-high Sendzimir mill to thickness in the range of 25 to 50 µm (0.001 to 0.002 inches). Surface roughness for the substrate material is typically in the 1 to 1.5 µ-inch Ra (25 to 40 nm) range. This extremely smooth surface finish is developed during cold rolling and not in a separate polishing step. By finishing the cold rolling mill work rolls by lapping to a very low Ra value, the smoothness required for photovoltaic cell manufacture is achieved as the stainless steel is rolled to finish thickness. After rolling, the substrate material is cleaned to remove any traces of the rolling mill oil, trimmed in width and cut into coils of several thousand feet in length. These coils are then sent to the photovoltaic cell manufacturer to be made into flexible photovoltaic cells utilizing combinations of CVD and PVD coating processes. Finished photovoltaic cells are then made from the coated substrate. Summary Metal foils have found use in several clean, renewable energy applications. Palladium alloy foils have the ability to purify fuel reformer output for generation of pure hydrogen for fuel cell operation. The thinner the foil membrane is made, the higher the hydrogen flux. This allows for minimum amount of costly palladium in the membrane, lowering the cost of balance of plant equipment for the fuel cell. Stainless steel foils in the 25 to 50 µm thickness range can be used as a flexible substrate for second generation thin film photovoltaic solar cell manufacture. The use of a flexible substrate allows for coil-to-coil manufacturing techniques to be utilized, significantly lowering the cost to produce the solar cells.

Proceedings of the SEM Annual Conference June 7-10, 2010 Indianapolis, Indiana USA ©2010 Society for Experimental Mechanics Inc.

Experimental and Probabilistic Analysis of Asymmetric Gear Tooth

S. Ekwaro-Osire 1, I. Durukan, F.M. Alemayehu Mechanical Engineering Department Texas Tech University Lubbock, TX 79409-1021 ABSTRACT Photoelasticity has been used previously in experimental studies on gears. With the advantages of today’s high capability of computing and software, photoelastic analysis can be combined with probabilistic analysis. The objective of this study was to perform an experimental study and probabilistic analysis of an asymmetric gear tooth. The uncertainty considered involves the material properties and geometry. In this study, different asymmetric gear tooth profiles were manufactured from PSM-5 photoelastic material. . This study shows that the combination of photoelastic experiments and probabilistic analyses may be an effective tool to study the advantages of asymmetric gear over the symmetric gears. BACKGROUND Asymmetric Gears Lately, there have been a number of research activities on spur gears with asymmetric teeth. One of the reasons is that new gear designs are needed for an increasing performance requirements placed on wind turbines, such as high load capacity, high endurance, low cost, long life, and high speed. In wind turbine gearboxes, the gears experience only uni-directional loading since the drive train rotates in one direction. In this instance, the geometry of the drive side does not have to be symmetric to the coast side. Since asymmetric tooth is shaped using different pressure angles on the drive and coast side, it makes asymmetric gears to be the best choice to be used for such an application. In previous studies, related to bending stress and load capacity, high performance has been achieved for gears with asymmetric teeth. If correctly designed, they can make important contributions to the improvement of designs of gears in wind turbine industry [1-4]. Most of the recent research on the benefits of involute spur gears with asymmetric teeth has focused on geometrical design and stress analysis [4-8]. It has been shown that as the pressure angle increases, the root fillet stress and contact stress decrease significantly. In a few studies [9], the effects of various parameters, such as pressure angle and tooth height on the dynamic load and the static transmission errors of spur gears with asymmetric teeth, were investigated. On comparing the spur gears with asymmetric and symmetric teeth, it was shown that for asymmetric teeth, increasing the addendum leads to a significant decrease in the dynamic factor, static transmission error, and root fillet stress. Karpat and Ekwaro-Osire [10] studied the wear of involute spur gears with asymmetric teeth under dynamic loading. They observed an interaction between wear and dynamic loads for spur gears with asymmetric teeth. It was shown that, as the pressure angle on the drive side increases, wear depth decreases considerably. Photoelasticity Photoelasticity is an experimental technique for stress and strain analysis which is specifically useful for systems having complicated geometry, complicated loading conditions or both. For such cases, analytical solutions may be complex or impossible and analysis by an experimental approach may be more appropriate [11]. Certain noncrystalline transparent materials, notably some polymeric plastics, are optically isotropic under normal conditions but become doubly refractive or birefringent when loaded. This effect normally persists while the loads are

1

Corresponding author: [email protected]

T. Proulx (ed.), Experimental Mechanics on Emerging Energy Systems and Materials, Volume 5, Conference Proceedings of the Society for Experimental Mechanics Series 16, DOI 10.1007/978-1-4419-9798-2_25, © The Society for Experimental Mechanics, Inc. 2011

207

208 maintained but vanishes almost instantaneously or after some time interval depending on the material and loading conditions. The photoelastic stress analysis can be implemented using two modes: first, a scale model shaped out of photoelastic material is tested on a loading frame. Such studies are particularly important during product design and evaluation stages. In the second mode, a photoelastic coating is applied on the surface of the component to be analyzed, and tested on reflection polariscope equipment. This mode of study is rather useful for analyzing deformed components [12]. Microcomputers and image processing tools have enabled automated determination of stress fields [12]. Probabilistic Analysis In probabilistic design concept, design variables such as geometry, applied loads, and material properties are defined in terms of their distribution and then statistically defined. Through this, entire system is developed as a probabilistic model and analysis is performed to yield failure probabilities. The concept is that once the final reliability driver contributions are identified, the design can be optimized by altering some of input parameters under given constraints, while maintaining the overall reliability at an acceptable level. The first step in evaluating the reliability or probability of failure of a system is to decide on specific performance function and relevant load and resistance parameters, called basic variables Xi, and the functional relationship among them, corresponding to each performance criterion, mathematically [13]

Z = g ( X1, X 2 ,, X n ).

(1)

The limit state of interest can then be defined as Z = 0, which defines the boundary between safe and unsafe regions in the design parameters space. Failure occurs when Z < 0. Objectives The primary motivation of this study was to contribute to the body of literature on the potential of asymmetric gears. Additionally, there is also a need to develop inexpensive experimental techniques for asymmetric gears such as using photoelasticity. Furthermore, there is also a need to extend probabilistic analysis in the study of asymmetric gears. It is also recognized that the mechanical properties of photoelastic materials change with time. The author investigated how the change in the photolelastic mechanical properties would impact the resulting maximum stress distribution of asymmetric gear tooth. The objective of this study was to perform an experimental study and probabilistic analysis of an asymmetric gear tooth. MATERIALS AND METHODS Experimental Setup and Design For the photoelastic experiment, circular polariscope is used to analyze the stress on tooth profile as shown in Figure 1. The circular polariscope consists of the following main elements; light source polarizer, first quarterwave plate, second quarter-wave plate, and analyzer arranged in the order. Digital image of the loaded gear tooth is captured by a digital camera placed on tripod. Load, that is applied to gear tooth, is measured by a load cell place on the loading mechanism. The photoelastic specimens were made out of PSM material. This material is an annealed polycarbonate plastics specially suited for two-dimensional photoelastic models. Besides having excellent transparency, this material is ductile rather than brittle. A single asymmetric tooth was machined out of this material. The geometry of the asymmetric gear tooth used in this study is shown Figure 2.

209

Figure 1 Photoelastic Experimental Setup

Probabilistic Analysis In this study, ANSYS PDS was used for probabilistic analysis. There gear parameters were cast as random variables. These random variables included Young’s modulus, Poisson’s ratio, and gear thickness (See Table 1)[14]. ANSYS PDS has two kinds of methods, namely, Monte-Carlo simulation and response surface. In this study, Monte-Carlo simulation was selected. Rey et al. [15] noted, apart from the assumption related to the number samples, the Monte-Carlo simulation method does not make any simplification or additional assumptions in the probabilistic model. In this study, for the Monte-Carl simulation was executed with 50 samples. Table 1 Random Variables

No.

Name

Type

Mean

Standard Deviation

1

Elastic Modulus, E Normal 4.49617E+05

44962.

2

Poisson’s Ratio, υ Normal

0.35000

3.50000E-02

3

Tooth Thickness, t Normal

0.24700

2.47000E-03

RESULTS AND DISCUSSIONS Experiments The geometry of the gear tooth profile was constructed in ProEngineer Cad software and imported into ANSYS in “igs” format Figure 2. Plane stress problem was defined using ANSYS with Plane 82 element type. The meshed model had 758 elements (see Figure 3). The batch file of the process was stored and used in the ANSYS PDS module for the probabilistic analysis.

210

Figure 2 Geometry of

Figure 3 Meshed Model

Asymmetric Gear Tooth

Figure 4 shows the fringe patterns for (a) symmetric tooth under 61.10 lb, (b) asymmetric tooth under 62.41 lb on drive side, (c) asymmetric tooth under 72.25 lb on coast side. The experimental result shown on Figure 4b was used for the probabilistic analysis.

(a)

(b)

(c)

Figure 4 Fringe patterns (a) symmetric tooth under 61.10 lb, (b) asymmetric tooth under 62.41 lb on drive side, (c) asymmetric tooth under 72.25 lb on coast side.

Probabilistic Analysis Figure 5 shows the histogram of the random maximum stress. This histogram can be approximated by a normal distribution. The plot shown on Figure 6 reveals the sensitivity of the maximum stress to the variability of the input random variables i.e. the Elastic modulus (E), the Poisson’s ration (υ) and the thickness of the asymmetric gear tooth. The Poisson’s ratio is the most influencing random variable and the thickness is insignificant. The negative values indicate that an increase in the input parameters results in a decrease of the maximum stress.

211

Figure 5 Histogram of Output Stress Parameter

Figure 6 Sensitivity Plot for Input Random Variables

CONCLUSIONS The objective of this study was to perform an experimental study and probabilistic analysis of an asymmetric gear tooth. Asymmetric gear tooth was machined from a photoelastic material PSM-5 and was loaded with 62.41 lb in the polariscope. The experimental results have been used to drive the random variables for the probabilistic analysis. ANSYS PDS was used for the probabilistic analysis. The sensitivity of the maximum stress to the variability of the input random variables i.e. the Elastic modulus (E), the Poisson’s ration (υ) and the thickness of the asymmetric gear tooth were analyzed. Results showed that the Poisson’s ratio was more influencing random variable. The negative values indicate that an increase in the input parameters results in a decrease of the

212 maximum stress. This study shows that the combination of photoelastic experiments and probabilistic analyses may be an effective tool to study the advantages of asymmetric gear over the symmetric gears. REFERENCES 1. Karpat F, Ekwaro-Osire S (2008) Influence of tip relief modification on the wear of spur gears with asymmetric teeth. Tribology Transactions, 51:581-588. 2. Karpat F, Ekwaro-Osire S, Cavdar K, Babalik FC (2008) Dynamic analysis of involute spur gears with asymmetric teeth. International Journal of Mechanical Sciences, 50:1598-1610. 3. Karpat F, Ekwaro-Osire S, Chapman J, Swift A, "Wind power test bed," in WindPower 2007, Los Angeles, California, 2007. 4. Litvin FL, Lian Q, Kapelevich AL (2000) Asymmetric modified spur gear drives: Reduction of noise, localization of contact, simulation of meshing and stress analysis. Computer Methods in Applied Mechanics and Engineering, 188:363-390. 5. Kapelevich AL, Shekhtman YV (2003) Direct gear design: Bending stress minimization. Gear Technology:4449. 6. Kapelevich A (2000) Geometry and design of involute spur gears with asymmetric teeth. Mechanism and Machine Theory, 35:117-130. 7. Di Francesco G, Marini S (2005) Asymmetrical gear wheels: Automatized procedure for the design. VDI Berichte, 1904 II:1735-1742. 8. Brecher C, Schafer J (2005) Potentials of asymmetric tooth geometries for the optimisation of involute cylindrical gears. VDI Berichte, 1904 I:705-720. 9. Karpat F, Cavdar K, Babalik FC, "An investigation on dynamic analysis of involute spur gears with asymmetric teeth: Dynamic load and transmission errors," in Power Transmissions 2006, Novi Sad, 2006, pp. 69-74. 10. Karpat F, Ekwaro-Osire S, "Wear of involute spur gears with asymmetric teeth under dynamic loading," in ASME International Mechanical Engineering Congress and Exposition, Chicago, 2006. 11. Doyle JF, Phillips JW, Manual on experimental stress analysis. Society for Experimental Mechanic, Bethel, Conn. (1989). 12. Ng TW (1997) Photoelastic stress analysis using an object step-loading method. Experimental Mechanics, 37:137-141. 13. Haldar A, Sankaran M, "Probability, reliability, and statistical methods in engineering design." New York John Wiley, 2000. 14. Lolge G, Analysis of a notched bimaterial using an inverse problem method and a probabilistic analysis, Masters, Texas Tech University, 2007. 15. Reh S, Beley JD, Mukherjee S, Khor EH (2006) Probabilistic finite element analysis using ansys. Structural Safety, 28:17-43.