SHAPE CASTING: 4th International Symposium 2011 in honor of Prof. John T. Berry
** GH is the partition coefficient. The non-dimensional parameters are: π
Reynolds number Re =
VH
v
,
1 Darcy coefficient Da = —rK, H2
Froude number rr = Stefan number St =
v
U2
r-,
ßTgGH2
GHcp( — L v
Prandtl number Pr = —, Schmidt number Sc = a Ds In the above expressions v is the kinematic viscosity of the fluid phase, /?, is the thermal coefficient of volumetric expansion, K is the permeability, L is the latent heat at the reference temperature TH, a is the thermal diffusion coefficient and Ds is the solutal diffusion coefficient. The permeability tensor K has been assumed to be diagonal and has been obtained from empirical data and numerical calculations as a function of the volume fraction of liquid φ and the primary dendrite arm spacing rf, [21, 22], The fractional step method is used for the solution of the momentum equation. In this method, the velocity components are solved explicitly and only the pressure equation is solved implicitly. This semi explicit formulation of the momentum equation results in improved computational efficiency and less memory requirements. Full details of the formulation are given in [16], and will not be repeated here. Adaptive Meshing Solidification models that consider thermosolutal or double diffusive convection and channel formation, involve different length scales. A very small solute boundary layer develops ahead of the solidification front due to large (several orders of magnitude) differences in thermal and solute diffusivities. Hence, for accurate macrosegregation computations, proper resolution of the fluid flow close to the tip of the dendrites in the mush zone and in the liquid just ahead of the solidification front is very important. This necessitates the use of very fine meshes in those critical regions. The use of mesh adaptation enables accurate prediction of macrosegregation while reducing the computational cost. In the present work, an adaptive meshing scheme based on linear triangular elements is used. In the mushy zone, only the regions where the liquid metal can still flow leading to the formation of channels and those close to the solidification front are
56
discretized with fine meshes. Deep in the mushy zone, the flow velocities are one or more orders of magnitude smaller compared to the velocities in critical regions and can easily be captured by the coarse mesh. It is known that the channel regions are the last regions to solidify in the whole solidification domain. Hence the gradient of the volume fraction of liquid is also used as an additional criterion for identifying the regions at which finer discretization is needed. In this work, the meshes are generated using an efficient unstructured grid generator AFLR2 developed at Mississippi State University [19, 20]. In this mesh generator, the points (nodes) are created iteratively at the desired spacing using an advancing front type point generation algorithm and the connectivity is optimized locally based on a quality criterion such as minimizing the maximum angle or maximizing the minimum angle. In order to generate the solution-adapted meshes, this mesh generator allows the use of mesh adaptation sources. At these adaptation sources, the point distribution function (or length scale) of the standard automatic point generation algorithm can be modified to a smaller desired value. The mesh is allowed to grow from these adaptation sources, with the value of the point distribution function varying between the small and a predefined large value with a specified growth rate. This algorithm produces a smooth adaptive mesh. In the present problem, initially a coarse mesh with a uniform spacing is generated by prescribing a larger point distribution function (or length scale) for the whole computational domain. Once the solution with the coarse mesh is obtained, the nodes that satisfy the following criteria are chosen for identifying the regions where fine space discretization is needed: 1) the nodes in the mushy zone at which the volume fraction of liquid is greater than a prescribed value and less than one; 2) the nodes at which the gradient of volume fraction of liquid is higher than a prescribed value. All these nodes act as the adaptation sources for the mesh generation algorithm and a smooth solution adapted mesh will be generated. The generated meshes are updated periodically at fixed intervals of solidification time chosen as a function of the solidification speed determined by the cooling rate. Whenever the mesh is updated, the solution variables at the new mesh nodes are interpolated from the corresponding old mesh data. Numerical Results The solidification model with the projection method for solving the momentum equation is implemented by means of a stabilized Petrov-Galerkin formulation based on solution adapted linear triangular finite elements. Numerical simulations for directional solidification of a binary Pb-Sn (23wt.pct) alloy were performed in a two dimensional domain of 30 mm x 50 mm. The domain is enclosed by solid walls with no-slip boundary conditions on all surfaces. Initially, the alloy is all liquid with temperature varying linearly from 546.5 K at the bottom to 596.5 K at the top. The lateral walls are insulated and a constant temperature gradient of 1000 K/m was imposed at the top. At the bottom boundary, the temperature varies with time according to the prescribed cooling rate 1.0 K/min. The properties of the alloy are listed in Table 1. The simulation starts with a coarse mesh of uniform spacing 1 mm. To perform mesh adaptation, the nodes in the mushy zone with a volume fraction of liquid greater than 0.9 and also the vertices of the elements for which the gradient of volume fraction of liquid exceeds 500 m"' are identified as the adaptation sources described above. At these adaptation sources, a value of 0.2 mm is prescribed as the length scale which is the desired smallest elements' size. The mesh starts
57
Figure 1: Directional solidification of Pb-23wt%Sn alloy with adaptive finite elements. Finite element mesh, contours of volume fraction of liquid and mixture solute concentration of Sn after 1165s.
Figure 2: Detail of mesh refinement showing contours of volume fraction of liquid and solute concentration of Sn after 1165s. to grow smoothly from these nodes with a specified growth rate such that the maximum spacing does not exceed the initial coarse spacing of 1 mm at the farthest regions. With this criterion for adaptation, the channel regions and the regions near the solidification front are discretized with fine elements. These channel regions are enriched in solute concentration and are surrounded by solute depleted regions. Figure 1 shows the adapted mesh together with the contours of volume fraction of liquid and solute concentration after 1165 s of solidification time. The result shows the development of the channels along each wall of the domain. In Figure 2, magnified views of the critical regions where the mesh is refined are shown along with the contours of volume fraction of liquid and solute concentration with superimposed discretization. It can be observed that the refinement follows the evolution of freckles quite well.
58
Conclusions A numerical model that can predict the occurrence of freckle defects in directionally solidified castings has been presented. Adaptive re-meshing with linear triangular elements has been used in conjunction with Galerkin finite element method. A simple mesh adaptation strategy that can produce a fine mesh in the critical regions and coarser mesh in other regions has been implemented. The fluid flow equations are solved using an efficient fractional step formulation. Simulation results of the binary Pb-Sn alloy solidification demonstrate the ability of the method to capture the evolution of freckle defects. The method is currently being improved with the goal of capturing freckles in simulations of large castings with complex geometries. Table 1: Thermodynamic and transport properties of Pb-23 wt%Sn alloy used in calculations Reference concentrations (wt pet): Sn=23
Specific heat of liquid (J kg"1 K"1): cpt = 190
Reference temperature (K): TR = 546.494
Specific heat of solid (J kg"1 K"1): cps = 160
Eutectic temperature (K): TE=456
Latent heat of fusion (J kg"1): Z, = 3.76xl0 5 Density of liquid (kg m"3): p ; = 8800
Temperature at which latent heat is given (K) = 528 1
Density of solid (kg m"3): ps = 9700
Thermal expansion coefficient (K" ): βτ = -1.2x10^" 1
Solutal expansion coefficients (wt pet" )ßc = -5.15x10 Thermal conductivity of liquid (W K"1 m"1): κι = 18.4 Thermal conductivity of solid (W K"1 m"1): Ks = 36.8
3
Viscosity (Ns m"2): μ = 2.1736χ10~3 Equilibrium partition ratio: 0.31 Slope of liquidus: -2.32633
Solute diffusivity in liquid (mV 1 ): D = 3x10"9 Melting temperature of the pure substance (K): 600 :200 Acknowledgements This work was funded by the National Science Foundation through grant number CTS-0553570. Support with the AFLR mesh generation software by Prof. David Marcum at Mississippi State University is gratefully appreciated. References 1. A.F. Giamei and B.H. Kear, "On the Nature of Freckles in Nickel Base Superalloys," Metallurgical Transactions, 1 (1970), 2185-2192. 2. S.M. Copely, A.F. Giamei, S.M. Johnson, and M.F. Hornbecker, "The Origin of Freckles in Unidirectionally Solidified Castings," Metallurgical Transactions, 1 (1970), 2193-2204. 3. J.R. Sarazin and A. Hellawell, "Channel Formation in Pb-Sn-Sb Alloy Ingots and Comparison with the System NH4-C1-H20," Metallurgical Transactions A, 19 (1988), 1861-1871. 4. S.N. Tewari, R. Shah, and M.A. Chopra, "Thermosolutal Convection and Macrosegregation," Metallurgical Transactions A, 24 (1988), 1661-1669.
59
5. W.D. Bennon and F.P. Incropera, "A continuum model for momentum, heat and species transport in binary solid-liquid phase change systems-I. Model formulation," International Journal of Heat and Mass Transfer, 30 (1987), 2161-2170. 6. W.D. Bennon and F.P. Incropera, "A continuum model for momentum, heat and species transport in binary solid-liquid phase change systems-II; applications to solidification in a rectangular cavity," Internationaljournal of Heat and Mass Transfer, 30 (1987), 2171-2187 7. C. Beckermann and R. Viskanta, "Double diffusive convection during dendritic solidification of a binary mixture," PhysicoChemical Hydrodynamics, 10 (2) (1988), 195-213. 8. S. Ganesan and D.R. Poirier, "Conservation of mass and momentum for the flow of interdendritic liquid during solidification," Metallurgical Transactions B, 21 (1990), 173-181. 9. D.R. Poirier, P.J. Nandapurkar, and S. Ganesan, "The energy and solute conservation equations for dendritic solidification," Metallurgical Transactions B, 22 (6) (1991), 889-900. 10. S.D. Felicelli, J.C. Heinrich, and D.R. Poirier, "Simulation of Freckles during Vertical Solidification of Binary Alloys," Metallurgical Transactions B, 22 (1991), 847-859. 11. H. Combeau and G. Lesoult, "Simulation of freckles formation and related segregation during directional solidification of metallic alloys," Modelling of Casting, Welding and Advanced Solidification Processes IV. The Minerals, Metals & Materials Society, Warrendale Pennsylvania (1993) 201-208. 12. J. Guo, and C. Beckermann, "Three-dimensional Simulation of Freckle Formation during Binary Alloy Solidification: Effect of Mesh Spacing," Numerical Heat Transfer A, 44 (2003), 559-576. 13. G. Amberg, "Computation of macrosegregation in an iron-carbon cast," International Journal of Heat Mass Transfer, 34 (1) (1991), 217-227. 14. D. Xu and Q. Li, "Numerical method for solution of strongly coupled binary alloy solidification problems," Numerical Heat Transfer A, 20 (1991), 181-201. 15. D. G. Westra, "Simulation of directional solidification in a binary alloy using the fractional step method" (PhD. Dissertation, The University of Arizona, Department of Aerospace and Mechanical Engineering. 2003). 16. J.C. Heinrich, U.K. Sajja, S.D. Felicelli and D.G. Westra, "Projection method for flows with large local density gradients: Application to dendritic solidification," International Journal for Numerical Methods in Fluids, 57 (2008), 1211-1226. 17. U.T. Kämpfer and M. Rappaz, "Modelling of macrosegregation during solidification processes using an adaptive domain decomposition method," Modelling and Simulation in Material Science and Engineering, 11 (2003), 575-597. 18. W. Liu, C. Xie, M. Bellet, and H. Combeau, "2-Dimensional FEM modeling of macrosegregation in the directional solidification with mesh adaptation," Ada Metallurgica Sinica (English Letters), 22 (2009), 233-240. 19. D.L. Marcum and N.P. Weatherill, "Unstructured grid generation using iterative point insertion and local reconnection," AIAA Journal, 33 (1995), 1619-1625. 20. D.L. Marcum and N.P. Weatherill, "A procedure for efficient generation of solution adapted unstructured grids," Computer Methods in Applied Mechanics and Engineering, 127 (1995), 259-268. 21. S. Ganesan, C.L. Chan and D.R. Poirier, "Permeability of flow parallel to dendrite arms," Material Science Engineering A, 151 (1992), 97-105. 22. M.S. Bhat, D.R. Poirier and J.C. Heinrich, "Permeability for cross flow through columnardendritic alloys," Metallurgical and Material Transactions B, 26 (1995), 1049-1056.
60
Shape Casting: The 4Ih International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
A Mathematical Model for Simulating the Microporosity of Squeeze Casting of Aluminum Alloy Zhiqiang Han1, Jinxi Li1, Wen Yang1, Baicheng Liu'· 2 1 Key Laboratory for Advanced Materials Processing Technology (Ministry of Education), Department of Mechanical Engineering, Tsinghua University, Beijing 100084 2 State Key Laboratory of Automotive Safety and Energy, Department of Automotive Engineering, Tsinghua University, Beijing 100084 Keywords: Aluminum Alloy, Squeeze Casting, Microporosity, Modeling and Simulation Abstract A mathematical model for simulating the microporosity of squeeze casting of aluminum alloy has been developed, in which the heat transfer, solidification shrinkage, feeding flow, pressure transfer, and hydrogen conservation were taken into account. The shrinkage induced flow and the pressure drop in the mushy zone were calculated by solving mass and momentum conservation equations. A mechanical model was solved for obtaining the pressure transferred into the central area of the casting. By coupling the pressure drop with the pressure transferred into the central area, the pressure in the mushy zone was calculated. Based on the hydrogen conservation equation, the microporosity volume fraction was estimated by referring to the pressure in the mushy zone. The squeeze casting processes of aluminum alloy under different process conditions were simulated and the simulation results agree well with experimental results. Introduction Squeeze casting is an advanced metal processing technology where solidification is promoted under a high pressure to produce castings with compact interior and excellent mechanical properties [1-3]. It is important to appropriately control the process to avoid the formation of porosity in the casting as the existence of porosity significantly reduces the tensile strength, elongation, fatigue strength, and toughness of the castings [4-6], which fails the attempts to improving casting quality through the squeeze process. Porosity is resulted from volume shrinkage caused by cooling and phase change as well as the precipitation of dissolved gas in the liquid metals. In the squeeze casting, the pressure transferred into the casting decreases with the increasing of the thickness of the solidified shell, and when the pressure inside the casting drops to a certain extent, porosity may form due to insufficient feeding and gas precipitation. The scope of this research lies in modeling and simulation on microporosity of squeeze casting to understand the effect of process parameters on the formation of microporosity. Some efforts have been made on the prediction of microporosity [7-19]. Lee et al [7] and Stefanescu [8] made comprehensive reviews on the research work of predicting microporosity. Kubo and Pehlke [9] proposed an interdendritic flow model, and then many authors made contributions in this field [10-14]. Lee et al [15] developed a pore growth model based upon the diffusion limited growth of pores. Then, they developed a model of pore formation coupled with microstructure simulation using Cellular Automaton method [16, 17]. Backer et al [18] combined the interdendritic flow model with the pore growth model to improve the precision of microporosity prediction. Carlson et al [19] analyzed microporosity nucleation and growth by a
61
volume-average model considering the local, finite-rate diffusion of dissolved hydrogen in the liquid towards the pores, and with that calculated the volume fraction of microporosity. However, the reported work mainly focused on the solidification process under normal pressure. There is few work reported on the microporosity modeling of squeeze casting process. We conducted research on modeling and simulation of microporosity in squeeze casting, aiming at developing a simulation tool facilitating the analysis and optimization of the process design. Mathematical Model Microporosity forms due to volume shrinkage and gas precipitation during solidification. Local volume shrinkage resulted from cooling and solidification in the mushy zone induces feeding flow, and a local pressure drop develops when the volume shrinkage cannot be fully fed. As a result of local pressure drop, the solubility of hydrogen in the liquid reduces and hydrogen precipitation takes place. Moreover, with the solidification carrying on, hydrogen concentration in the liquid of mushy zone increases since the solubility of hydrogen in solid is far lower than that in liquid. It is well understood that the volume shrinkage and hydrogen precipitation are the primary factors resulting in microporosity defect in aluminum alloys. In squeeze casting process, the applied pressure is fully or partially transferred into the casting and creates a pressure distribution inside the component, which has a crucial effect on the microporosity. Hence, the following physics must be taken into account in the modeling: (1) the local pressure drop in the mushy zone induced by the volume shrinkage and feeding difficulties, (2) the transfer of the applied pressure during the solidification, and (3) the conservation of hydrogen and the formation of microporosity. 2.1 Conservation Equations The following assumptions are introduced into the present model: (1) the densities of solid and liquid phase are constant, (2) the liquid flow is lamellar flow and the viscosity of liquid phase is constant, (3) the effect of flow on the thermal field is not considered, and (4) the precipitated gas phase does not move. Base on the above assumptions, the conservation equations of energy, mass, momentum and hydrogen are as follows. The energy conservation: riH p— = V.(iNT) (1) at where T is temperature, κ is thermal conductivity, t is time, H is enthalpy, P = P,g,+ Ptëi+ Ppëpls the average density, where g is volume fraction and the subscripts s,l,p
mean solid, liquid and gas phases, respectively. The mass conservation:
^ + V.(p,F) = 0 (2) ot where V = g,Vl is superficial velocity, g, is the volume fraction of liquid, V, is the velocity of liquid. The density of hydrogen is far less than that of the liquid metal, so we have P = P,g,+P,g, This equation can be further simplified as follows, V.F = - 0 ^ L + - ^ (3) at dt
62
whereß = (p3 - p , ) / p , gives the solidification shrinkage rate. The momentum conservation:
p,^+p,v.i-^J=//iv 2 K+Iv(v.F)j-g / |-F-g / v/'+pg,i
(4)
where P is pressure, μ is viscosity, g i s gravity acceleration, K is permeability described as follows Λ
=-ΐ—^-,-
(5)
180 (1-g,) 2 where λ^ is the secondary dendrite arm spacing. The hydrogen conservation: P,g,C'H + P,g,C'„ + P.g.Cf, = pCl (6) where C°H is the initial hydrogen concentration, C'H, C\, and CH are the concentration of solid, liquid and gas phases, relatively. 2.2 MicroDorositv Formation The formation of pores is judged by using the following criterion: where PG is the gas precipitation pressure, P is the local pressure in the casting during squeeze process, Plhr is reduced pressure caused by solidification shrinkage, Ργ = 2γ/τρ is the additional pressure caused by interfacial energy, where y is the gas-liquid interfacial energy and r is the pore radius. The relationship between the gas precipitation pressure and the hydrogen concentration is described by the Sievert law, (8) C'H=Kiy[p^ where K, is a coefficient depending on alloy composition and temperature and can be calculated by using the following equation [20]: K,=K.lfH (9) where Ke is a parameter depending upon temperature. For aluminum alloy [20]: .og,^=-^*-1.32 / „ is a parameter depending on alloy composition [20]: log,„/ H =Z[ e ;iC; v +^(C, A -) 2 ] Λ'
(„)
where the superscript X denotes alloying element, Cf is the concentration of X, ef,, r^ are the impact factors of X on hydrogen solubility. When the pore has formed, the pressure inside the pore can be described by Pa=P^-P,*r+P,
63
(12)
Normally, microporosity forms in the dendritic array, so it is assumed that the size of the pores equals to the secondary dendrite arm spacing, ! ■ „ = —
(13)
Λ=[(Λ°)3+Λ^
(14)
' 2 where the secondary dendrite arm spacing Xj can be described by the following coarsening model:
where λ°2 is the initial secondary dendrite arm spacing, M is the coarsening coefficient and tf is the solidification time. Numerical Algorithm 3.1 Finite Element Equation The model was solved by using finite element method. The details of finite element discretization and solution method of the energy equation can be found elsewhere [21]. Standard Galerkin method was used for the discretization of the mass and momentum conservation equations. In order to ensure the stability and convergence of the equations, mixed interpolation method was used, where quadratic interpolation function was used for velocity and linear interpolation was used for pressure. The space-discretized equation may be written in the following matrix form ~MU
0
0
0 0
Mv 0
0 0
u V
~Km Km + Km K„
P
cl cl
c„~ II
c. 0
V
P
\F"
= Fv
U
(15)
where M is the mass matrix, K is the velocity stiffness matrix which contains the advection and viscous terms, CT is the divergence matrix and F denotes the force vector. A fully implicit (backward Euler) method was used for the temporal discretization, finally yielding a system of nonlinear equations. A detailed description of the discretization and solution can be found in [22], By solving the equation, the pressure difference in the mushy zone as well as the feeding flow can be obtained. 3.2 Coupling with the Mechanical Model During the squeeze casting process, a thin solid shell forms immediately after the pressure has been applied on the metal by a punch or an upper die. The punch or the die squeezes the casting and keeps the solid shell continuously deformed, and a static pressure develops in the liquid or mushy core of the casting. At the same time, volume shrinkage happens near the outer part of the casting and feeding flow from the center to the outer part takes place. The static pressure inside the casting was calculated by using a mathematical model developed earlier [21, 23] for describing the deformation and stress of squeeze casting. On the other hand, the pressure difference in the mushy zone was calculated by solving the mass and momentum conservation equations. The absolute pressure value in the mushy zone was determined by referring both the
64
static pressure and the pressure difference, which was used for judging whether a pore forms and for calculating the volume fraction of porosity. 3.3 Calculation of Volume Fraction of Microporositv When the porosity has not formed ( gp = 0 ), the volume fraction of solid phase can be calculated directly by g =
J
i— =
K+V,
=
· ^'
f./p.+f./p,
il!-L
(16)
f,P,+f,P,
where Vs and V, denote the volume of solid and liquid phases. The mass fraction of solid phase fs can be calculated according to temperature. The hydrogen redistributes during solidification: C'H=kHC'„
(17)
where kH is the partition coefficient of hydrogen. By using Eq. (6) and Eq. (17), the hydrogen concentration of liquid phase can be calculated (18)
C'„ = pC°H/(kHp,g,+plgl)
By using the Sievert law, the hydrogen precipitation pressure can be calculated and used to judge whether the porosity forms based on the criterion defined by Eq. (7). When the porosity has formed, the initial volume fractions of solid and liquid phases, g] and g* can be calculated by .
V
1
V,
ν,+ν,'>gi = v,,s+v,,.
(19)
The relationship between gt andgj, as well as g, andg*, is as follows:
ν. + ν, + ν,
'
v,+v, + vP
(20)
The density of porosity can be calculated based on the ideal gas assumption: P p=a— p Ύ
(21)
r
where a is a constant. As only hydrogen is considered, the hydrogen concentration in the pore is CH = 1. By substituting Eq. (6) for Eq. (8), Eq. (20) and Eq. (21), the volume fraction of microporosity can be calculated as follows "
{p,g',+p,g',)c°H -{kHp,gl '
apG/T+(PX
+psg;)c°H-(kHp,g;
+p,gl)K,4^ +p,gl)K,4FG
(22)
When the solid fraction reaches a certain extent, the dendrites grow into a close skeleton, which makes the liquid and solid phases separated from each other. As a result, the fraction of microporosity calculated using pressure drop in the mushy zone has a large deflection from the
65
experimental data. In this paper, it is assumed that the feeding flow is cut off when the solid fraction is higher than a critical value g and the volume fraction of porosity can be calculated as follows gP=g'P+{\-gAi-g'P)ß
(23)
where g' is the volume fraction of microporosity wheng 5 = glc. Simulation Examples A casting of A3 5 6 aluminum alloy with symmetric geometry was simulated. Fig. 1 shows the casting geometry and simulated results. The vertical edge at the left is the symmetric line of the casting. In the figure shows the calculated velocity of feeding flow and the pressure distribution in the castings with a pressure of 40MPa applied. The pressure distribution is not shown in the area that has already solidified in order to display the pressure change in an appropriate scale for the mushy zone. As shown in Fig. 1, when the pressure has been applied on the casting for 10s, the upper and right part of the casting has almost solidified, and the liquid metal flows from the central area to the mushy zone adjacent to the solidified part. The casting with low die temperature solidifies more quickly than the casting with high die temperature. The pressure in the center of the casting with low die temperature is approximately 9.5MPa while in the casting with high die temperature is 18.5MPa, as shown in Fig. 1 (a) and (b). At the time of 15s after the pressure is applied, most part of the casting with low die temperature has solidified, and the pressure in the center is around 0, which implies that the applied pressure cannot be transferred into the casting effectively whereas in the casting with high die temperature, a large portion of the central area has not solidified yet, where a pressure of 14MPa still exists, as shown in Fig. 1 (c) and (d).
Fig. 1 The calculated flow velocity and pressure distribution in the casting at 10 and 15s after the pressure is applied. The process parameters: applied pressure 40MPa, punch temperature 100"C, die temperature 150'C and ejector temperature 150'C for (a) and (c); applied pressure 40MPa, punch temperature 200°C, die temperature 250°C and ejector temperature 250°C for (b) and (d). Fig. 2 shows the calculated feeding flow velocity and pressure distribution in the castings with pressures of 40 and 60MPa applied. The pressure in the central area of the casting with a pressure of 60MPa applied is approximately 18MPa while that with a pressure of 40MPa applied is about lOMPa. It is shown in Fig. 2(c) that at the time of 22s after the pressure is applied the pressure of the unsolidified area is inadequate for preventing the porosity from being formed for the casting with a pressure of 40MPa applied. In contrast, in the casting with a pressure of 60MPa applied, a pressure of 8MPa in the unsolidified area still exists. Fig. 3 shows the calculated results of microporosity in the casting with different process parameters. In the case with low die temperature, a pressure of 40MPa is not sufficient to ensure a sound casting. There is some microporosity in the hot spot part of the casting, see Fig. 3(a). The casting with higher die temperature has less microporosity, as shown in Fig. 3(b), indicating
66
that an appropriate increasing of the die temperature is helpful to the pressure transfer and further to microporosity reduction. When the applied pressure increases to 60MPa, there is almost no obvious porosity in the whole casting. These results agree well with experimental results.
Fig. 2 The calculated flow velocity and pressure distribution in the casting at 18 and 22s after the pressure is applied. The process parameters: applied pressure 40MPa, punch temperature 200°C, die temperature 250°C and ejector temperature 250'C for (a) and (c); applied pressure 60MPa, punch temperature 200'C, die temperature 250°C and ejector temperature 250'C for (b) and (d).
(a) (b) (c) Fig. 3 The calculated volume fraction of microporosity in the castings. The process parameters: (a) applied pressure 40MPa, punch temperature 100 "C, die temperature 150"C and ejector temperature 150'C; (b) applied pressure 40MPa, punch temperature 200°C, die temperature 250 'C and ejector temperature 250*C; (c) applied pressure 60MPa, punch temperature 200'C, die temperature 250'C and ejector temperature 250°C. Conclusions Based on the understanding on the formation mechanism of microporosity, a mathematical model for simulating the microporosity of squeeze casting of aluminum alloy has been developed, in which the heat transfer, solidification shrinkage, feeding flow, pressure transfer, and hydrogen conservation were taken into account. The shrinkage induced flow and the pressure drop in the mushy zone were calculated by solving the mass and momentum conservation equations. A mechanical model was solved for obtaining the pressure transferred into the central area of the casting. By coupling the pressure drop with the pressure transferred into the central area, the pressure in the mushy zone was calculated. Based on the hydrogen conservation equation, the microporosity volume fraction was estimated by referring to the pressure in the mushy zone. The squeeze casting of aluminum alloy under different process conditions was simulated and the simulation results agree well with experimental results. Acknowledgement The research work is funded by the National Natural Science Foundation of China (No. 50675113 and No.50875143). One of the authors ZQHAN also would like to appreciate the support of the
67
Scientific Research Foundation for the Returned Overseas Chinese Scholars, Ministry of Education of China, and the support of State Key Laboratory of Materials Processing and Die & Mould Technology, Huazhong University of Science and Technology. References [1] M. R. Ghomashchi, A. Vikhrov, Journal of Material Processing Technology, 2000, Vol.101 (l),ppl-9. [2] P. X. Qi, Special Casting and Nonferrous Alloys, 1998, Vol.4, pp 32-36. [3] S. J. Luo, B. G. Chen, P. X. Qi, Liquid Forging and Squeeze Casting Technology, Beijing: Chemical Industry Press, 2007, pp 1. [4] J. F. Major, AFS Transactions, 1998, Vol.105, pp 901-906. [5] A. M. Samuel, F. H. Samuel, Metallurgical Transactions A, 1995, Vol.26, pp 2359-2372. [6] C. D. Lee, Materials Science and Engineering A, 2007, Vol.464, pp 249-254. [7] P. D. Lee, A. Chirazi, D. See, Journal of Light Metals, 2001 (1), pp 15-30. [8] D. M. Stefanescu, International Journal of Cast Metals Research, 2005, Vol.l8(3), pp 129143. [9] K. Kubo, R. D. Pehlke, Metallurgical Transactions B, 1985, Vol.16, pp 359-366. [10]D. R. Poirier, K.. Yeum, A. L. Maples, Metallurgical Transactions A, 1987, Vol.18, pp 19791987. [11]S. Shivkumar, D. Apelian, J. Zou, AFS Transactions, 1990, Vol.98, pp 897-904. [12]A. S. Sabau, S. Viswanathan, Metallurgical and Materials Transactions B, 2002, Vol.33(2), pp 243-255. [13]C. Pequet, M. Gremaud, M. Rappaz, Metallurgical and Materials Transactions A, 2002, Vol.33, pp 2095-2106. [14]H. D. Zhao, C. Z. Wu, Y. Y. Li, I. Ohnaka, Acta Metallurgica Sinica, 2008, Vol.44(ll), pp 1340-1347. [15]R. C. Atwood, S. Sridhar, W. Zhang, P. D. Lee, Acta Metallurgica, 2000, Vol.48(2), pp 405417. [16] P. D. Lee, A. Chirazi, R. C. Atwood, W. Wang, Materials Science and Engineering A, 2004, Vol.365, pp 57-65. [17]J. S. Wang, P. D. Lee, International Journal of Cast Metals Research, 2007, Vol.20(3), pp 151-158. [18]G. Backer, Q. G. Wang, Metallurgical and Materials Transactions B, 2007, Vol.38, pp 533540. [19]K. D. Carlson, Z. P. Lin, C. Beckermann, Metallurgical and Materials Transactions B, 2007, Vol.38, pp 541-555. [20] P. N. Anyalebechi, Acta Metallurgica, 1995, Vol.33, pp 1209-1216. [21]Z. Q. Han, W. Zhu, B. C. Liu, Acta Metallurgica Sinica., 2009, Vol.45, pp 356-362. [22] R. W. Lewis, Z. Q. Han, D. T. Gethin, Comptes Rendus Mécanique, 2007, Vol.335, No.5-6, pp 287-294. [23] W. Zhu, Z. Q. Han, B. C. Liu, Acta Metallurgica Sinica, 2009, Vol.45, pp 363-368.
68
Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
SHAPE CASTING: 4th International Symposium 2011 in honor of Prof. John T. Berry
Solidification Session Chairs: William Griffiths Peter Schumacher
Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
REVIEW OF DEFECT BEHAVIOR IN Ni-BASED SUPERALLOYS John Campbell Emeritus Professor, Metallurgy and Materials, University of Birmingham, B15 2TT, UK, je @ campbelltech.co.uk. Abstract The Ni-base superalloys, normally melted and cast in vacuum, entrain their surface oxide film during turbulent pouring of the melt, which unfortunately at this time, is universally practiced for investment castings of these materials. The entrained film automatically becomes a bifilm crack, so that cast alloys have a large population of cracks that controls their failure behavior. The problems of the growth of single crystals, and the welding of polycrystalline alloys are reviewed to illustrate the central role of bifilms in the cracking of turbine blades, the heat affected zones of welds and the reliability of properties. It has been demonstrated that improved gravity pouring systems can significantly reduce these problems, but only counter-gravity filling of molds is expected to result in defect-free castings. Key words: Ni superalloys; cracks; single crystals; welds; HAZ. Introduction There is a growing body of evidence that Ni base superalloys harbour cracks because of the poor casting techniques that are currently used to shape these materials [1-3]. These defects arise naturally during the turbulent pouring of metals (Figure 1). The importance of this subject was confirmed to the author in the tragic case of the last failed turbine blade that he examined that had caused the plane to crash, costing lives. While it is acknowledged that in other cases engine manufacturers go to great lengths to ensure a 'blade off event does not cause the engine to fail, it seems not helpful to continue to put huge efforts into metallurgical research on alloy development while ignoring the major casting defects necessarily introduced by current manufacturing techniques. The defects in metallic liquids and final castings are principally bifilms. They appear to give rise to a wide spectrum of phenomena including porosity, hot tearing, cold cracking, stray grain initiation, fatigue initiation and corrosion. This is a formidable list. Bifilms are a serious issue that so far has been overlooked or ignored. Bifilms are created easily and rapidly during casting. The surface of the melt oxidises rapidly (whether in air or in so called 'vacuum') so that when folded in, or when experiencing collisions between droplets, the surface oxide contacts dry-face-to-dry-face when impinging against other masses of liquid. The resulting unbondable interface, as a double film called a 'bifilm', is then entrained in the bulk liquid as a crack. Our existing turbulent pouring systems fill the liquid metal with cracks. The defects remain in suspension sufficiently long to become frozen into the casting.
71
At this time, the preparation of superalloys and the manufacture of the majority of turbine blades involves pouring of the liquid metal from considerable heights into molds designed with poor filling systems; the melt experiencing energetic turbulence, guaranteeing that most if not all Ni-base feed stock, and Ni-base blades will contain large populations of defects, some of serious size. For instance nearly the whole of the fracture surface seen in Figure 2 is covered with an oxide bifilm (EDX spectra from this film are added in Figure 5). Although many blades are cast by bottom-gated designs of filling system, these systems are not necessarily free from problems, and in any case the damage by the trauma of the prior pour event usually cannot be reversed. Similarly D/S and single crystal blades are grown in a temperature gradient with a relatively quiescent, planar front, but they too suffer irreversible damage caused by the severely turbulent filling of the mold. The general belief that melting and casting in vacuum avoids the problems of oxidation during casting is seen in general to be an unfortunate and serious error, (even though there is recent evidence [4] that in some conditions a high quality vacuum might avoid this problem). The design of vacuum melting and casting furnaces used for practically all investment cast Ni-base alloys enhances the turbulence problem because of the significant fall of metal from the lip of the melting crucible to the mouth of the mold. This height is often in excess of a meter, whereas it is known that any fall distance greater than about 10 mm for heavy metals can cause entrainment of the surface film and the consequent formation of bifilm cracks [1]. For wrought Ni-base alloys, the situation is even worse, with the melts falling in air through heights of several meters, via poorly designed ceramic channels that ensure the mixing of large quantities of air into the melt during the casting of bottomfilling of tonnage-sized ingots. The ingots are destined for subsequent hot plastic working such as forging, rolling or extrusion. These processes ensure that the entrained bifilms may become tightly closed, making them more difficult to detect, but the bifilms will be expected to be resistant to welding. The working processes will extend the length of the defect, but residual air trapped in folds and creases of the bifilm is likely to continue oxidation or nitridation of freshly extended surfaces, preventing welding until the reservoir of air is finally consumed. The bifilms introduced by turbulence during casting are usually invisible, or at least difficult to see, as a result of their extreme thinness, often measured in nanometres. This contrasts with their impressive surface areas, sometimes measured in square millimetres or even square centimetres. These extensive natural cracks are probably the most important defects in both cast and wrought metals. Evidence for bifilms in metals in general is summarised elsewhere [1-3]. 'Brittle' intermetallics and second phases. The wetted outer interfaces of oxide bifilms (those surfaces from which the oxides grew, atom by atom, and so in perfect atomic contact with the melt) appear to be favoured substrates for the precipitation of second phases in a wide variety of different matrices, including Al, Cu, Ni and Ti alloys [3].
72
In particular, in Ni-base alloys the association of cracks with so-called 'brittle' grain boundary phases has been observed by many workers [5,6]. It has been observed [6] that the outside surfaces of the bifilms act as favoured substrates for the precipitation and growth of carbides containing Cr, Ti, W and Mo. The cracks in the carbides are the visual reminder of the presence of the bifilm that initiated the formation of the carbide. The presence of the bifilm is usually not detectable by optical inspection and therefore unsuspected, but becomes completely clear at higher magnifications of the fracture surface observed in the scanning electron microscope (SEM). Rashid found that the bifilms themselves were oxides rich in Cr and Al (Figures 3 and 5). This study was carried out on an industrially cast blade for power generation made under similar conditions to an aero-engine blade. Exactly similar features were found in blades for aero engines cast in China in 2005 [7]. Furthermore, it is worth emphasising that the carbides themselves are not expected to crack, since it is likely that they will be extremely strong. The tensile failure of the casting would take place by a crack that followed the unbonded bifilm interfaces, apparently following the carbide cracks, giving the appearance that the carbides have caused the failure by their brittleness. D'Souza [8] observes linear features in the microstructure of his CMSX4 on which Nb, Zr and Cr rich phases have formed. He indicates that the composition in which these phases occur is the most likely to form hot tears. Both these observations are consistent with the presence of oxide bifilms. Furthermore, the stray crystals seen in his single crystal castings may have origins associated with the presence of bifilms acting as barriers to the advance of grains as described below. In polycrystalline superalloys Qin et al. [9] observed that long term thermal exposure created chains of carbides that formed pathways for the spread of cracks, and the subsequent deterioration of properties. Once again, the carbides would be expected to have precipitated on oxide bifilms, and so effectively would be pre-cracked; the long term thermal exposure merely gradually opening these features, possibly by the increased pressure caused by the expected precipitation of hydrogen into the cracks, and the creep of the surrounding solid to allow some slight opening. Sidhu et al. [10] also note intergranular cracks associated with continuous films of M23C6 and MC carbides in their welded and heat treated Inconel 738LC alloy. Similar interdendritic carbides are seen associated with film-like defects on fracture surface of a Co base alloy [11]. The report by Mälzer and colleagues [12] on the creep failure of single crystal superalloy LEK94 clearly shows films (assumed to be oxides, but possibly nitrides) on fracture surfaces and microcracks associated with as-cast pores. One would expect pores and bifilm-type cracks to be associated, because both initiate from entrainment mechanisms during casting [1]. In fact bubbles and bifilms are hard to differentiate at times; both are entrained defects; the major difference being the amount of gas that each contains. However, the difference in their gas contents is sometimes not clear, blurring the distinction between them. Also, in passing, it is worth noting that pores and cracks are not solidification defects but casting defects.
73
Grain Boundary Phenomena The commonly accepted reason for the approximately three orders of magnitude benefit of creep life for directionally solidified Ni base alloy structures compared with conventional equiaxed structures is the absence of transverse grain boundaries, the assumption being that these boundaries are weak, resulting eventually in decohesion. Since bifilms are to be expected in gravity poured castings, and because their preferred siting will be between dendrites and grains (cannot penetrate the microscopic air layer between the films so that they are mainly pushed ahead into interdendritic and intergranular spaces) the presence of bifilms as invisible unbonded interfaces easily explains the rupture of transverse boundaries. Moreover, as in the case of intermetallics, there is little reason to suppose that grain boundaries are actually weak; it is almost certain that they are extremely strong, even if not quite as strong as the matrix. Crack formation along the longitudinal grain boundaries of directionally solidified Ni base superalloys during solidification has been attributed to so called grain boundary decohesion in the absence of any really consistent explanation resulting from many studies over past decades [7]. Once again, the presence of bifilms is to be expected, and can be predicted to result in grain boundary cracking [1,3]. Furthermore, these authors found that stray grain formation was increased, once again likely to be the result of a higher density of bifilms, or possibly mechanically stronger bifilms, that mechanically obstruct the advance of the desired single crystal. The blocked advance of the dendrite front, while the withdrawal of the mold continues, will ensure that the liquid above the blockage will progressively undercool as it is withdrawn down the furnace temperature gradient. Eventually the undercooling will become sufficient to nucleate a new grain. The bifilm, with its internal layer of air, will ensure that the new grain will have no benefit of contact with the blocked original crystal, with the result that the new grain will have a totally independent growth orientation. In agreement with this proposed mechanism, Carney and Beech [13] identified oxides at the root of most of the stray grains in single crystals. Furthermore, the incidence of stray grains was reduced by filtering the metal. Welding of Ni-Base Superalloys Up to this time it has been understandable that most authors studying welding have overlooked the probability that oxide bifilms will be present in their alloys, so that the materials that they study are already effectively pre-cracked. Welding provides the opportunity for the cracks to open and become visible. The author [14] has suggested a mechanism for the damaging effect that incipient grain boundary melting has in the heat affected zone (HAZ). For instance, if a grain boundary phase melts in the heat of the weld, but subsequently re-solidifies, why should the properties not be fully recovered, if not improved, as a result of the rapid freezing and consequent fine structure? If a bifilm occupies the grain boundary, the melting of a
74
nearby phase will be associated with (usually) an expansion, with the necessary plastic yielding of the matrix. However, on re-freezing, the volume contraction will merely open the bifilm, creating an open crack, and thus lowering properties (a closed bifilm can at least support some shear stress as a result of friction between the surfaces, the nonplanarity provided by jogs and folds). The differing phases on either side of the crack are another feature to be expected of a bifilm crack; many bifilms are asymmetrical, with one thick side consisting of an older film, often having a spinel structure, whereas its opposing side consists of a pure oxide with a different structure. These differing sides favor the precipitation and growth of different phases during the solidification of the surrounding alloy. Wang at al. [5] present data comparing the behavior of laser beam welding of two single crystal alloys, CMSX-4 and -486, finding increased cracking in the high Zr and Hf -486 alloy. This seems likely to be a result of the higher reactivity of Zr and Hf with oxygen, strengthening the oxide film and enhancing its damage potential during entrainment. Unreliability of tensile properties The statistical chance of oxides being folded in by chance events of turbulence to create scatter in the tensile properties of Ni-base superalloys has been clearly shown by the work of Cox et al [15] who compared top-filled with bottom-gated molds (Figure 4) for alloy IN939 in the hipped condition. Interestingly, both casting techniques show some scatter at lower strength levels, indicating the filling system designs used in this work could be improved. This seems typical of gravity filling where it is difficult or impossible to suppress the formation of all damage. Concluding Remarks Nearly all the above work has been carried out on alloys poured freely under gravity into molds of various kinds and so is expected to contain a generous quantity of bifilm cracks. The rules for the design of gravity casting techniques to avoid the entrainment of the oxidised surface during the filling of the mold have been developed over recent years [2]. Different authors [15,16] have demonstrated that the application of these rules for gravity pouring can reduce the number of casting defects by a factor of 10. At first sight the achievement of a reduction in defects by a factor of 10 might seem impressive. However, if metals were cast with a good counter-gravity technique the factor would be expected to approach infinity. This is because the number of entrained defects can, in principle, fall to zero [2]. It is to be hoped that both researchers and industry will convert to either improved gravity pouring systems, or better, countergravity for the future casting of Ni base and other alloys. Finally, it is worth emphasising that bifilms in cast metals and their remnants in wrought metals probably permeate nearly all our engineering materials. However, they need not be present. Although there are now improved designs of gravity filling systems for castings, any system of alloy production or casting manufacture that involves the pouring
75
of metals involves the production of damage to the liquid metal. A degree of unreliability in properties and performance in the final cast material seems then unavoidable. Our technology for the production of such safety critical components urgently requires to be changed. Only the avoidance of pouring, by proper design of tilt techniques, or preferably full counter-gravity handling of melts and filling of molds, will ensure perfectly reliable metals. This technology is already available and proven [2]. The transformation of many of our engineering materials into materials without bifilms has potential to bring a revolution in properties and performance. References 1. J. Campbell: 'Castings', 2nd edn; 2003, Elsevier, Oxford, UK. 2. J. Campbell: 'Castings practice - the 10 rules for casting' 2004, Elsevier, Oxford, UK 3. J. Campbell: Mater. Sci. TechnoL, 2006, 22, 127-145; discussion 2006, 22, 999-1008 4. D Giuranno, E Ricci, E Arato, P Costa; Acta Materialia 2006 54 2625-2630 5. Y. L. Wang, O. A. Ojo, R. G. Ding and M. C. Chaturvedi: Mater. Sci. TechnoL, 2009, 25, 68-75. 6. K. M. B. Rashid and J. Campbell: Metall. Mater. Trans., 2004, 35A, 2063-2071. 7. Huang Aihua: Proc. 68th World Foundry Cong. 2008, 215-218 8. N. D'Souza: Mater. Sci. TechnoL, 2009, 25, (2), 170-185. 9. X. Z. Qin, J. T. Guo, C. Yuan, C. L. Chen and H. Q. Ye: Metall. Mater. Trans., 2007 38A, (12) 3014-3022. 10. R. K. Sidhu:, N. L. Richards and M. C. Chaturvedi; Mater. Sci. TechnoL, 2007, 23, (2)203-213 11. Montero-Ocampo, M. Talavera and H. Lopez: Metall. Mater. Trans., 1999, 30A, 611-620 12. G. Malzer, R. W. Hayes, T. Mack and G. Eggeier: Metall. Mater. Trans. 2007, 38A, (2), 314-327 13. A. Carney and J. Beech: Proc. Solidification Processing Conf., Sheffield, UK, 1997, University of Sheffield, (ed. J. Beech and H. Jones), 33-36. 14. J Campbell; Mater Science & Technol 2000 25 (1) 125-126 15. M. Cox, M. Wickins, J. P. Kuang, R. A. Harding and J. Campbell; Mater. Sci. TechnoL, 2000, 16, 1445-1452 with additional personal communications from Cox reported in [1] pp 57-61. 16. Z. Li, J. Campbell and Y. Y. Li: J. Processing TechnoL, 2004, 148, (3), 310-316.
Figure 1. Entrainment of a bifilm in a liquid metal.
76
Figure 2. Fracture surface of a turbine blade casting.
Figure 3. Close up of fracture surface showing line scan confirming O, Al and Cr in oxide region. 'Brittle' carbide seen on right, precipitated on oxide [6].
Figure 4. Two-parameter Weibull plot of vacuum melted and cast top poured (squares) and bottom filled (circles) of IN939 test bars into investment molds [15].
77
78
Shape Casting: The 4' International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
P r e m i u m Quality Super Duplex Stainless Steel Castings Without Secondary Refining Bob Puhakka Alloy Casting Industries, New Hamburg, Ontario Canada Keywords: super duplex stainless steel, oxide bifilms, naturally pressurized fill system Abstract The ASTM A890/ A995 (25Cr 7Ni 3.5Mo) super duplex stainless steels are a popular family of high alloy cast steels used extensively in the power generation and energy sectors. The manufacturing steps for these alloys have up to now required the use of costly secondary refining processes such as Argon Oxygen Decarburization (AOD). This paper will describe and validate a set of process parameters that obviate the need for secondary refining. Initial results indicate that the principle embrittling features, carbides and sigma phases, are not found, confirming the proposal that these phases form on bifilms entrained by surface turbulence during pouring. Other significant benefits from the absence of bifilms appear to include the elimination of cracking and leakage defects. Introduction Over the past five years there has been a great deal of truly first-rate research performed toward the goal of understanding the behavior of the superduplex alloy family [1]. The variety in this report targets the nominal composition 25Cr 7Ni 3.5Mo. We now have a detailed comprehension of the chemical and thermal processing requirements needed to produce quality components. However, the processing of the alloys used in such studies has been lamentably poor, ensuring the cast material studied has been polluted with air bubbles, entrained oxide films and reoxidation inclusions. In addition to previous observations that sigma phase can nucleate at ferrite-ferrite boundaries, ferrite-ferrite-austenite triple points [1] and at precipitated carbides [6], it seems likely that sigma phase and carbides precipitate on oxide bifilms, often at grain boundaries, and hence displaying cracks that give these phases the appearance of brittleness [4]. The resulting 'embrittled' matrix has served as the most significant challenge for casting facilities working with these alloys assumed to be the result of the presence of intermetallic phases and carbides at boundaries. The view presented in this work is that the sigma and carbide phases are strong and would therefore be expected to be resistant to cracking. The cracks are present simply because the phases form on the pre-existing oxide bifilms which are effectively cracks; the intermetallics and carbides are not believed to be 'brittle' of themselves. One might speculate that in the absence of doubled oxide films, the sigma phase would probably not precipitate (or not nearly to the degree traditionally observed), since it may find no other suitable substrate. In this case the sigma phase constituents would simply remain in supersaturated solution (it is just possible that they might precipitate later during a heat treatment or slow cooling cycle as an extremely fine phase, possibly now of a completely different composition and structure, and contributing to strength rather than embrittlement).
79
The molten metal is likely to be practically defect-free when it leaves the furnace because of the large density difference between oxides and steel, leading to rapid flotation of oxides and assimilation into the surface slag layer. It is the pouring of the metal into the ladle, and the following pour into the mold that can create so much damage. The prior use of a secondary refining process such as AOD (argon oxygen decarburization) treatment probably helps to eliminate much of the damage introduced by the pour into the ladle. Such treatment adds significant cost to the production of super duplex stainless steel. However, of course, it can do little to assist with the damage caused by a turbulent filling system in the mold. Initial results presented in this report indicate the naturally pressurized filling systems for super duplex stainless steel castings appear to obviate the need for secondary refining. The implication of this result is that the damage introduced during the pour from the furnace into the ladle floats out quickly, prior to the arrival at the mold, and the pouring into the mold, in agreement with expectations [4]. Processing Logic for Super Duplex Stainless Steel For the ASTM A890/ A995 family of alloys the utilization of a custom-batch induction melting process, constituted from washed, dry, virgin raw materials, alloy- specific ingot and controlled, internal, alloy- specific returns provides the foundation for a clean, workable base melt chemistry. Without introducing moisture, carbon-containing liquids and surface-oxidized charge material the melt will be clean and free from undesirable contaminants. Even with all of these precautions, however, the process metallurgist can still expect to find a heavy population of precipitated carbides and nucleated intermetallics such as sigma[l]. The mechanism for this phenomenon is the deterioration of a metastable super saturated ferrite phase that decomposes into a secondary austenite phase and alloy-rich intermetallics. The out-of-mold microstructure is certain to be frilly damaged and unusable without a proper solution anneal. However, even in taking all these precautions, many metal casters continue to experience the serious difficulties resulting from castings that crack during these initial and critical processing steps. As the result of an intensive twenty-four month long foundry trial it is the experience of the author that the difficulties encountered w;ith casting this family of alloys - difficulties that have been traditionally identified as solely solidification phenomenon - are in fact the result of oxide bifilms and air bubbles generated and introduced during the filling of the casting. Taylor reflects the traditional metallurgical view in his statement [2] "The gating design must also deliver the metal to the mold cavity with a minimal degree of turbulence. With the high nitrogen content of the metal, this smooth flow is particularly important to prevent subsurface gas indications. Pouring of duplex stainless steels has been compared to pouring a bottle of beer into a glass. The beer has a high quantity of dissolved gas. If the beer is poured slowly, very little foam develops. When the beer is poured quickly, a large amount of foam develops. Turbulent gating systems can produce gas indications in the casting. A gas problem may first appear as a slight mushrooming of the risers. The majority of the indications will be found on the cope in the highest parts of the castings, either after heat treatment, or when the skin of the casting is broken."
80
-Traditionally the use of the phrase "turbulence" when referring to the fill system has almost universally been associated with the bulk turbulence of the fluid system as quantified by the Reynolds number. The entrainment mechanism described here, however, is the results of surface turbulence resulting in a stainless steel. folding and mixing-in of the surface oxides, air, and other debris such as molding material. Understanding the difference between bulk and surface turbulence is essential for grasping the defect mechanisms described within this paper. The author describes elsewhere [3] explicitly why it is that the current industry-standard filling system design theory is incapable of insuring an adequately tranquil and aspiration-free delivery of the molten metal into the mold cavity. In fact, it has become clear that the consequences of the current fill system design practices are responsible for the embrittlement and porosity mechanisms occurring within these alloys. The use of a naturally pressurized filling system, however, completely eliminates the troublesome failures experienced when processing this family of alloys. Briefly, the naturally pressurized fill system is a system in which the areas of the filling channels are calculated by finding the velocity, V, at each fall distance, h, from the melt level in the pouring basin, assuming no friction. The approach is therefore a simple balance between potential energy, mgh, and kinetic energy, mV2/2. The approach is well known to casting method engineers. The significant difference in this application is to accept these areas and provide the filling system with only these calculated areas at every point throughout the downsprue and runners. Only the gates would be increased in size to reduce the velocity of entry to the mold to the critical 0.5 m/s if possible (on occasions this would be raised to 1.0 m/s if necessary, but not beyond this already 'stretched' limit to the Rule). A typical 'sprue exit/runner/gate' ratio for such a system might vary from 1:1:4 to 1:1:20. It must be stated however that the pre-selection of such a 'ratio' has no part to play in the design of a proper fill system; the ratios simply happen, occurring as a result of the design process. Confirmation by Results Empirical validation of the newly introduced processing concepts took place over a twenty-four month period as an extended foundry trial. The compilation of testing below was performed using test specimens sampled from a typical production-run pour. The test specimens were solution annealed with the castings immediately following shakeout and shot blasting. The alloy is an ASTM A890 Grade 6A super duplex stainless steel, air induction melted without secondary refining.
81
Table 1- Results of Chemical Analysis of Exemplar Heat ASTM A890 Grade 6A Required, \vt. percent Carbon 0.03 max 1.00 max Manganese Silicon 1.00 max Phosphorus 0.030 max Sulphur 0.025 max Chromium 24.0-26.0 Nickel 6.5-8.5 Molybedenum 3.0-4.0 Copper 0.5-1.0 Tungsten 0.5-1.0 Nitrogen 0.20-0.30
Results, wt. percent 0.014 0.79 0.79 0.019 0.007 25.49 6.86 3.62 0.758 0.581 0.22
Mechanical Testing Testing was carried out to meet the specification NORSOK M-630 MATERIAL DATA SHEET MDS D56 Rev. 3 TYPE OF MATERIAL: Ferritic/Austenitic Stainless Steel, Type 25Cr 'fable 2 - Results of Mechanical Testing on Exemplar Heat Results Required NORSOK M-630 MDS D56 Rev 3 Yield Strength, MPa 514 450 min 779 Ultimate Tensile Strength, MPa 700 min Elongation percent 18 min 36 301 max 233 Hardness Testing, BHN 45 min 110.6 Charpy Impact Testing -46 C, Joules Corrosion Resistance AST M A923 Ferric Chloride corrosion Test, Method C, 24 Hours Table 2 - Results of Corrosions Testing on Exemplar Heat Results Required Corrosion Rate, mdd* 10 max 0.94 *mdd = mass loss(mg) per total exposed surface area (dm2) per day Non-Destructive Assessment Pressure cover castings poured from the exemplar heat were subjected to 100% radiographie inspection. The castings were found to free of all defects, requiring no upgrading or rework. The elimination of defects traditionally identified as nitrogen gas porosity and microslinnkage is, frankly, nothing short of amazing. Furthermore the use of liquid penetrant inspection was dreaded for the defects that would be revealed as the nonn. The norm now has become a relaxing test expecting and confirming a routinely defect-free appearance.
82
Microstructural Assessment
Figure 2 ASTM A890 6A KOH electrolytic etch x3100 SEM, the microstructure appears perfectly clean; free from all carbides, intermetallics or other inclusions.
Figure 3 ASTM A890 6A KOH electrolytic etch x775 SEM, the microstructure appears perfectly clean; free from all carbides, intermetallics or other inclusions.
The specification NORSOK M-630 Material Date Sheet MDS D56 Rev. 3 requires the following: "The ferrite content shall be determined according to ASTM E 562 or equivalent and shall be within 35 - 55 %. The microstructure on a suitably etched specimen shall be free from intermetallic phases and precipitates." Naturally, the microstructure is normally checked using optical microscopy. However the author elected to examine the microstructure using scanning electron microscopy (SEM) to be certain that the examination was as stringent as possible. SEM not only has the benefit of greater resolution but has better phase differentiation than traditional reflected light microscopy for these alloys. As observed in Figures 2 and 3, the results are truly outstanding. There is an ideal ferrite-austenite phase balance, and a complete lack of nucleated intermetallics, precipitated carbides or other inclusion matter. It bears repeat for the sake of complete certainty, that this test specimen was air melted in an induction furnace with no covering shrouds and not subjected to any secondary refining. Furthermore, although there is a continued wide debate on whether the use of traditional degassing practices such as the addition of a calcium-silicon additive is necessary for these alloys, in fact, during the developmental stages of this research the author was able to demonstrate clearly that no degassing additives are required. In fact, it was clearly shown that the addition of degassing materials can result in the creation of dispersed inclusions. [5] Other benefits accruing from these new procedures is the experience of zero leakage defects or cracking during the production of approximately three hundred castings over the past 24 months. Once again this is perhaps to be expected in the absence of bifilm defects that would be expected to form excellent cracks and leak paths [4].
83
Validation The introduction of a new set of operating concepts accompanied by bold claims is necessarily subject to scrutiny. In the spirit of the scientific method, new concepts and theories must possess (as rightly suggested by Popper) the opportunity for falsifiability. The empirical evidence presented here demonstrates that the concepts introduced produce excellent results. These results have been repeatedly verified, effectively over 300 times. What the research does not answer, however, is what the out-of-mold microstructural differences are between castings poured using traditional methods and those poured using the new concepts. What is the quantified difference in the presence of sigma and carbides? This study has not yet been attempted so far, so that further work is clearly required. Thus although no direct 'before and after' comparison is currently available for microstructures of eastings by traditional filling and new techniques (as is common for industrial developments, in contrast to laboratory research), the current general appearance, properties and freedom from defects of the improved castings are strongly suggestive that the microstructures are uniquely free from undesirable embrittling phases. Conclusions The use of a precisely designed naturally pressurized fill system coupled with stringent melting practices allows for the production of premium quality super duplex stainless steel castings with the following unique list of properties. 1. No necessity for secondary refining processes such as AOD. 2. Mechanical properties, corrosion resistance and microstructural condition of castings far exceed industry guideline requirements. 3. The loss of castings due to cracking during subsequent thermal processing is completely eliminated. 4. Sub-feeder cracking following cut-off is completely eliminated. 5. Leakage of castings is eliminated. 6. Surface finish is uniquely good. 7. Upgrading of castings by weld repair is eliminated. 8. The study confirms the recent view that entrainment defects (air bubbles and bifilms) are the major defects in super duplex stainless steels, and that bifilms in particular appear to constitute the substrates for the precipitation of the major undesirable constituents such as sigma phase and carbides. Acknowledgements The author would like to thank John Campbell for his advice and assistance during the development stages of this project. The fill system concepts proposed by Campbell served as the foundation upon which these processing steps were built. The application of the naturally pressurized fill system to the family of high alloy steels has yielded phenomenal results not previously believed attainable. Additionally, the group of projects of which this paper was a part were conducted in a production metal casting facility with the all the pressures therein contained. Without the great support and patience of Steve Blenkhorn, such developments would never have happened.
84
References 1. Metallurgical Evaluation Of Cast Duplex Stainless Steels And Their Weldments, U.S. DEPARTMENT OF ENERGY Award Number - DE-FC07-00 ID13975, Songqing Wen, Carl D. Lundin, Greg Batten 2. Duplex Stainless Steel Production, Taylor, Steel Founders' Society of America Technical & Operating Conference Chicago, IL November 1994 3. Advanced Methoding Concepts for the Gravity Casting of Steel Alloys, Bob Puhakka, TMS 2011, San Diego CA. 4. Castings (2003), J. Campbell, Elsevier Butterworth-Heinemann 5. http:/fàobpuhakka.blogspot.com/2010/09/using-energy-dispersive-x-ray.html 6. Duplex Stainless Steels, A State-of-the-Art Literature Review, SFSA, March 2001.
85
Shape Casting: The 4 International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals <S Materials Society), 2011
IN SITU HIGH SPEED X-RAY OBSERVATION OF THE SOLIDIFICATION OF AL15CU WITH AND WITHOUT AL 2 0 3 COMPOSITE ADDITION R. W. Hamilton1", A. Leung1, A. B. Phillion2, P. Rockett1, T. Connolley3, and P. D. Lee1, 'Department of Materials, Imperial College London Prince Consort Road, London SW7 2BP, UK 2 School of Engineering, The University of British Columbia 3333 University Way, Kelowna, BC, Canada VIV 1V7 3 Diamond Light Source Ltd, Harwell Science & Innovation Campus, Didcot, OX11 ODE, UK Keywords: X-ray tomography, aluminium, porosity Abstract The use of a novel high temperature furnace in combination with a high speed camera capable of acquiring a complete 3D tomographic image in 3 s at DLC (Diamond Light Source) has enabled in-situ imaging of the formation of porosity during solidification, using monochromated X-rays. This study focuses on the comparative differences between Al-15wt.Cu with and without composite addition, and shows clearly the final stages of pore growth in 3D. Post-solidification high-resolution X-ray tomography was also performed to characterize the final structure, helping to explain the full mechanism of pore growth. Although at relatively coarse spacial resolution, the in-situ results demonstrate the effectiveness of the technique, and allow for the quantification of the rapid changes that occur during solidification. Introduction The development of high-speed X-ray tomographic capabilities on Beamline 112 at Diamond Light Source+ has enabled an in-depth study of the final stages of pore evolution and growth in aluminium alloys. This final stage is generally complex and is of key interest in many industrial applications. The ability to resolve pore development in short time steps (~1.5 sec) allows for direct observation of the complex interactions between the evolving primary Al, the appearance of the eutectic and the influence of third phases (e.g. particulates, intermetallics and contaminants). Although there is plenty of literature covering the characterisation of porosity by X-ray tomography[l-3], until recently it has been confined to post solidification observation. In the present work, novel results are presented comparing in-situ pore evolution in a simple Al 15%Cu binary alloy both with and without AI2O3 reinforcement. Experimental methods Material Two aluminium alloy cylindrical specimens, ~ 2 mm in diameter by 3 mm in height were examined in this study, Specimen 1 was a high-purity Al-15wt.%Cu alloy. Specimen 2 was the same Al-15wt.%Cu binary but containing 9.6vol.% AI2O3 particulate with a particle size range of 10-30 μιη. Corresponding author:
[email protected] Diamond Light Source Ltd, Harwelî Science & innovation Campus, Didcot, 0X11 ODE, UK
87
Apparatus The test apparatus, Figure 1, consisted of a small furnace with an internal bore of 13 mm surrounded by six alumina coated resistance elements and capable of maintaining a constant temperature of up to 900 °C. A 9 mm square hole in the furnace, covered with a thin layer of Al foil to prevent heat loss, allowed the X-rays to pass through the specimen from the source to the detector. The furnace was fixed in space above a rotation table, and held a specimen mount that was aligned perpendicular to the beam. This specimen mount could rotate freely within the furnace while acquiring the radiographs needed for computed X-ray tomography. Specimens were held in a boron nitride (BN) cup with a wall thickness of 0.5 mm. Two K-type thermocouples were mounted within the furnace, on opposite sides and as close as possible to the specimen-holder and at the same level as the specimen (and the observation window). There was a small air gap (~1 mm) between the holder and the thermocouple to ensure that there was no contact during rotation. Temperature was controlled via a National Instruments interface using a NI cRIO module. 3D X-ray tomographic scans were collected on Beamline 112 at DLC using monochromated Xrays (55keV), and a high speed Phantom camera with module 3 optics* collecting 120 frames per s at a voxel size of 12 μπι. Combined with the stage rotating continuously at 60° s"1, this enabled the capture of one complete tomograph every 3 s, as only 180° rotation is needed for tomography with monochromated X-rays. The on-board buffer had space for 17000 frames, enabling the capture of 47 complete scans.
Figure 1: Schematic of the test apparatus Test Methodology Each sample was heated in the furnace to 700°C and then allowed to stabilize at this temperature. During heating and stabilization, the samples were continuously observed using X-ray radiography to ensure that füll melting had taken place, and in the case of sample 1, all porosity had dissipated. In the case of sample 2, it was not possible to remove entirely the pre-existing bubbles due to the particulate addition, and so the experiment was run with the residual bubbles. After stabilization was completed, a cooling rate of 1 °C/s was applied, and the behaviour of the samples was observed in-situ. The camera continuously captured radiographs, storing the images in the on-board circular buffer, with the addition of a time stamp. Once the buffer was filled with data, the newest images automatically overwrote the earliest ones. At the point when solidification was observed to have completed, the capture function in the camera was triggered, www.diamond.ac.uk
88
then the buffer was downloaded to a hard disk after one final tomograph was acquired. The timestamp in the radiographs was used to synchronise them with the temperature data recorded from the thermocouples. One challenge in the apparatus design was the fact that the thermocouples were not mounted directly in the sample, but were instead recording the temperature within the furnace. These thermocouples were therefore separated by a small air gap, ~lmm, from the outer surface of the BN cup. Although this design would not be an issue in isothermal experiments, prior experience in cooling tests has shown that there may be an offset in temperature between the thermocouple and the sample during heating or cooling. Reconstruction and image processing A code written specifically for JEEP was used for the tomographic scan reconstruction^]. In total, 47 tomographic images were acquired for each specimen during the solidification process. Data processing was kept to a minimum by selecting scans at suitable time intervals, based on observed changes in the microstructure. Image processing was conducted using Avizo5. An initial 3x3 median filter was applied, and the porosity identified using a simple grey-level threshold. The porosity was quantified using internal sub-routines, and checked manually by reference to arbitrarily selected representative pores. In particular, apparent pores of less than 9 voxels in volume were excluded as noise. In the case of sample 2, regions between the large residual pores were selected to analyze the newly nucleated pores while excluding much of the residual porosity. Results and Discussion The reconstructed and rendered images of typical pores in the Al-15wt.%Cu are provided in Figure 2 and show clearly pore evolution with time at a cooling rate of 1 °C/s. Due to the limited memory buffer in the camera, it was not possible to capture the entire duration of the cooling process, so the pores are already visible in the first image at a size of several voxels. However, over the subsequent image sets at time steps corresponding to temperatures of approximately 598, 574, and 544 °C, Fig 2a,b,c, the pores can be seen to grow and develop complex shape. A multitude of smaller pores can be seen in the final image of the sequence (Fig 2c). The nucleation of these pores is due to the lack of feeding at high fraction solid, when the final phases solidify[3, 5] and hence fluid-flow cannot compensate the additional volume created by solidification shrinkage. Similar images were obtained from sample 2 containing reinforcing paniculate; however, due to the nature of the residual porosity previously mentioned, the overall images are not particularly informative. By focusing on areas which initially contained no porosity, as shown in Figure 3, it was possible to observe pores growing that did not exist at the onset of solidification.
Avizo v6.2, www.vsg3d.com
89
Figure 2.3D rendering of pores in Al-15wt.%Cu in real time; a) 598 °C, b) 574 °C, c) 544 °C In order to characterize the distribution and size of the pores, a volume of 6 mm3 was cropped from the reconstructed tomographic images and analyzed to quantify typical microstructural features, i.e. equivalent spherical radius, volume and surface area as shown in Figure 4, along with the determination of overall pore number density and porosity fraction. The graph of total porosity, P%, and pore number density, Nv, as a function of temperature in the Al-15wt.%Cu alloy shown in Figure 5 indicates that P% increased continuously during cooling, with a rapid increase in P% at high fractions solid. In contrast, the number of pores, Nv, remained low and relatively constant until a high fraction solid was reached at which point there was also a rapid increase in Nv. This agrees well with previous observations[6] and the concept of initial pores growing steadily due to relatively free diffusion of hydrogen in the liquid, until the final solidification when the remaining liquid solidified rapidly. At this point, the rapid reduction in liquid flow caused by the narrowing intergranular regions and increased liquid viscosity is accentuated by the presence of particulate, and results in reduced pressure within the liquid, void nucleation, and void growth. The decrease in Nv seen at the end of solidification cannot currently be accounted for, but it could be an effect of the relatively coarse resolution, which could have caused nearby pores to be counted as separate when they are actually connected. In the final stages of solidification, shrinkage allows them to grow enough to be identified as one interconnected pore.
Figure 3. 3D rendering of pores in Al-15wt.%Cu with 9.6vol.% AI2O3 particulate; a)618°C, b)565°C
90
Figure 4. Typical microstructural features of as-cast micro-porosity Also shown in Figure 5 is the development of P% and JVv in the Al-15wt.%Cu alloy with 9.6 νο1.%Αΐ2θ3. As can be seen in the figure, the initial percentage porosity was larger compared to the alloy without particulate, due to the residual porosity, but the increase in Nv during cooling occurred at a similar rate. It also appears that there was a tendency for more pores to grow above the observable threshold, rather than relatively few pores to grow at the expense of others. In comparison the Al-15wt.%Cu showed a similar magnitude of porosity, with markedly fewer (1/3) actual pores. This hypothesis is also partially validated by the fact that the sharp increase in pore nucleation occurred at ~590'C in this alloy, at least 20' C above the start of pore nucleation in the alloy without particulate. Thus, the particulate phase can significantly restrict the growth of the pores, as well as acting as sites for further heterogeneous pore nucleation, leading to a larger number of smaller pores. It should be noted that the presence of residual porosity would also deplete the amount of hydrogen available for newly nucleated pores to grow.
Figure 5. The P% and Nv for the Al-15wt.%Cu alloy and Al-15wt.%Cu with 9.6vol.% A1203 particulate as a function of temperature.
91
By isolating an individual, but representative pore in sample 1 and sample 2, it was possible to examine the development of specific pores starting from the point at which the resolution threshold was exceeded (in this case at least 9 voxels in volume) until final solidification. In the Al-15wt.%Cu alloy both with and without paniculate, the pore growth, shown in Fig 6 as pore equivalent radius, was predominantly as expected from the literature[7]: 1. Initial nucleation and initial growth (which cannot be observed here) 2. A rapid growth phase with a rate which then decreases with decreasing temperature. A plot of the ratio of the pore surface area as compared to the expected surface area of a sphere containing the same volume (i.e. a measure of sphericity), also provided in Fig 6, indicates that initially pores in both samples grew reasonably spherically. It might be expected that some impingement occurs during the final stages of solidification, but this detail was below the resolution of this in-situ tomographic experiment. However, it can be seen that, in the presence of paniculate, the pore stopped growing sooner and smaller, suggesting that growth became restricted by the paniculate. During the final stages of solidification the pore in sample 2 grew considerably less, and becomes more tortuous as it follows the intergranular boundaries[8].
Figure 6. Equivalent radius and surface area ratio for a single pore in Al-15wt.%Cu with and without paniculate. Conclusions High-speed tomographic images were successfully captured of the solidification of an Al15wt.%Cu alloy both with and without reinforcing A1203 paniculate, allowing for visualization and quantification of porosity growth as a function of temperature in the semi-solid regime.
92
Despite limitations in the optics available giving a relatively large spatial resolution of 12 μιτι, and the fact that it was not possible to monitor the exact temperature of the sample due to the high speed rotation, the results are surprisingly clear and show good agreement to the theories of pore nucleation, growth, and coalescence. There is good agreement with the expectation, that the growth of porosity will be constrained where there is a significant addition of paniculate, resulting in a larger number of smaller pores. In particular, the final evolution of shrinkage porosity that occurs at final solidification is observed, and shown to be more tortuous (less spherical) in the paniculate containing sample. These are significant initial results, which will be strengthened by future planned work at higher time and spacial resolution. Acknowledgements The authors would like to thank all the staff at Diamond Light Source, Beamline 112; (EE2071) and the EPSRC (EP/FOO1452/1). References [1] J.Y. Buffiere, S. Savelli, E. Maire, "Characterisation of MMCp and cast Aluminium alloys," X-ray Tomography in Material Science, ed. Baruchel, Buffière, Maire, Merle, Peix, Ed. Hermes, Paris, Fr. (2000), 103. [2] P.D. Lee and J.D. Hunt, "Hydrogen porosity in directional solidified aluminium-copper alloys: In situ observation," Acta Materialia, 45(10) (1997), 4155. [3] O. Lashkari, L. Yao, S.L. Cockcroft, D.M. Maijer, "X-ray microtomographic characterization of porosity in aluminum alloy A356," Metallurgical and Materials Transactions A, 40A(4) (2009), 991. [4] V. Titarenko, S. Titarenko, P.J. Withers, F. De Carlo, and X. Xiao, "Improved tomographic reconstructions using adaptive time-dependent intensity normalization," Journal of Synchrotron Radiation, 17 (2010), 689. [5] J.D. Zhu, S.L. Cockcroft, D.M. Maijer, "Modeling of microporosity formation in A356 aluminum alloy casting," Metallurgical and Materials Transactions A, 37A(3) (2006), 1075. [6] M.A. Easton and D.H. St.John, "The effect of grain refinement on the formation of casting defects in alloy 356 castings," International journal of Cast Metals Research, 12(6) (2000), 393. [7] R.C. Atwood, S. Sridhar, W. Zhang, P.D. Lee, "Diffusion-controlled growth of hydrogen pores in aluminium-silicon castings: In situ observation and modelling," Acta Materialia, 48 (2) (2000), 405. [8] R. Sasikumar, M. Rettenmayr, S. Savithri, H.E. Exner, "A model for the formation of interdendritic cavities from pores pre-existing in the melt," Zeitschrift für Metallkunde, 92 (2) (2001), 158.
93
Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011 IN-MOLD THERMAL ANALYSIS OF DUCTILE CAST IRON Morten I. Onsoien SINTEF Materials and Chemistry, N-7465 Trondheim, Norway Keywords: Ductile cast iron, In-mold melt treatment, In-mold thermal analysis, Microstructure
Abstract The objective of the present work is to study characteristic solidification data of three experimental ductile irons, produced using a flow through in-mold melt treatment technique, by means of in-mold thermal analysis. All produced irons were in addition subjected to chemical analysis, quantitative metallography, as well as an evaluation of shrinkage porosity and carbide forming propensity. Applied melt treatment alloys were based on FeSiMg with small contents of cerium, lanthanum or misch-metal. Based on the obtained results it was concluded that in-mold thermal analysis provides an effective way of monitoring the graphite nucleation throughout the solidification. Furthermore, the selected melt treatment resulted in ductile irons having very high nodule counts, up to 1162 nodules mm"2, very low shrinkage porosity and very low carbide forming propensity. Introduction In-mold melt treatment of ductile iron with ferrosilicon based treatment alloys was first developed and introduced to the foundry practice in mid-sixties [1]. In-mold processes for making ductile iron has gained increased popularity over the years due to advantages such as consistently high magnesium recovery, an almost complete absence of glare and fume, simultaneous nodularization and inoculation, and avoidance of problems related with magnesium fade [2-7]. Conditions for a successful in-mold procedure require a balanced amount of ferrosilicon-magnesium (FeSiMg) based alloy in the reaction chamber under the runner to last throughout the pour. A filter in the runner system is often used in combination with in-mold melt treatment to avoid problems with dross related defects in the casting [2], Thermal analysis is commonly used to study phase transformations, and this technique is frequently used in establishing equilibrium phase diagrams [8]. Under given kinetic conditions, measured deviations from equilibrium reaction temperatures can be used to study nucleation and growth processes [9]. Solidification characteristics can be investigated using a simple cooling curve analysis applied to standardized castings, and thus, together with chemical analyses, provide rapid characteristic information that can be used to tailor the cast iron melt [10-12]. In the current work use of thermal analysis in material and process control of ductile iron produced using a flow through in-mold treatment has been explored. The major goal was to study the characteristic solidification data of the ductile iron within the mold immediately after melt treatment. In addition the produced irons were characterized by means of chemical analyses, quantitative metallography, relative shrinkage porosity measurements and carbide forming propensity.
95
Experimental procedure Figure 1 shows schematically the applied casting procedure using a flow through in-mold system. An experimental ductile cast iron charge of 80 kg was made based on re-melting of iron returns, pig iron, high purity ferrosilicon and graphite in an induction furnace. The molten iron reached a temperature of 1510 °C prior to pouring the melt into the holding ladle. The melt was kept in the holding ladle until it reached a temperature of 1400 °C. Three batches of 20 kg each were then poured into three different sand molds containing a reaction chamber and a runner system for nodularization using a flow through in-mold melt treatment technique. Prior to pouring, the first reaction chamber mold was charged with a commercial lanthanum containing FeSiMg alloy (La-FeSiMg), the second with an experimental cerium containing FeSiMg alloy (Ce-FeSiMg) and the third with an experimental FeSiMg alloy containing misch-metal (MMFeSiMg). The amount of treatment alloy corresponded to 1 wt.% of the melt. All FeSiMg alloys contained 5.6-5.7 wt.% magnesium and 0.3 to 0.4 wt.% rare earth elements. The melt was allowed to flow through the reaction chamber mold and into four different sample molds (Casting A-D). The sample molds were filled, by passing them continuously on a conveyor, under the liquid iron flowing from the reaction chamber mold (sequence A-B-C-D) at a filling rate of around 1.8 kg s"1. The sample molds consisted of a 20 mm thick plate, a 5 mm thick plate, a chill wedge sample, and a hemispherical sample for shrinkage porosity evaluation. In addition a single thermocouple Quik-Cup for thermal analysis was embedded in the drag part of the mold. This experimental setup allowed sample extraction for "in-situ" thermal analyses, chemical analyses and metallographic analyses at different times after the magnesium reaction start without any risk of extended mixing of the melts.
Figure 1. Flow diagram of casting procedure using the flow through in-mold system. Analysis of the collected temperature data was done using a commercially available software for thermal analysis. The following parameters were extracted; undercooling, recalescence, graphite factor 1 (GRF1) and graphite factor 2 (GRF2). GRF1 is an indicator of eutectic graphite precipitation rate, while GRF2 is an expression of the amount of eutectic graphite formed during the last stage of solidification [13]. Samples for chemical analysis were extracted from the 20 mm thick plates. Metallographic examination of the cast irons were performed on samples extracted from a cross section cut at the centre of both the 5 mm and the 20 mm thick plates. The metallographic samples were prepared according to standard metallographic techniques, i. e., polished to a Ιμτη diamond spray finish. The graphite nodules were then characterized using an automated image analysis system. For analyses of characteristic graphite data such as nodule count, nodule diameter, nodule shape factor and nodularity, only graphite nodules larger than 5 μιη were measured. Nodule diameter was defined as the diameter of a circle with the same area as the graphite nodule under consideration, whereas the shape factor S is an area to perimeter function expressed as: S^iTtA/P2, where A is the nodule area and P is the perimeter. Nodularity was defined as the percentage of nodules with a shape factor larger than 0.65. The polished
96
samples were then etched in 2 % nital for quantification of the microstructure constituents such as ferrite, pearlite and carbide by means of counting minimum 1000 grid points at a magnification of 200 X using an optical microscope. The cast hemispherical body was used to determine the shrinkage tendency of the different irons by means of density measurements using Archimedes principle of density measurement, while carbide forming propensity was evaluated by means of a standard chill wedge sample. Results and discussion Chemical composition The chemical composition of the produced ductile irons is summarized in Table 1. Within each series the magnesium level increases from the first mold to be filled to the last one, i.e. from mold A to mold D. Magnesium treatment using the Ce-FeSiMg alloy seems to give higher magnesium yield than magnesium treatment using the two other FeSiMg alloys, especially in the firstfilledmold. The rare earth levels in the samples are, as expected, quite low as a result of the low amount of treatment alloy added. Still the levels are not to far away from the optimum levels of rare earth metals reported by Lalich [14], such that a positive effect on nodule count is expected. Table 1. Chemical composition of experimental ductile irons. Sample La-FeSiMg A La-FeSiMg B La-FeSiMg C La-FeSiMg D Ce-FeSiMg A Ce-FeSiMg B Ce-FeSiMg C Ce-FeSiMg D MM-FeSiMg A MM-FeSiMg B MM-FeSiMg C MM-FeSiMg D
C 3.81 3.70 3.89 3.81 3.82 3.73 3.94 3.75 3.87 3.83 3.83 3.80
Si 2.46 2.40 2.51 2.55 2.55 2.64 2.60 2.59 2.40 2.55 2.37 2.55
0.23 0.22
Elements P 0.023 0.022 0.022 0.022 0.023 0.021 0.022 0.023 0.022
0.23 0.23 0.24
0.023 0.022 0.023
Mn 0.23 0.25 0.23 0.24 0.23 0.24 0.24
(wt.%) S 0.011 0.011 0.011 0.012 0.011 0.010 0.010 0.010 0.011 0.010 0.010 0.011
Mg 0.021 0.028 0.029 0.030 0.027 0.028 0.027 0.031 0.018 0.027 0.030 0.030
Ce f
"
*
(13)
Here,/; = \lq\,fc is the critical void volume fraction, and,//.- is the failure void volume fraction. Their values should be different for different materials. In current hot tearing indicator calculation, constant
105
values are used in [18]. Gurson's coefficient characterizes the rapid loss of material strength due to the growth of void volume fraction/,. When/. =//.-, then / * = / , =l/3.5 very high hot cracking susceptibility Microscopy. SEM examination was performed at 20 kV in BSD-mode to compare the as-cast microstructure with results from TCC and to investigate fracture surfaces. Results As-Cast Microstructure. Existing phases in the as-cast microstructure of various alloys were calculated by TCC (equilibrium conditions) and are shown in Fig. 3. Microstructure examination with SEM confirmed the theoretical predicted results. Alloy AlSi7Mg0.6Cu0.5 is given as an example in Fig. 4 to compare forecast phases by TCC and detected phases by SEM. Qualitatively, it is apparent from 50 EDX point analysis that in the sand mold a higher fraction of Mg2Si can be found.
Figure 3. As-cast phases at room temperature, calculated by TCC in equilibrium.
Figure 4. SEM, BSD, AlSi7Mg0.6Cu0.5, as-cast phases, (a) die mold, (b) sand mold. 116
Crack Surfaces. Crack surfaces initiated during casting of the HCI-samples in the dog bone shaped die were investigated by SEM. Samples with a small hot cracking level, i.e. samples not completely separated by a crack, were mechanically opened to subsequently observe the crack surface. Fig. 5 shows three SEM pictures of various hot cracking levels. SEM results indicate that at areas next to hot cracks no or insufficient eutectic phase exists. Furthermore, detailed SEM investigation of the fracture surfaces revealed no presence of bifilms as these may act as crack initiation sides within interdendritic liquid.
Figure 5. SEM, fracture surfaces, (a) dendrites in fully broken sample, WF=1, (b) dendrites and eutectic phase in sample with modest crack, WF=0.5 - mechanically opened, (c) eutectic in sample with hair crack, WF=0.25 - mechanically opened. TFR. Table 1 shows the TFR of all alloys. It is evident that the Cu-content has the dominating influence on TFR over that of Mg-content. Firstly, a high Cu-content results in a large TFR. Secondly, a low Mg-content results also in large TFR. Hence, the largest TFR is obtained in the alloy AlSi7Mg0.1Cu0.5 (see Fig. 6), the smallest TFR is obtained in the alloy AlSi7Mg0.6Cu0.05 (see Fig. 7). Table 1. TFR of evaluated alloys, calculated with TCC Alloy TFR [°C] AlSi7Mg0.1Cu0.5 46.0 27.0 AlSi7Mg0.6Cu0.5 AlSi7Mg0.1Cu0.05 17.0 AlSi7MgO.3CuO.05 9.5 AlSi7Mg0.6Cu0.05 4.0 CSC. Table 2 shows the CSC of three evaluated alloys. Again Cu has the dominant influence on the CSC. A high Cu-content results in a high CSC, a low Mg-content results also in a high CSC. Furthermore, the CSC results show that the CSC is much lower in sand casting than in die casting. The reason for this is a longer solidification time in sand casting and the larger amount of eutectic present which may induce a healing process for cracks. Table 2. CSC of evaluated alloys. CSC [-] Alloy Die Mold Sand Mold 0.69 7.3 AlSi7Mg0.1CuO,5 0.36 4.5 AlSi7Mg0.6CuO,5 0.33 3.7 AlSi7Mg0.1Cu0.05
117
Figure 7. TCC, calculation of TFR (4°C), AlSi7Mg0.6Cu0.05. HCI. Table 3 shows the HCI and subsequent resulting hot cracking susceptibility. For every alloy five hot cracking samples were investigated (NOF=5). Again Cu has a dominant effect on HCI. A high Cu-content results in a high HCI, a low Mg-content results also in a high HCI. Furthermore, all hot cracking susceptibilities for alloys in sand casting are negligible. Table 3. HCI and hot cracking susceptibility of evaluated alloys. HCI [-] HCI [-] Hot Cracking Hot Cracking Alloy Susceptibility Susceptibility Die Mold Sand Mold 0.8 0.01 AlSi7Mg0.1Cu0.5 no susceptibility small susceptibility AlSi7Mg0.6Cu0.5 0.6 no susceptibility 0.01 small susceptibility AlSi7Mg0.1Cu0.05 0.3 0.01 no susceptibility no susceptibility AlSi7Mg0.3Cu0.05 0.22 no susceptibility no susceptibility AlSi7Mg0.6Cu0.05 0.01 no susceptibility no susceptibility Summary of Results. Figure 8 shows in a summary of results the theoretical models and the experimental hot cracking index method for different AlSi7MgCu-alloys. On the left y-axis TFR values are plotted. On the right y-axis CSC and HCI values are plotted, the HCI values are multiplied by 10 so that it was possible to show both measurement values on one axis.
118
Figure 8. Trend lines of TFR, CSC and HCI for different AlSi7MgCu-alloys for theoretical and experimental methods for measuring hot cracking susceptibility. Discussion A brief overview of influencing factors on hot cracking was given. Five different AlSi7MgCu-alloys with varying Mg and Cu content were evaluated with three methods: theoretical TFR (Gulliver-Scheil condition), semi-empirical CSC model (Gulliver-Scheil condition) and experimental HCI examination. In contrast to the review for DC casting by Eskin et. al [4] all three performed examinations indicate the same trend (see also Fig.8): The Cu-content has a dominating influence on hot cracking susceptibility in AlSi7MgCu-alloys. A high Cu-content results in a large hot cracking susceptibility (large TFR, high HCI and high CSC), a high Mg-content results in small hot cracking susceptibility (small TFR, low HCI and low CSC). Furthermore, theoretical predicted phases were also found in SEM investigations. At higher Cuconcentrations Cu-phases segregate in form of Al2CuMg, Al5Cu2SÌ6Mg8 and Al2Cu during solidification; this has a negative effect and depletes the alloy of eutectic available for micro feeding. Despite the fact that the grain size in sand casting is larger, in general a lower hot cracking susceptibility is observed in sand casting. The amount of precipitated Mg-containing phases in the eutectic in as-cast alloys is higher in sand casting than in die casting. Moreover, the soft sand mold can accommodate shrinkage strains. For AlSi7MgCu-alloys of similar grain size a good correlation between theoretical models and the experimental hot cracking index method was observed as a material property. Especially for the development of new casting alloys a theoretical tool to forecast the hot cracking susceptibility is of major interest. Experimental evaluation of hot cracking tendency is intricate. TCC calculations are an adequate method of predicting the hot cracking susceptibility qualitatively. Acknowledgement Part of this work was financially supported by the Austrian Research Promotion Agency FFG.
119
References [I] [2] [3] [4] [5] [6] [7] [8] [9] [10] [II] [12] [13] [14] [15] [16]
[17] [18] [19] [20] [21] [22]
F. Matsuda, K. Nakata, K. Tsukamoto, S. Johgan, "Combined Effect of Current Pulsation and Zr Addition on Improvement of Solidification Cracking of Al-Zn-Mg Alloy Weld Metal," Transactions ofJWRl, 14, No. 2 (1985), 99-104. F. Matsuda, K. Nakata, and Y. Shimokusu, "Effect of Additional Element on Weld Solidification Crack Susceptibility of Al-Zn-Mg", Transactions ofJWRl, 12, No. 1 (1983), 81-87. G.L. Petrov, A.G. Makarov, "The sensitivity of Al-Zn-Mg Alloy to Hot Cracking During Welding," Avtomaticheskaya Svarka, No. 9 (1961), 18. D.G. Eskin, L. Katgerman, "A Quest for a New Hot Tearing Criterion," Metallurgical and Materials Transactions Λ, 38 (2007), 1511-1514. E. Cicala, G. Duffet, H. Andrzejewski, D. Grevey and S. Ignat, "Hot cracking in Al-Mg-Si alloy laser welding - operating parameters and their effects," Materials Science and Engineering A, 395 (2005), 1-9. E. Schubert, M. Klassen, J. Skupin, G. Sepold, "Effect of filler wire on process stability in laser beam welding of aluminium-alloys," Proceedings of the 6th International Conference on C1SFFEL, Toulon, France (1998), 195-203. T.W. Clyne, G.J. Davies,"The influence of composition on solidification cracking susceptibility in binary alloy systems," The British Foundryman, 74 (1981), 65-73. E. Brunhuber, Giesserei-Lexikon (Berlin: Schiele & Schön, 14. Auflage, 1988), 1100-1102. A.A. Gokhale, "Solidification Cracking: A Review," Transaction of the Indian Institute of Metals, 39 (1986), 153-164. M.B. Djurdjevic, R. Schmid Fetzer, "Thermodynamic calculation as a tool for thixoforming alloy and process development", Material Science and Engineering A, 417 (2006), 24-33. S. Lin, "A study of hot tearing in wrought aluminum alloys" (Ph.D. thesis, University of Quebec, 1999), 7-68, 69-90. H.F. Bishop, CG. Ackerlind, W.S. Pellini, "Investigation of metallurgical and mechanical effects in the development of hot tearing", Trans. AFS, 65, 1957, 247-258. D.C.G. Lees, "The Hot Tearing Tendencies of Aluminium Casting Alloys," The Journal of the Institute of Metals, 72 (1946), 343. J.A. Spittle, A.A. Cushway, „Influences of superheat on grain structure on hot-tearing susceptibilities of Al-Cu alloy castings," Metals Technology, 10 (1983), S. 6-13. J.A. Dantzig, M. Rappaz, Solidification (Lausanne: EPFL Press, CRC Press, 2009), 519565. T.W. Clyne, G.J. Davies, "Comparison between experimental data and theoretical predictions relating to dependence of solidification cracking on composition," Proceedings of the Conference on Solidification and Casting of Metals, Metals Society, London (1979), 274-278. L. Katgerman, "A Mathematical Model for Hot Cracking of Aluminum Alloys During D.C.Casting," Journal of Metals (1982), 46-49. U. Feurer, "Mathematisches Modell der Warmrissneigung von binären Aluminium Legierungen," Giesserei Forschung, 28 (1976), 75-80. M. Rappaz, J.M. Drezet, M. Gremaud, „A New Hot-Tearing Criterion," Metallurgical and Materials Transactions A, 30A (1999), 449-455. B. Lenczowski, H. Koch, K. Eigenfeld, "Neue Entwicklungen auf dem Gebiet der warmfesten Aluminium-Gusswerkstoffe," Gießerei, 8 (2004), 32-38. A. Franke, „Design of new high-performance aluminum casting alloys" (Ph.D. thesis, University of Leoben, 2006), 50-61. C. Kneissl, T. Pabel, G. Dambauer, P. Schumacher, "Formenkonzept und Ergebnisse gießtechnologischer Versuche zur Legierungsentwicklung im Niederdruckkokillenguss," Giesserei-Rundschau, 56 (2009), 120-125.
120
Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
HYDROGEN AND COOLING RATE EFFETCS ON MICROPOROSITY FORMATION IN THE PRODUCTION OF DEFECT-CONTROLLED FATIGUE SPECIMENS Rosario Squatrito1, Ivan Todaro1, Lorella Ceschini2, Andrea Morri2, Luca Tomesani1 ΌΐΕΜ (Department of Mechanical Engineering Constructions) Viale Risorgimento 2,40126 Bologna, Italy 2 SMETEC (Department of Metallurgy) Viale Risorgimento 4,40126 Bologna, Italy Keywords: Gravity casting, Aluminum Alloy, Fatigue Specimen, Microporosity Abstract In experiments aimed at the production of fatigue specimens, the increased number of nearly identical specimens needed for each processing condition, together with the high sensitivity to pore size, call for very strict requirements of both the casting tool and the processing conditions. An experiment for producing aluminium alloy fatigue specimens by gravity casting with controlled microstructure and defects is presented here. The main requirements to be obtained on a set of specimens (extracted from a single casting block) were to have the near identical microstructure and gas porosity content. The main process parameters were the hydrogen level of the melt, the addition of oxides for improving the number of pore nucleation sites and the cooling rate within the casting mould. The distribution of the relevant properties (SDAS, %area of porosity) was measured throughout the casting plates in order to validate the design criteria of both the experiment and the mould. Introduction Al-Si casting alloys, such as A356/A357, find extensive applications in the transport field, due to their excellent castability, corrosion resistance, and especially their high strength-to-weight ratio which increases performance and fuel economy. However, the casting process inevitably introduces solidification defects, which can significantly reduce the mechanical properties, mainly the elongation to failure and, above all, the fatigue strength of the final cast component. Several studies have shown that fatigue resistance of cast Al-Si alloys, is dominated by casting defects, such as gas and shrinkages pores, which considerably decrease the fatigue life with respect to defect-free cast components [1]. Only when the porosity is negligible (as in the case of castings subjected to hot isostatic pressing, HIP), the negative effect on the fatigue life of others solidification defects (such as oxide films) becomes dominant [2-3]. In order to evaluate the effect of solidification defects and other microstructural features on fatigue strength, the production of fatigue specimens with controlled microstructure and defects is crucial. In fact, when setting a tool for casting experiments, many sources of defects generate at both the filling and solidification stages have to be considered. To this aim, it is not only important to follow all the rules for good mould design, but it is also essential to carefully evaluate the specimen casting process by numerical analysis as if it was a process itself.
121
To meet these needs, the experimental activity presented in this paper was aimed to produce several sets of nearly identical casting specimens with different microstructure and amount of porosity. This goal was reached by means of a permanent gravity casting device, properly studied to ensure a sufficient control of the casting process, and by implementing a precise melt treatment procedure. 2. EXPERIMENTAL The aims of the experiments were: 1. Producing simple castings of A356 aluminium alloy, that would allow the extraction of two sets of nearly identical specimens in terms of microstructure and defects distribution. 2. Producing different castings, each one with different microstructure and porosity. 3. Producing a gradient of microstructure and porosity throughout each casting cross section. To achieve these aims it was necessary to have a high level of control of process conditions in terms of local cooling rate and preliminary melt treatment. 2.1 Design of the casting mould The general features for the casting mould, were set as follows: 1. Possibility to vary the cooling conditions in different casting experiments 2. Forcing of heat flux direction during solidification inducing planar isothermal surfaces, thus obtaining a microstructure gradient inside the casting 3. Filling conditions to avoid metal flow surface turbulence and trapping of surface oxides. These requirements resulted in the tool geometry outlined in Fig. 1. The global mould concept follows the classical arrangement of a gravity die tool, based on a vertical tapered sprue, followed by a 90° curve, a horizontal runner, a single gate connected at the bottom side of the casting and having the same section length of cast plate, which is filled from the bottom up. The casting geometry is a simple vertical plate, 30 mm in thickness, 300 mm in height and 250 mm in width. The mould was made of two elements (A and B in Fig.l) machined from two blocks of CK45 steel. In order to control the cooling conditions in different castings and the initial temperature distribution of the mould, the element (A) was crossed by a system of cooling channels positioned at 20 mm from the casting/mould interface, to be fed either by water or air. The element A could be heated by oxyfuel combustion as well. In order to obtain a gradient in the cooling rate throughout the casting thickness, a 30 mm layer of insulating material (C) was placed in a pocket machined in the element (B), preventing the casting from cooling in that direction. In order to check the initial temperature distribution of the mould, and to record the evolution of temperature with time, eight holes 1.5 mm in diameter were drilled through the back of element (A) to allow thermocouples embedded at 5 mm and 10 mm from the casting/mould interface. The filling system dimensions (runner sections reduction, gate length and depth) were studied by means of numerical simulation with PROCAST v. 2008, in order to avoid early metal flows into the plate during the filling of the runner (metal stops) and to fill the mould evenly.
122
Fig. 1: Outline of the casting tool A critical limit around 0.5 m/s of melt velocity inside the gate was assumed as an optimization criterion to define the dimensions of the inlet system, in order to avoid any disturbance coming from the unevenness of structure that could be related to the filling phase: in particular gas entrapment, turbulence, excessive speed, filling stops [4- 5]. To meet these requirements two devices were used: 1. a cylindrical filling reservoir placed above the filter (positioned horizontally) to absorb the peaks of metallostatic pressure due to the fluid-dynamics behaviour of metal during the filling of the basin and the downsprue; 2. a trap at the end of the rise channel, to avoid the flow into the plate cavity of oxides that form on the metal free surface in the first steps of riser filling. The filling analysis of the adopted single gate solution showed the formation of a large fluid recirculation (Fig. 2) in the vertical plane of the casting that could force convective flows during cooling and solidification. The optimal inlet-gate casting dimensions were then evaluated by minimizing the momentum of the recycled fluid structure, taking into account the massive dimensions of the cast and the presence of an imposed thermal gradient along the plate section. Numerical results showed that the ratio between first cooling time (above T|jq) and insurgence time of convective flow meant that the temperature differences inside the casting at the end of filling phase were negligible. In all of the experimental conditions, the maximum temperature difference between the top and bottom of the casting was below what was considered to be consistent with the requirement of 'near identical' specimens production (Fig. 3). The numerical thermal analysis performed with this mould showed that a planar solidification front proceeds from the cold side to the hot side of the mould. The predicted extensions of isothermal surfaces and thermal gradient along the cast depth, suggested the capability to obtain two casting portions where specimens can be extracted, thus producing two sets of specimens (one of faster solidification, the other of lower solidification rate) (Fig. 4), each one characterized by nearly identical cooling rate conditions.
123
Fig. 2: Flow Streamlines during rilling
Fig. 3: Temperature distribution at the end of filling
2.2 Melt Treatment In order to achieve different amounts of porosity in the castings the initial hydrogen content in the melt was varied. To get different and controlled hydrogen content in the molten material, it was necessary to build a simple gassing feeding system to the rotary impeller and to formalize a melt treatment procedure. In order to always have the same initial conditions, each charge of molten material in the furnace underwent the same preliminary treatment, consisting of Sr addition for eutectic Si modification ,Ti-B addition for grain refinement and dross removal. Several Ar degassing cycles by means of a rotary impeller (time of up to 45 minutes) and then the measurement of hydrogen content through Foseco Alspek H probe completed the start-up procedure of the experiment. The initial targeted hydrogen content was 0.08 - 0.1 mlHi/lOOg. After the preliminary treatment, it was possible to reach the desired hydrogen levels (Tab.l) simply by up-gassing the melt through the rotary impeller fed by a 10% II2- 90% Ar mixture . Table 1 Hydrogen Level Hydrogen Level (HL) Val.(mlH2/100g)
HL0 - very low
HL1 -low
HL2 -medium
HL3-high
0.08
0.13-0.16
0.23-0.24
0.30-0.33
Another aim of the experimentation was to study the effect of oxide content on porosity nucleation. For each hydrogen level, the oxide content was varied by introducing dry air through the gassing system. The oxidation level was determined by the amount of gassing treatment time (Tab. 2): Table 2 Oxide Level Oxide Level (OL) Val (minutes)
OL0 0
OLI 1
OL2 3
OL3 5
2.3 Microstructural characterization To evaluate the correspondence between the targets of the experiment and the actual achieved the central zones of plates were investigated, corresponding to the part of the casting designed for the extraction of fatigue-test specimens (Fig. 4).
124
Fig. 4: Cast plate and fatigue specimen Fig. 5: Metallographic specimen extraction For this purpose , 5 mm thick slices were cut from each cast plate at 130 mm from the gate (Fig. 5) then 30x10 mm metallographic specimens were extracted in the center of every slice to perform the microstructural analysis along their whole thickness. The specimens were prepared using standard metallographic techniques, according to ASTM E3 [6]. Qualitative and quantitative metallographic analyses were carried out, using an optical microscope (OM) and an image analysis software (Image pro-Plus®), on approximately 30 optical micrographs for each specimen. The microstructural characterization was focused on the evaluation of: • secondary dendrite arm spacing (SDAS [μηι]), • percentage area fraction of defects (DAF%) • number of defects per cm"2 (Nd [cm"2] ) • maximum Feret diameter (largest side of the rectangle enclosing the defect [μηι]) • roundness R of every defect, defined as: (Perimeter d e f e r t ) 2 4-rtAreadefect Where R=l means a circular shape and higher values means more irregular shape =
All of the microstructural data were evaluated as a function of the distance from the uninsulated side of the mould, the hydrogen content, the oxide level and the different cooling system of the mould (air or water cooling). 3. RESULTS AND DISCUSSION SDAS trends on the casting thickness The analysis of all of the average experimental SDAS data highlighted an increase with the distance from the mould side (Fig. 6), testifying to the actual gradient of cooling rate achieved. In Fig 6. average SDAS values are separated into two classes depending on the way the mould was cooled. Water cooling induced a difference in the SDAS values of up to 10 μτη in the central part of the plate while no variation was obtained close to the mould and insulator sides due to their starting temperature. Effect of hydrogen content Hydrogen content in the melt proved to have a high influence on porosity. Low hydrogen level (HL0) castings showed no porosity but in the side near to the insulator (Fig.7). Moreover the
125
number of pores was low (Fig.8), and they had a high R value (Fig.9) so that they appear to be "mainly shrinkage driven".
Distance from Che not insulated side of the castings, mm
Fig.6: SD AS as a function of the distance from the not insulated side of the mould, for air /water cooled castings.
Fig. 7: Average values of the percentage area fraction of defects, as a function of the distance from the mould, at the different hydrogen levels.
Fig.8: Average values of the number of defects as a function of the distance from the mould, at the different hydrogen levels
Fig. 9: Average R values of defects, as a function Fig. 10: Average values of the max ferèt of the distance from the mould, at the different diameter of defects, as a function of the hydrogen levels. distance from the mould, at the different hydrogen levels.
126
As the hydrogen level went up Percentage Area Fraction of defects, Maximum Feret and Number of Pores/cm2 increased accordingly (Figg. 7,8,10). A little decrease of the R values was observed as well, (the lower values for higher hydrogen level castings (HL3)), testifying a trend towards more rounded shape, thus an effect of hydrogen content on pore growth. As expected a dependence of porosity on cooling rate (i.e. the distance from the cold side of the mould) was observed. Fig.7 and Fig.8 show an increase of Percentage Area Fraction of pores and Number of Defects going from the cold to the hot side of the casting. Effect of oxide content Oxides are supposed to act as nucleation sites for porosities. The higher the amount of oxides in the melt the higher the number of pores. Moreover, considering the same amount of hydrogen in the melt, it is supposed to get an increase of the number of pores with a decrease of pore dimensions. This effect could be useful in the production of fatigue specimen to "drive" the porosities dimension. Owing to the high standard deviation of measurments, the expected behavior could not be consistently inferred, nevertheless Fig. 11(a) suggests a general increase in the number of pores nucleated and grown, in function of the oxidation time. Moreover, Fig 11 (b) suggests a trend of Feret max decreasing with oxidation time, for the highest hydrogen level only. As a final remark it must be underlined that the control of the melt oxidization by dry air addition in the melt was really a challenging issue: the fine and dispersed bubbles of dry air delivered to the melt with the rotary impeller, several times removed the hydrogen from the melt and cleaned the melt itself from the largest part of oxides, making them float towards the surface.
(a) (b) Fig. 11: Pore Number per cm2 (a) and max Feret of the pore (b) in function of oxide level 4. CONCLUSIONS In this work, an experimental method to obtain sets of fatigue specimens with nearly identical microstructure and defects content was proposed. First, a casting tool was designed and simulated in order to allow the production of a set of specimens in nearly identical processing conditions.
127
Second, a casting experiment was performed by controlling all of the process variables that have an influence on the generation of microporosity: cooling rate, hydrogen level, oxide presence. For this aim, a specific melt treatment procedure based on the set up of hydrogen content and oxide generation was developed and validated. Metallurgical analysis on cast material confirmed the influence of increased hydrogen level on pores growth [7]-[8] and showed the increment of porosity percentage area fraction, pores dimensions and the evolution towards more rounded pore morphology. The control of the melt oxidation was difficult to achieve. This is probably due to the particular way to increase the oxide level that adopted the use of dry air. Despite this, the experience showed a good repeatability of experimental results, suggesting its usefulness as a prospective evaluation of numerical models for the prediction of gas porosity defects. REFERENCES [1] Wang QG, Apelian D, Lados DA. Fatigue behavior of A356-T6 aluminum cast alloys. Part I. Effect of casting defects. J Light Metals 2001;1:73. [2] Wang QG, Apelian D, Lados DA. Fatigue behavior of A356/357 aluminum cast alloys. Part II. Effect of microstructural constituents . Journal of Light Metals 1 (2001);85:97. [3] Ceschini L, Morri, A, Sambogna G. The effect of hot isostatic pressing on the fatigue behaviour of sand-cast A356-T6 and A204-T6 aluminum alloys. Journal of materials processing technology 2008;204:231-238. [4] J.Campbell and R.A. Harding, "TALAT: The freezing of Castings", European Aluminum Association, (1994) Lecture 3204. [5] J. Campbell, Casting Practice - the 10 rules of castings, Elsevier, Oxford, 2004 [6] ASTM E3-01 (2007) Standard Practice for Preparation of Metallographic Specimens [7] J-Y. Buffiere, S. Savelli, P.H. Jouneau, E. Maire, R. Fougères, "Experimental study of porosity and its relation to fatigue mechanisms of model Al-Si7-Mg0.3 cast Al alloys", Material Science and Engineering A316 (2001) 115-126 [8] D. Dispinar, A. Nordmark, J. Voje, and L. Amberg, Influence of Hydrogen Content and Bifilm Index on Feeding Behaviour of A1-7SÌ Alloy, Shape Casting: 3rd International Symposium 2009, pp. 63-70 [9] J. A. Dantzig, M. Rappaz, Solidification, EFPL Press, 1st edition, 2009, Lausanne Acknowledgements The authors would like to acknowledge Ferrari SpA and in particular Eng. Gianluca Pivetti for general collaboration, Fonderie Scacchetti and Eng. Lorenzo Pivetti, FOSECO-Italy for providing technical facilities. Great acknowledges also to Franco Iorio (Modelleria CPC), for the mould production.
128
Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbeil, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
EFFECTS OF GRAVITY ON THE COLUMNAR TO EQUIAXED TRANSITION IN DIRECTIONAL SOLIDIFICATION. Wajira U. Mirihanage, David J. Browne School of Electrical, Electronic and Mechanical Engineering, University College Dublin, Belfield, Dublin 4, Ireland Keywords: CET, Grain transport, Solidification Abstract In industrial casting processes microstructure plays a major role in determining the properties of the final cast product. Columnar to equiaxed transition (CET) is a frequent result of the evolving grain structure during alloy solidification. In this contribution, we analyze CET in directional solidification via numerical simulations. The numerical model employs front tracking to track columnar growth and a volume average approach to account for the evolution of the equiaxed zone. The effects of gravity, thermal natural convection and dendrite transport were integrated into the model. Simulations of vertical directional solidification of an Al-7%wt.Si alloy both in and opposite the direction of gravity were conducted for different cooling conditions. Here, we present a preliminary analysis of these numerical simulations and a comparison of the predictions with previously published theoretical and experimental work. Introduction Dendritic microstructures are the most common type in cast alloys. Columnar and equiaxed grains are normally present in the macrostructure of alloy castings. When both types co-exist, a distinguishable change, known as the Columnar-to-Equiaxed Transition (CET), is often visible. Much research attention has been paid to analytical, experimental and computer modeling of CET phenomena. Various CET models have been presented at the macro/mesoscopic scale [1-8] and most of this work has recently been reviewed by Spittle [9]. However, many CET models ignore the effects of gravity during solidification. Only a few CET models (e.g. ref. [6,7]) consider the effects of gravity on solidification, but they lack explicit tracking of the columnar dendrite front. In the present contribution, a Front Tracking (FT) algorithm [8,10-12] is used to track the columnar growth under the effects of natural convection. A volume average model of equiaxed solidification [13,14] which considered the effects of gravity is used to model the equaxed dendrite evolution during solidification. Both the models are then combined together via considering macroscopic energy, momentum and mass conservation. The CET model was used to simulate vertical directional solidification of an Al-7%wt.Si alloy both in and opposite the direction of gravity, with different cooling rates. The simulation results are analyzed by contrasting the effects of gravity and previous experimental results [15,16].
129
The Computational Model The CET prediction model presented here is a combination of the columnar FT model [10] and the volume averaged equiaxed solidification model [13]. Descriptive details of these two models can be separately found in the previous literature [8,10-14]. Previously, combined CET models were presented for the pure thermal diffusive condition [17] and for the thermal diffusive/convective condition, but ignoring equiaxed dendrite sedimentation [18]. Hence, only the basic details of the model are outlined here. For the columnar front tracking and equiaxed volume averaging model, the transient conservation equations govern the transport of energy, momentum and mass.
ar dt
:
d(uT) t d(vT) dx
dy
du du1 d(uv) — + + ——'- = dt dx dy dv d(uv) dv2 σί dx dy du dv — + — dx dy
-
k pC 1 dp p dx 1 dp pdy
d2T dx2
+
d2T dy2 d2u dx2
+
(1) d2u dy2
μ a^v as, p\_dx2
dy2
(2)
+ {T-Tre/)ßgy+ sy + p
(3) (4)
= 0
Where, u, v ,T, t, p, μ, p, Cp, k and β are velocity in x direction, velocity in y direction, temperature, time, pressure, dynamic viscosity, density, specific heat, thermal conductivity and volumetric thermal expansion coefficient, respectively; Tre/ is the reference temperature; S, P and E are source terms relating to flow resistance from the porous medium, momentum effects from dendrite sedimentation, and latent heat, respectively; g is an external force (gravity). Heterogeneous nucleation at the mould wall is considered for the columnar grains. So at a given nucleation undercooling, columnar grains start to grow from the domain boundaries. For equiaxed grains, free growth from inoculant particles is computed as explained in rei [13], considering a log-normal distribution of commercial inoculant particle diameters [19]. In the current work, growth of dendrite envelopes is computed using a simplified dendrite tip velocity relationship [20]. A low thermal gradient is assumed and the following modified equation is used to calculate dendrite tip velocity v, of Al-7%Si at undercooling AT, (5)
C AT" where, C and n are alloy dependent constants.
130
The transportation of nucleated equiaxed grains with melt flow is performed by considering the system as a slurry [13], but subjected to sedimentation setting. Therefore, dendrite velocity in the vertical direction is the vector sum of the flow velocity and the settling velocity. So, solid crystals do not necessarily stay on the same transport path as the molten metal flow. For this deviation, a momentum difference to the single phase conservation equation arises and is accounted for via a source term {equation no.(3)}. The source term, f i s defined as [14], P
= / , £ at
(6)
For the slurry, viscosity is a function of the volume fraction solid volume/; modified viscosity is obtained using Thomas' empirical approximation, as cited in [21]. The equiaxed dendrites are free to move until dendrites form a coherent network locally. Instead of using a pre-defined coherency fraction, the coherency point is defined as the point where equiaxed envelopes first fill the local space (control volume), i.e. when equiaxed envelope volume fraction reaches unity [14]. The detailed consideration of prediction of dendrite coherency fraction and comparison with experimental observation is the subject of a separate article under preparation. The average dendrite sedimentation/settling speed w is calculated using averaged characteristic dendrite parameters. The complete mathematical description of these calculations is presented in ref. [22]. Average and equal constant density for both solid and liquid (p=ps=pi) is assumed, except in the settling calculations where appropriate densities are considered (p*s>p*i). Where, Ps, pi, P*S and p*i are nominal density of the solid, nominal density of the liquid, actual density of the solid and actual density of the liquid, respectively. The columnar zone and coherent equiaxed zone are treated as a porous medium. In porous media, the phase interaction forces are proportional to the liquid velocity (liquid velocity relative to the medium), and source terms Sx and Sy are given by [11,12], S , = — P
(7)
Ku
Sy = -Ü2. K v
(8)
where AT is a component of the permeability tensor - a physical property of a porous medium. It is common practice to simplify these equations by the assumption that the mush is isotropie [11,12]. Therefore, the permeability tensor K is defined by the Blake-Carman-Kozeny model using a morphological constant and the local solid volume fraction. The source term E in the energy equation accounts for the latent heat effects. Latent heat effects are incorporated by considering changes in the grain volume and local solid fraction [8,10-14]. These changes include the latent heat release during growth as well as absorption of latent heat during re-melting of the solidified grains.
131
*-[«■£♦"
(9)
Here, g s is the local solid fraction, and is calculated by assuming no diffusion in the solid phase and complete mixing in the liquid phase between secondary dendrite arms; so the Scheil approximation is used to calculate internal solid faction. The evolution of equiaxed and columnar solid fractions in each CV are separately but simultaneously calculated by average equiaxed calculation and columnar front tracking, respectively. The columnar front is free to progress until it meets a control volume that is fully occupied by the equiaxed grains. At this point, no further growth of the columnar front is possible and CET is predicted [17,18].
Results and Discussion Model simulations that include gravity-induced natural thermal convection and sedimentation effects were carried out for directional solidification of Al-7%Si. For these simulations, a mould 10cm high and 3cm wide was chosen and the dissipation of heat (cooling) from either top or bottom surface was investigated. All of the other surfaces were treated as adiabatic walls. A mixed boundary condition controls heat extraction from this cold surface, primarily set by choice of value for the heat transfer coefficient h. The chilled surface temperature was kept at a constant 473K. The total number of grain refiner particles in each melt was set to 0.3 x 106 particles per m2 (for 2D) and the initial melt temperature was set at 896K for all cases. Two different heat transfer coefficients (ft = 1,000 Wm"2K~'and h = 3,000 Wm~2K~') were applied for the different bottom and top cooled solidification simulations. For all of these simulations, primary dendrite arm tip velocity (for both columnar and equiaxed growth) was calculated according to the empirical relation given for Al-7%Si in ref. [20]. In these simulations, the columnar solidification front is either moving against the direction of the gravity vector or moving with it. According to basic casting experience, equiaxed zone formation is sensitive to the heat extraction rate from the solidifying melt [23]. Predicted equiaxed and columnar volume fractions are shown in Figure 1. These four simulation results can be compared and analyzed against the previously published experimental results. The possible engulfment of very small equiaxed dendrites in between primary columnar dendrite arms was considered in these simulations. For these computations, columnar primary interdendritic space was calculated according to the dimensionless relationship of Hunt and Lu [24]. It was assumed that some of the equiaxed dendrites are captured in between the columnar dendrite trunks during their growth. This is possible when the columnar front reaches an undercooled region and if equiaxed dendrites are not coherent and are smaller than this columnar spacing. Those equiaxed dendrites with growth orientations that are not parallel to the columnar
132
growth can be blocked by the columnar dendrites. Only equiaxed crystals with growth orientations that are aligned (parallel) with the columnar growth direction can still grow. Such equiaxed dendrites can grow further in an elongated manner, and may look like columnar grains in any post-mortem macrostructure analysis. Thus, no further growth can be expected for such small dendrites in an equiaxed manner. Therefore in the model computations, further growth of these encapsulated equiaxed dendrites is halted. According to the current simulations, the diameters of these dendrites are well below the average grain diameters of the fully equiaxed zone. Similar equiaxed islands encapsulated between columnar dendrite trunks were predicted in previous Cellular Automata -finite difference (CA-FD) directional solidification simulations by Dong and Lee [5]. Recently reported experimental observations on directionally solidified Al3.5%Ni alloy [16] also provide very good experimental evidence of the presence of such equiaxed grains stuck in between the columnar channels.
Figure 1 : As-cast columnar and equiaxed volume fractions (al) bottom cooled and h = 1000 Wm"2K_1 (a2) bottom cooled and h =3000 Wm"2K"' (bl) top cooled and ft = 1000 Wm"2K"' and (b2) top cooled and h =3000 Wm"2K"'. According to the simulation results in Figure 1, one can observe that low cooling rates (low h) contributed to increase the equiaxed zone size and shorten the columnar length, irrespective of the direction of solidification. This simulation outcome agrees with the indirect predictions [23] of formation of the equiaxed zone ahead of a columnar front. A similar experimental outcome was found [15] from a large number of directional solidification experiments. Furthermore, the simulation results agree with general casting experience [23], A snapshot view of the solid and liquid flow transport processes during bottom and top chilled directional solidification (h = 1,000 Wm"2K"'), 30 seconds after the cooling begins, is shown in Figure 2. Here liquid flow and sedimentation velocities are shown to the same vector scale. With the temperature gradient, we can see the liquid metal is circulating due to convection currents
133
and free solid equiaxed dendrites are falling towards the growing columnar front. In comparison to the top chilled solidification, convection currents seem to be relatively weak in the bottom chilled simulation (average flow speeds in the centre: top down - 1 . 0 mras'1 , bottom up ~ 0.2mms"'). In this bottom chilled simulation, the columnar front has progressed slightly further (2.7cm from the chill) than in the counterpart top chilled case (2.6cm from the chill). In the top chilled solidification, hot liquid is rising towards the columnar front and a resulting slight reduction in the undercooling is assumed to be the main reason for this difference. This could be a significant cause of the difference between the CET positions (see Figure 1) for top and bottom cooled solidification, where the top chilled solidification always has slightly lower columnar length. The shape of the columnar front can be seen to deviate slightly from planar, and the horizontal temperature gradient created due to the natural convection may be the cause of this. However, equiaxed dendrite sedimentation toward the columnar front can set conditions that favour early CET in the bottom chill solidification if they can become coherent. If not and they are small enough, free dendrites can sediment between columnar trunks and this can delay CET further. (a)
(b)
Figure 2 : Bottom (a) and top (b) chilled directional solidification (i) temperature and fluid flow (ii) solid fraction and equiaxed dendrite transport, 30 seconds after solidification starts, for h = 1000Wm"2K"' As shown in figure 1, the simulated CET has occurred through a small transition zone rather than at a sharp boundary. In this mixed region, equiaxed dendrites co-exist with columnar trunks. This nature of CET with a mixed zone rather than a sharp boundary between the columnar and equiaxed zone was experimentally observed in the directional solidification experiments conducted by Ares et al. [15]. A similar mixed region between the columnar and equiaxed zones was also reported from the directional solidification experiments with grain refined Al-7%Si and
134
Al-3.5%Ni [16]. The appearance of such a mixed region is understood to be dependent on the equiaxed dendrite nucleation, undercooling and dendrite transport conditions present ahead of the columnar front. There are a number of factors that can cause the rate of equiaxed growth to decrease ahead of the columnar front. These include (i) fewer nucleation events due to low undercooling, (ii) lower dendritic growth rate due to low undercooling, and (iii) high rate of equiaxed dendrite transport towards the superheated liquid. At a high cooling rate, the columnar front can progress quickly and can reduce the undercooled zone. A high rate of cooling can increase convective flow also. Therefore, one or more of these factors can delay the formation of a coherent equiaxed network ahead of the columnar front. However, in instances where no interconnected equiaxed network is present, columnar trunks are expected to grow through the growing equixed dendrites without facing any effective mechanical barrier, and thus a mixed region can be formed. The blocking fraction, the equiaxed envelope volume = 1.0, used here is different to the established mechanical blocking fraction of 0.49 [1], But, notably no direct physical explanation was given for choice of the extended volume fraction value of 0.66, which yields the volume fraction 0.49. Biscuola and Martorano [25] reviewed this mechanical blocking fraction value with deterministic and stochastic models. According to their analysis, an equiaxed volume fraction value of 0.2 for the mechanical blocking criteria closely agreed with their stochastic CAFD model with a solutal blocking criterion [4]. But it should be noted that the equiaxed envelope that was used to define the equiaxed volume fraction in the stochastic CA-FD model was different to that of the Hunt model [1]. According to Hunt, a circular equiaxed envelope was assumed, and the total equiaxed envelope volume was considered as the extended volume. In the stochastic CA model (in ref. [25]), the equiaxed envelope was based on a fine mesh. The latter part of the same ref. [25] describes the way that the columnar front is blocked: via fully impinged equiaxed grain envelopes. This is equivalent to the blocking of columnar front by the justcoherent equiaxed dendrites. Here it would seem that the most important factor is the presence of a continuous mechanical barrier to the growing columnar grains. Such a continuous barrier exists when equiaxed dendrites form an interconnected network.
Conclusions A model of CET in alloy solidification is presented and directional solidification of Al-7%Si is simulated. The CET model is a combination of a columnar front tracking model and an equiaxed volume average model. The combined model considers the effects of gravity on solidification such as thermal natural convection and equiaxed dendrite transport. A modified mechanical blocking criterion is employed in the model. Upward and downward directional solidification cases were simulated with different cooling rates. Simulations show gradual CET transition rather than a sharp boundary between columnar and equiaxed zones. According to the simulations, the effects of gravity promote equiaxed solidification. The simulations results are in agreement with previous directional solidification experiment results found in the literature.
135
Acknowledgements The authors wish to acknowledge the support of the European Space Agency (ESA) via PRODEX funding (contract number 90267). This work is part of the ESA-MAP (Microgravity Applications Promotion) project CETSOL.
References 1. J.D. Hunt, Mater. Sei. Eng., 1984, 65, pp. 75-83 2. S.C. Flood and J.D. Hunt, Journal of Crystal Growth,, 1987, 82, pp.552-560 3. C.Y. Wang and C. Beckermann, Metallurgical and Materials Transactions A, 1994, 25, pp. 1081-1093 4. M.A. Martorano, C. Beckermann, Ch.-A. Gandin, Metall. Mater. Trans. A, 2003, 34, pp.1657-1674 5. H.B. Dong, P.D. Lee, Acta Mater., 2005, 53, pp.659-668 6. A.Ludwig, M. Wu., Mater. Sei. Eng. A, 2005,413-414,ρρ.109-114 7. M. Wu, A. Ludwig, Metall. Mater. Trans. A,, 2006, 37, pp.1613-1631 8. S. McFadden, D.J. Browne, App. Mathematical Modelling, 2009, 33, pp. 1397-1416 9. J.A. Spittle, Inter. Materials Reviews, 2006, 51(4), pp 247-269 10. D.J. Browne, Hunt J.D., Num. Heat. Trans. B, 2004, 45, pp 395-419 l l . J . Banaszek, D. J. Browne , Mater. Trans., 2005, 46(6), pp 1378-87 12. J. Banaszek, S. McFadden, D.J. Browne, L. Sturz, G. Zimmermann, Metal. Mater. Trans. A.., 2007, 38A, pp 1476-84 13. W.U. Mirihanage, D.J. Browne, Comput. Mater. Sci., 46(4), 2009, pp.777-784 14. W.U. Mirihanage, D.J. Browne, Proceedings of Global Innovations in Manufacturing of Aerospace Materials: The 11th MPMD Global Innovations Symposium at the TMS Annual Meeting, Seattle, USA, Feb 14-18, 2010, pp. 249-256 15. A.E. Ares, S.F. Gueijman, R. Caram, CE. Schvezov, J. Crystal Growth, 2005, 275, pp. e319-e327 16. H. Jung, N. Mangelinck-Noël, H. Nguyen-Thi, B. Billia, J. Alloys and Compounds, 2009, 484, pp. 739-746 17. W.U. Mirihanage, S. McFadden, D.J. Browne, Materials Science Forum, 2010, 649, pp.355360 18. W.U. Mirihanage, S. McFadden, D.J. Browne, Proceedings of 3rd International Symposium on Shape Casting held at the 2009 TMS Annual Meeting, TMS, Warrendale, PA, USA, pp. 257-263 19. T.E. Quested, A.L. Greer, Acta Mater., 2005, 53, pp. 4643-4653 20. Ch.-A. Gandin, Acta Mater., 2000, 48, pp. 2483-2501 21. R.S. Qin, Z. Fan, Matter. Sci. Technol. 2001, 17, pp 1149-52 22. W.U. Mirihanage, D.J. Browne D.J., Comput. Mater. Sci., 2010, 50 (1), pp.260-267 23. D.J. Browne, ISIJ International, 2005, 45(1), pp.37-44 24. J.D. Hunt, S.Z. Lu, Metall. Mater. Trans. A, 1996, 27, pp.611-623 25. V. B. Biscuola, M.A. Martorano, Metall. Mater. Trans. A, 2008, 39A, pp.2885-2895
136
Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
SHAPE CASTING: 4th International Symposium 2011 in honor of Prof. John T. Berry
Properties Session Chairs: Glenn Byczynski Sergio Felicelli
Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
Fracture Surface Facets and Fatigue Life Potential of Castings Murat Tiryakioglu School of Engineering University of North Florida Jacksonville, FL 32224 USA e-mail:
[email protected] John Campbell Department of Metallurgy and Materials University of Birmingham Edgbaston, B15 2TT, UK e-mail:
[email protected] Christian Nyahumwa Department of Mechanical Engineering Dar es Salaam Institute of Technology Dar es Salaam, Tanzania e-mail:
[email protected] Keywords: Facets, casting defects, Bifilms, Fatigue potential Abstract Fatigue potential has been studied in cast aluminium alloys with regard to fatigue crack initiation mechanism at casting defects, particularly surface and subsurface defects. The significance of facets are interpreted as the presence of defects in the interior of castings. Furthermore, two varieties of facets have been identified, one originating as a dendrite-straightened bifilm, and the other originating from a slip plane mechanism. Implications of the findings are discussed in terms of the fatigue potential of castings in the absence of defects. Introduction Fatigue failure in metals accounts for 90% of all in-service failures due to mechanical causes [1]. Consequently, much effort has been made to determine the mechanisms of fatigue failure, namely crack initiation and propagation before the final rupture. In castings, cracks initiate almost always from defects, such as inclusions and pores. However facets have been observed on fatigue fracture surfaces in cast alloys, including Al alloys [2,3,4,5], cast irons [5,6,7,8], steels [5], magnesium alloys [9] and Ni-base superalloys [10,11,12,13], These facets have been interpreted as "persistent slip bands" [2,3] and assumed to represent the metallurgical or ideal fatigue failure in the absence of defects. In one of these studies, Nyahumwa et al. [2] investigated the effect of casting techniques and hot isostatic pressing (HIP) on fatigue life on A356 aluminum alloy castings. For each fatigue fracture, the type of fatigue crack initiators was determined through extensive fractography. Nyahumwa et al. found that a majority of the specimens failed at fatigue cracks initiated at oxide bifilms and pores, which degraded fatigue life significantly. In the specimens with the highest fatigue lives, however, facets, originally interpreted as persistent slip bands, were reported to be involved in the process of fatigue crack initiation and Stage I crack growth. Assuming that facets
139
represented defect-free failure, Nyahumwa et al. [14] stated that the fatigue life potential of cast aluminum alloys is several orders of magnitude higher than the values usually obtained in fatigue tests, due to the presence of casting defects. However, there is considerable evidence in the literature that facets form around casting defects in fatigue testing. Therefore assumptions and interpretations about facets in fatigue failure, in particular the assumption that the appearance of facets represents an 'ultimate' metallurgical limit to the fatigue life of castings, need to be revisited. The present study is intended to review the observations about facets in fatigue failure of castings and build a case for the mechanism of their formation and their implications on fatigue life potential in castings. Fatigue Crack Initiation Mechanism for Formation of Facets The fatigue crack initiation mechanism that leads to the formation of facets on fatigue fracture surfaces of aluminium alloy castings has been a source of speculation. One of the authors [15] interpreted facets found by Jang et al. [4] in aluminum castings as oxide bifilms straightened by the advance of the dendrites during solidification. This dendrite pushing process creates large planar areas, which are separated by the planar unbonded interface between the two films, thus forming extensive transgranular, and sometimes intergranular, cracks. However, Wang et al. [16] conducted energy dispersive spectrometry (EDS) on the facets in A356 castings presented in Figure 1. They found only Al with small amounts of Si and Mg, no Fe, and a trace of O. Hence the facets in Figure 1 are planes through a primary Al dendrite. However, Wang et al. also identified oxide bifilms associated with the some facets, including the ones in Figure 1, and provided two scenarios: (i) casting defects (bifilms) cause slip concentrated in the resulting shear plane, or (ii) no significant defect present initially and the crack initiates from a shear plane. Wang et al. indicated that there is no totally convincing evidence to favor either hypothesis, and it seems quite likely from their observations that both cases occur. Evidence for two quite different facet forming mechanisms is presented in Figures 1 and 2. In Figure 1 the clean, mirror-smooth facets with sharply delineated steps and edges are expected to be typical of those generated by a slip plane mechanism. Figure 2 shows a facet that does not display mirror smoothness, and is characterised by what appear to be a distribution of pore fragments across its surface. Interestingly, close examination reveals its surface to flow seamlessly in places into the surrounding oxide folds. In particular, the oxide that surrounds the sand grain, necessarily present as a result of its entrainment via the oxidized liquid surface, is seen to connect smoothly and become contiguous with the facet surface, suggesting the facet has an origin as an oxide film. It is suggested that this variety of facet is formed by the advance of dendrites, straightening oxide bifilms. The bifilms would contain minute irregular bubbles from its own entrainment event, explaining the features observed on the face of the facet. In addition, the facet is not particularly smooth, and will clearly depend on the degree of perfection of supply of feed metal. If the supply is poor liquid will be sucked away from the bifilm into the dendrite mesh, leaving only dendrites clearly revealed. Alternatively if feeding is over-generous liquid will exude slightly between dendrite arms. Figure 3 illustrates the situation commonly observed by Nyahumwa [17] that the two varieties of facet can occur together. In fact in Figure 3 most of the observed surface consists of the Type 2 (bifilm) facets, with a small central region of Type 1 (slip plane) facet, and the remaining area in the bottom left hand comer appears to be a non-straightened area of bifilm. Kunz et al. [10,11] investigated the high cycle fatigue failure in cast Inconel Ni-based superalloys and found facets in areas adjacent to large casting defects, which is consistent with the observations
140
of Price [13] in another Ni-based superalloy. Kunz et al. attributed the formation of facets to the stress concentration effect of casting defects. They increase the local stress amplitude, promote the slip on adjacent slip planes and contribute to the decohesion process. Kunz et al. also observed that facets (i) enclose the casting defect, which was identified as the crack initiator, (ii) are in the vicinity of the crack initiator, or (iii) intersect with the casting defects, which "evokes impression that the development or growth of facets is not influenced by this porosity."
Figure 1. Facets formed in a cast A356 alloy near an oxide bifilm [16]
Figure 2. Faceted fatigue fracture in a specimen with a sand inclusion [17].
Fatigue studies by Reed [18] found facets on fracture surfaces of single crystal U720 alloy. She argues that facets are the result of a slip-band decohesion process that can operate under the conditions of cyclic stress which seems likely for this special failure mode. Boyd-Lee [19] investigated the formation of facets in a forged Ni-based superalloy and found the fatigue crack growth rate in the facet to be an order of magnitude higher than in the other parts. Boyd-Lee suggested the following mechanism for the formation of facets:
141
Figure 3 : A faceted trans-granular appearance on fatigue fracture surface indicating that persistent slip bands (PSBs) were involved in the process of crack initiation and Stage 1 crack growth in an unfiltered and HIPed Al-7Si-Mg alloy casting. Nf = 4.4 x 105 cycles at a maximum stress of 240 MPa and stress ratio R =+0.1 [17]
1. Active slip bands form to dissipate stress concentrations. 2. During load cycling, strains in active slip bands become decreasingly reversible. 3. Stress concentrations increase until slip bands form on parallel planes. This and initial slip band formation result in the observed 'intense' slip bands. 4. Eventually there occurs a zone where the dislocation density has built sufficiently resulting intense plane-normal stresses exceed the bond strength of the material. 5. Hence, a ligament in the slip band unzips, and part of the released energy contributes to surface area increase, resulting in the exposure of fresh microstructure. The suggestion that slip bands form to dissipate stress concentrations is noteworthy. For persistent slip bands to appear, stress range and crack length need to be small [20]. However there has to be a preexisting crack that causes the stress concentration. This argument implies that a casting defect needs to be present to initiate the formation of a facet but the resultant stress concentration should be low. Such a defect has to be subsurface (internal) so that stress concentrations will be low as shown in detailed finite element studies. For instance Borbely et al. [21] showed that internal failure originating from pores should be limited to the ones in which defects lie very close to the surface, as observed by Staley et al. [22] in HIPped A206 castings. Such a defect is presented in Figure 2 [17] which shows facets formed around a sand inclusion in an A356 casting. As it happens, this particular facet does not appear to originate from a slip band mechanism but from a bifilm source. However, the behavior is essentially the same as discussed further below. The observed fatigue faceted failures were perhaps originated at the interface between aluminium matrix and the subsurface oxide film defects at which stress concentration generated persistent slip bands (PSBs) during cyclic loading. The stress intensity at the interface between a subsurface defect which is considered embryo crack and aluminium matrix is normally raised by a portion of material close proximity to a free surface. During the cyclic loading, the fatigue crack possibly starts growing towards the free surface of the specimen due to high stress concentration at the vicinity of the subsurface defect edge before it propagate in other directions. The slip bands were probably involved in Stage I crack growth at the material confined between the free surface and the oxide film defects as shown in Figures 2 and 3.
142
Effect of Subsurface Crack Initiators on Fatigue Potential of Castings When attempting to understand the action of slip plane or bifilm facets on failure during fatigue, it is perhaps a surprising simplification to realize that their actions will be expected to be nearly identical. This follows because the slip plane and the straightened bifilm can both be viewed as acting like cracks. Furthermore, their size is closely similar because both are initially limited to the grain size. Subsurface crack initiation in fatigue has received considerable attention recently [23,24,25,26,27,28,29], The S-N curve for the subsurface defects is different from the one for those on the surface, as shown schematically in Figure 4. The location of the S-N curve for internal defects will depend on the size of the internal defect. [30]. Separate S-N curves for surface and subsurface defects were found [27] for a cast Al-10%Si-4%Cu-0.6%Mg alloy when specimens were subjected to a surface treatment. In several studies, cracks were observed to grow from structural defects at or shortly after the first stress cycle [31,32,33]. However in cases where these is no defect on the surface large enough to initiate a propagating crack, fatigue crack initiation from a subsurface defect takes significantly longer, resulting in increased fatigue life. Results in a fatigue experiment may come from castings, some of which have defects on the surface while the rest have subsurface defects, we can expect to have two fatigue life distributions, as shown schematically in Figure 5. The fatigue life data for A356 castings from literature [17] are presented in Figure 6. The logaverage fatigue life of castings failed from subsurface defects is 11 times that of the castings failed at surface defects at a maximum stress of 150 MPa, and 7 times at 240 MPa. Moreover, the fatigue life results of HIPed castings were observed to have even higher fatigue life than those failed from subsurface defects [22]. The fatigue life of the HIPed castings ranged from 1.7 * 107 to 7.6 χ IO7 cycles for the specimens suspended from the experiment (run-outs). The castings were assumed to be defect free. This is a proof that fatigue lives of castings observed to fail from persistent slip bands emanating from subsurface oxide film defects in the same study are several orders of magnitude lower than the fatigue life potential. Hence, one would expect that the real fatigue potential of aluminium alloy castings in the absence of defects will be above 7.6 χ IO7 cycles.
S
Nf Figure 4. The schematic illustration of S-N curves corresponding to surface and internal (subsurface) defects.
143
s
Figure 5. The schematic illustration of S-N curves showing how a sample of fatigue specimens tested at a stress level may have data with different defect locations resulting in two distinct statistical populations. The existence of faceted fracture in specimens that fail from internal defects were observed in magnesium and titanium alloys as well. Tokaji et al. [34] found in an AZ31 magnesium alloy that in subsurface fracture, facets were always present at the crack initiation site, which were located mostly close to the surface, and the facet sizes were nearly the same as, or smaller than, the average grain size. The same authors [35,36] observed subsurface fracture in beta titanium alloys and indicated the presence of smooth facets at the crack initiation site. Implications of Faceted Fatigue Failure in Castings Based on the discussion presented above, we can state the following about faceted failure in castings: 1. Type 2 (Bifilm) facets exist in the as-cast matrix, and are common defects to initiate Type 2 (slip plane) facets. 2. In the absence of surface defects, subsurface casting defects cause local stress concentrations under cyclic loading, leading to the formation of slip bands around the casting defects. 3. With increasing number of fatigue cycles, the material along the slip bands work-hardens, reducing ductility and therefore the energy needed to separate the atoms is also reduced. 4. When the minute deformations in the material around the oxide defect accumulate and separate the two halves of the bifilm sufficiently, the material ruptures along the slip band and facets and a propagating fatigue crack form. This explains why the facets will follow crystallographic planes [10]. 5. During propagation the crack follows similar slip bands around other casting defects to minimize energy, but crack growth rate is higher in facets. Consequently, the fatigue life data of Nyahumwa et al. [2] can be reinterpreted as presented in Figure 7 [37]. The presence of the two fatigue life distributions is due to the location of defects when the specimens are excised which also determines whether facets form on fracture surfaces. In either case, the failure is due to the presence of casting defects, namely bifilms. Therefore the concept of fatigue life potential of aluminum castings proposed by Nyahumwa et al. [14] needs to
144
be revised. It is the authors' opinion that fatigue life of aluminum castings will be much higher than the data collected so far, all of which seem to have come from castings with defects.
a. E
σ
150
100 10"
IO5
IO6
Nf
Figure 6. S-N curves illustrating life of fatigue specimens tested at maximum stress levels 150 and 240 MPa at stress ratio +0.1 and failed at surface and subsurface defects resulting in two distinct SN curves (Data from [17]). Conclusions • Type 2 facets (flattened bifilms) are formed during casting and solidification. • Type 1 facets (slip planes) form around casting defects which are subsurface when there are no defects on the surface large enough to generate a propagating crack. • Types 1 and 2 facets act similarly to promote fracture. • A fatigue crack initiation mechanism at the subsurface defects for the formation of facets has been proposed. • Improvement in fatigue life of castings with subsurface defects is attributed to a longer crack initiation period. The average fatigue life of castings failed from subsurface defects is 14 times that of the castings failed at surface defects at 150 MPa, and 6 times at 240 MPa. • The fatigue life potential of aluminum castings is much higher than the data in the literature.
145
3
ff
0
-2
-5 11
12
13
14
15
16
17
ln(Nf) Figure 7. Weibull probability plot [37] of the fatigue life data by Nyahumwa et al. [2], References 1. G. E. Dieter. Mechanical Metallurgy.McGraw Hill, p. 375, 1986. 2. C. Nyahumwa, N.R. Green, and J. Campbell, "Influence of Casting Technique and Hot Isostatic Pressing on the Fatigue of an Al-7Si-Mg Alloy," Metallurgical and Materials Transactions A, Vol. 32A, (2001), pp. 349 - 358. 3. Q.G. Wang, D. Apelian, and D.A. Lados, "Fatigue Behavior of A356-T6 Aluminum Cast Alloys: Effect of Casting Defects - Part 2", Journal of Light Metals, vol. 1, issue 1, pp. 85-97, 2001. 4. Y.H. Jang, S.U. Jin, Y.I. Jeong, S.S. Kim: Metall. Mater. Trans. A, 2009, vol. 40A, pp. 157987. 5. E. Bayraktar, I. M. Garcias, C. Bathias: International Journal of Fatigue 28 (2006) 1590-1602. 6. J.H. Bulloch: Theoretical and Applied Fracture Mechanics 24 (1995) 65-78 7. J. Yang, S.K. Putatunda: Materials Science and Engineering A 393 (2005) 254-268. 8. G.L. Greno, J.L. Otegui, R.E. Boeri: International Journal of Fatigue 21 (1999) 35^13. 9. M.F. Horstemeyer, N. Yang, K. Gall, D.L. McDowell, J. Fan, P.M. Gullett: Acta Materialia 52 (2004) 1327-1336.
146
10. L. Kunz, P. LukâS, R. Koneöna: "Initiation and propagation of fatigue cracks in cast IN 713LC superalloy", Engineering Fracture Mechanics 77 (2010) 2008-2015 11. L. Kunz, P. Lukas, R. Konecnâ: "High-cycle fatigue of Ni-base superalloy Inconel 713LC", International Journal of Fatigue 32 (2010) 908-913. 12. D. Gelmedin, K.-H. Lang: Procedia Engineering 2 (2010) 1343-1352. 13. C E . Price: Metallography, Volume 17, Issue 4, November 1984, Pages 359-370 14. C. Nyahumwa, N.R. Green, and J. Campbell, "The Concept of the Fatigue Potential of Cast Alloys," J. of the Mechanical Behavior of Materials, Vol. 9, No. 4, (1998), pp. 227 - 235. 15. J. Campbell: Metall. Mater. Trans. A, 18—VOLUME 41A, JANUARY 2010 16. Q.G. Wang, C.J. Davidson, J.R. Griffiths, P.N. Crepeau: "Oxide Films, Pores and Fatigue Lives of Cast Aluminum Alloys", Metall. Mater. Trans. B, v. 37B, pp. 887-895, 2006. 17. C.W.M. Nyahumwa: Influence of Oxide film filling defects on fatigue properties of cast A1-7SÌMg Alloy, PhD. Thesis, University of Birmingham, UK, 1997. 18. P. A. S. Reed: Mater. Sci. Technol., 2009, 25, (2), 258-270. 19. A.D. Boyd-Lee / International Journal of Fatigue 21 (1999) 393^105 20. C. Laird: in: "Physical Metallurgy", ed. R.W. Cahn, P. Haasen, Elsevier, Volume 3, p.2379, 1996. 21. A. Borbely, H. Mughrabi, G. Eisenmeier, H.W. Höppel: International Journal of Fracture, v. 155, pp. 227-232,2002. 22. J.T. Staley, Jr., M. Tiryakioglu, J. Campbell: Materials Science and Engineering A 465 (2007) 136-145. 23. O. Umezawa, K. Nagai and K. Ishikawa: Internal crack initiation in high cycle fatigue for Ti-5 At-2.5 Sn ELI alloy at cryogenic temperatures, Tetsu to Hagane, 75(1) (1989), 159-166. 24. H. Mughrabi: "On Multi-Stage Fatigue Life Diagrams and the Relevant Life-Controlling Mechanisms in Ultrahigh-cycle Fatigue", Fatigue and Fracture of Engineering Materials and Structures, v. 25, pp. 755-764. 25. K. Sadananda, A.K. Vasudevan, N. Phan: International Journal of Fatigue 29 (2007) 2060-2071 26. C. Przybyla, R. Prasannavenkatesan, N. Salajegheh, D. L. McDowell: International Journal of Fatigue 32 (2010) 512-525 27. Y. Nakamura, T. Sakai, H. Hirano, K.S. Ravi Chandran: International Journal of Fatigue 32 (2010)621-626 28. C. Bathias: Fatigue Fract Engng Mater Struct 22, 559-565, 1999. 29. Q. Y. WANG, J. Y. BERARD, 1 A . DUBARRE, 2 G. BAUDRY, 2 S. RATHERY1 and C. BATHIAS: Fatigue Fract Engng Mater Struct 22, 667-672, 1999. 30. O. Umezawa, K. Nagai: "Subsurface Crack Generation in High-cycle Fatigue for High Strength Alloys", ISIJ International, Vol. 37 (1 997), No. 12, pp. 1170-1 179 31. B. Skallerud, T. Iveland and G. Härkegärd: Eng. Fracture Mech., 1993, v. 44, 857-874.
147
32. S.A. Barter, L. Molent, N. Goldsmith and R. Jones: J. Eng. Failure Analysis, 2005, v. 12, pp. 99-128. 33. B.R. Crawford, C. Loader, A.R. Ward, C. Urbani, M.R. Bache, S.H. Spence, D.G. Hay, W.J. Evans, G. Clark, A.J. Stonham: Fatigue Fract. of Eng. Mater. Struc, 2005, v. 28, pp. 795-808. 34. K. Tokaji, M. Kamakura, Y. Ishiizumi, N. Hasegawa: "Fatigue behaviour and fracture mechanism of a rolled AZ31 magnesium alloy", International Journal of Fatigue 26 (2004) 1217-1224. 35. Tokaji K, Bian JC, Ogawa T, Nakajima M. The microstructure dependence of fatigue behaviour in Ti-15Mo-5Zr-3Al alloy.Mater Sei Eng 1996;A213:86-92. 36. Tokaji K, Ohya K, Kariya H. Subsurface fatigue crack initiation in beta titanium alloys. Fatigue Fract Eng Mater Struct 2000;23:759-66. 37. M. Tiryakioglu, J. Campbell: Metall. Mater. Trans. A, v. 41A, pp. 3121-3129, 2010.
148
Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
EFFECT OF HOLDING TIME BEFORE SOLIDIFICATION ON DOUBLEOXIDE FILM DEFECTS AND MECHANICAL PROPERTIES OF ALUMINIUM ALLOYS M. El-Sayed1, H. Salem2, A. Kandeil1, W. D. Griffiths3 'Arab Academy for Science and Technology and Maritime Transport; Abu Qir, P.O. Box 1029; Alexandria, 21599, Egypt. 2
American University in Cairo; AUC Avenue, 5th district, P.O. Box 74; New Cairo, 11833, Egypt. university of Birmingham, Edgbaston; Birmingham, United Kingdom. B15 2TT. Keywords: Double oxide film defects, Aluminium, casting, mechanical properties Abstract
Double oxide films (bifilms) have been held responsible for the variability in mechanical properties of aluminium castings. It has been suggested that the air entrapped inside a bifilm can react with the surrounding melt leading to its consumption, which might improve the mechanical properties of the castings. In this work, the effect of the holding time of the melt before solidification on the distribution of mechanical properties, and by implication, on entrained double oxide films, was investigated for different aluminium alloys. The Weibull moduli of the plate castings were determined under tensile conditions, and their fracture surfaces examined for evidence of oxide films. The results suggested the occurrence of two competing mechanisms during the holding treatment. The consumption of air inside the bifilms due to reaction with the surrounding molten metal may lead to improvements in mechanical properties, but this may be accompanied by hydrogen passing into the bifilms, which has a deleterious effect on properties. Introduction One of the most important casting defects affecting the reproducibility of mechanical properties of aluminium castings is the double oxide film defect [1, 2], created due to surface turbulence of the liquid metal, a common feature during metal transfer and pouring in the shape casting process. When the liquid metal surface is exposed to air, a surface oxide film forms. As a result of surface disturbance, the liquid metal surface can be folded over onto itself, causing the oxidised surfaces of the folded-over metal to come together but not to fuse, trapping a layer of the local atmosphere between them, and creating a double oxide film defect or "bifilm" [1, 2] which can be entrained into the bulk metal, as shown in Figure 1. Such entrained double oxide film defects represent one of the easiest possible initiating features for cracks, since their unbonded inner surfaces can be separated with minimal effort. Also, gas dissolved in the liquid metal can precipitate inside the bifilm gap initiating porosity [3]. In addition, double oxide films are favourable sites for the nucleation and growth of intermetallic
149
compounds. These effects not only reduce the elongation, tensile strength and fatigue limit of aluminium alloy castings, but also increase their variability.
Figure 1: The formation of a double oxide film defect (1) surface turbulence leads to a breaking wave on the metal surface, and (2) the two unwetted sides of the oxide films contact each other leading to the submerging of the bifilm into the bulk liquid metal. Nyahumwa et al. [4] suggested that, due to the transformation of the oxide layer from γ-Αΐ2θ3 to α- ΑΙ2Ο3, ( a process thought to take about 5 hours), cracks are introduced into the oxide which allows the liquid aluminium to come into contact with, and react with, the air inside the oxide film defect (mainly oxygen and nitrogen). This mechanism could result in the consumption of the atmosphere inside the bifilm and possibly lead to its deactivation. The rate of consumption of the internal atmosphere has been examined by Raiszadeh and Griffiths [5], who trapped an air bubble inside liquid Al and monitored its change in volume with time using real-time x-ray radiography. Their results showed that the oxygen in the trapped air should be consumed first, to form AI2O3, then the nitrogen would react to form A1N. These reactions started immediately, (with no need for an initiating phase transformation). Also, if the initial hydrogen content of the melt was higher than the equilibrium associated with the ambient atmosphere, hydrogen diffused into the trapped air bubble, increasing its volume, which supported the idea that double oxide film defects could act as initiation sites for hydrogen porosity during the solidification of Al castings. The reaction rates of the trapped air with the melt were utilized to build a semi-empirical mathematical model capable of predicting the duration of the atmosphere inside the double oxide film defect, which suggested that the consumption of oxygen and nitrogen inside the defect would not take more than about three minutes. The aim of this work was to study the effect of the holding time of the melt before solidification on entrained double oxide films, and the corresponding change in the mechanical properties of aluminium alloy castings. Understanding these issues could lead to the development of techniques by which the effect of double oxide film defects might be reduced or eliminated in aluminium castings.
150
Experimental Procedure The experimental procedure involved the production of castings (by the investment casting technique), which contained oxide films of different ages; 0, 10 and 20 minutes. Three different aluminium alloys were considered in this work, commercial purity Al, Al-7wt.%Si-0.3wt.%Mg (2L99 alloy) and Al-5wt.%Mg aluminium alloy, so as to involve different oxide films which might have different behaviours, (AI2O3, MgAl2C>4 and MgO, respectively). In each experiment about 10 kg of the alloy was melted and held at about 800°C under a vacuum of about 80 mbar for one hour, a procedure intended to remove most, or all, previously introduced oxide films from the melt [6]. The liquid metal was then poured into preheated ceramic shell moulds, which were then placed in an induction furnace and stirred using a power setting of 7.5 kW and frequency of 2350 Hz, for one minute. This led to splashing of the liquid metal surface, and the creation and entrainment of new double oxide film defects, and their introduction into the melt. One casting was then allowed to solidify immediately, while two further castings were maintained in the liquid state by placing the filled ceramic shell mould in a furnace for 10 and 20 minutes, respectively, before removal and solidification. During holding, the hydrogen content of the melt was evaluated using a Hyscan H-measuring device. After solidification, each of the castings were machined into fifteen tensile test bars with the shape and dimensions shown in Figure 2, and tested using a Zwick 1484 tensile testing machine, with a strain rate of 1 mm min"'. Tensile results were evaluated using a Weibull statistical analysis approach to assess the influence of the holding treatment on the variability of the mechanical properties of the castings. Finally, SEM with EDX analysis was used to investigate the fracture surfaces of the test bars.
Figure 2: Sketch of the casting and the tensile test specimens taken.
151
Results Table 1 shows the results of the Weibull analysis of both of the UTS and percentage elongation values obtained from the different aluminium alloy castings, together with the position parameter and R2 values of the linear fits to the Weibull plots, as well as the results from the hydrogen measurements. Figure 3 illustrates the effect of the holding time before solidification on (a) Weibull Moduli of the UTS, (b) Weibull Moduli of the % Elongation and (c) the amount of H in solution with the liquid metal. Table 1: Results of the Weibull analysis for the test bars of different Al alloys that contained oxide films of age 0, 10 and 20 minutes.
H(cm 3 /100g)
UTS (MPa)
%
Elongation
Weibull modulus Position parameter (MPa) R2 Weibull modulus Position parameter (%E1.) R2
Commercial purity Al alloy 0 10 20 min. min. min.
Al-7Si-0.3Mg alloy
0.10
0.15
0.28
33
36
52
0 min
20 min.
0.08
10 min. 0.10
30
37
55
53
0.96
0.95
7.95
Al-5Mg alloy 10 min.
0.15
0 min. 0.91
1.0
20 min. 1.22
39
34
22
31
24
139
142
125
193
192
187
0.94
0.95
0.93
0.96
0.93
0.97
0.96
9.06
7.06
7.2
9.35
8.4
9.7
13.8
8.4
34
37
30
3.08
2.51
3.21
26
28
27
0.87
0.92
0.94
0.87
0.88
0.88
0.92
0.97
0.99
In all three alloys both the UTS and %Elongation Weibull Moduli were a maximum in the case of the casting held for 10 minutes in the liquid state before solidification, although in the case of the pure aluminium and Al-7Si-0.3Mg alloys the differences in Weibull Moduli were slight. In the case of the Al-5Mg alloy the increase in Weibull Moduli was most marked. The table also shows that the hydrogen content of the alloy consistently increased with holding time, but the Al-5Mg alloy possessed a greater hydrogen content due to the greater solubility of hydrogen in this alloy [7]. Figure 4(a) shows an SEM image inside a pore on the fracture surface of a specimen of Al-5Mg alloy. Many oxide fragments were visible inside the pore, and EDX analysis of the fragments indicated the presence of MgO, suggesting that the origin of this pore lay with a double oxide film defect. Also, Figure 4(b) shows an SEM image with the corresponding EDX analysis for the fracture surface of a specimen of Al-7Si-0.3Mg alloy, in which an iron intermetallic is associated with a spinel substrate, suggesting a nucleation relationship. This would be an indication of the role played by double oxide films in creating other defects such as porosity and intermetallics.
152
(e) Figure 3: Plot of the holding time versus (a) Weibull Modulus of UTS, (b) Weibull Modulus of % Elongation and (c) H content of the melt.
(a) (b) Figure 4: Consequences of entrainment of bifilms inside Al castings, (a) oxide-associated porosity and (b) oxide-related intermetallic formation.
153
Figure 5 shows whisker-like oxides found within pores of castings held for 20 minutes in the liquid state before solidification, (in the case of commercially-pure Al and Al-5Mg alloys). The pores were also associated with oxide films. The delicacy of these features suggests they formed during solidification, rather than earlier, say, within an oxide film defect floating within the liquid metal. Although the interconnections are too small to influence mechanical properties, their presence may be informative about conditions inside the pores during their formation.
(a) (b) Figure 5: Interconnections between adjacent oxide layers in castings containing 20-min. old oxide films, (a) commercially pure Al, (b) Al-5Mg alloy. Discussion As described in Table 1, the Weibull Moduli of the commercial purity Al alloy and A1-7SÌ0.3Mg alloy consistently exceeded 30, a value of Weibull Modulus often associated with a casting made with a well-designed running system, with reproducible properties [8]. Also, oxide films, as demonstrated by EDX analysis in the SEM, were mostly associated with pores, (as shown in Figure 4(a), rather than lying on the fracture surfaces. This could be an indication of the effect of the holding treatment on the elimination of oxide films. The Al-5Mg alloy exhibited the largest differences between the Weibull Moduli with different holding periods. This may be associated with the high solubility limit for H in this alloy, and also the permeability of its oxide film, MgO, that could lead to a more rapid reaction of O and N with the surrounding melt, as well as enhance the diffusion of H into the oxide film defect interior. These results may therefore be an indication of the significance of the H content in controlling the scatter of mechanical properties, as well as oxide film defects.
154
The consequences of the entrainment of double-oxide films were illustrated in Figure 4. They not only act as cracks in the solidified casting. Gas dissolved in the liquid metal can precipitate inside the bifilm gap initiating porosity, and in addition, double oxide films are favorable sites for the nucleation and growth of a wide variety of intermetallics. This might minimize the effect of the holding treatment on the mechanical properties of the castings, especially in the case of the commercial purity Al alloy and Al-7Si-0.3Mg alloy. The whisker-like structures within pores, shown in Figure 5, demonstrated by EDX to be an oxide, occurred mostly at holding periods of 20 minutes. The whisker-like growths are suggestive of ceramic structures grown from a vapour phase, and suggest the formation of oxide structures within porosity that contains an atmosphere. This in turn suggests that at 20 minutes holding time any double oxide film defects may have still contained an atmosphere, which may be separating their internal surfaces and preventing any bonding. The defects, therefore, would still be expected to have a deleterious effect on mechanical properties. In this work it has been found that, for all three alloys, (with their different oxides), the Weibull Moduli representing the UTS and % Elongation rose to a peak at holding times of 10 minutes, although the H content of each alloy rose progressively with holding time, as illustrated in Figure 3. The moduli decreased significantly when the holding period was increased to 20 minutes. This is suggestive of competing mechanisms affecting the distribution of mechanical properties. The first mechanism may be related to the consumption of air inside the bifilms due to reaction with the surrounding molten metal, while another mechanism may be related to the amount of hydrogen picked up by the liquid metal from the furnace atmosphere (and hence the porosity size in the resulting casting). Previous experiments had suggested that the interior atmosphere inside double oxide films could be consumed within a few minutes [5]. Hence some bonding of the two layers of the bifilm might then take place, or some reduction in the size and hence impact of the oxide films could occur, which might improve the mechanical properties of the castings. This mechanism would tend to increase Weibull Moduli with time. But on the other hand, the H content increased as the liquid metal was held in the furnace for longer periods, and this would lead to a decrease in overall mechanical properties due to increased porosity, and lead to a decrease in the Weibull Modulus with time. The Weibull Moduli for the castings allowed to solidify immediately represented the scatter of mechanical properties derived from castings that contained the lowest hydrogen contents, but that perhaps also contained oxide films then in the process of losing their initial atmospheres. The morphology of these defects seems to have been most harmful in the Al-5Mg alloy. The castings held for 10 minutes in the liquid state may therefore have possessed the best Weibull Modulus (narrowest distribution of properties) because at this time the defects may have lost all or most of their initial internal atmospheres, and may only have absorbed a little hydrogen from the surrounding melt. Under these circumstances the oxide film defects may have had a morphology that was least harmful to the properties of the castings. The mechanical properties of all castings subsequently declined with further holding up to 20 minutes, as the H content of the melt increased with holding time, increasing the deleterious effect of the double oxide film defects and perhaps leading to increased oxide-related porosity.
155
Conclusions 1. SEM examination of the fracture surfaces detected many oxide films, which demonstrated a role for such defects in influencing the failure of Al castings. 2. SEM examination also showed that the interior of porosity associated with double oxide film defects might develop whisker-shaped interconnections between them, apparent in the castings held for 20 minutes before solidification. The interconnections would not have a significant effect on mechanical properties, but could perhaps be indicative of chemical reactions resulting in deposition of ceramic whiskers, which in turn suggests an atmosphere present in the pores during solidification despite the (relatively) long holding periods. 3. The mechanical properties of the castings were most improved after a holding period for the liquid metal of 10 minutes, perhaps due to the consumption of air inside the bifilms and a reduction in their size and effect on casting properties. 4. Increasing the holding period to 20 minutes increased the H content of the alloy. This may have resulted a greater effect of porosity, which would reverse the initial enhancement in the distribution of mechanical properties (due to air consumption), leading to a decrease in the Weibull Moduli to values that were close to those of the castings that were solidified immediately after pouring. 5. The holding treatment may reduce the effect of double oxide film defects in Al melts, but would not prevent them from playing a role in the formation of other defects, such as serving as sites for the formation of hydrogen porosity or intermetallic phases. Acknowledgements The authors would like to thank Mr. Adrian Caden (of the University of Birmingham), Eng. Hanadi Hussein, Eng. Khalid Iraqi and Eng. Ramy Wasfi (of the American University in Cairo) for their technical support during the laboratory work. Also, the authors wish to acknowledge the sponsorship of this study by the Arab Academy for Science and Technology, Alexandria, Egypt. References 1. Campbell, J., Castings. 2nd. ed. 2003: Butterworth-Heinemann. 2. Campbell, J., Entrainment defects. Mat. Sci. Technol., 2006. 22: p. 127-145. 3. Griffiths, W. D. and Raiszadeh, R., Hydrogen, porosity and oxide film defects in liquid Al, J. Mat. Sci., 2009, 44, 3402-3407. 4. Nyahumwa, C , Green, N. R, and Campbell J., Influence of casting technique and hot isoslatic pressingon the fatigue ofanAl-7Si-Mgalloy, Met. and Mat. Trans. A, 2001, 32A: p. 349-358. 5. Raiszadeh, R. and Griffiths, W.D., A Semi-empirical Mathematical Model to Estimate the Duration of the Atmosphere within a Double Oxide Film Defect in Pure Aluminum Alloy. Met. and Mat. Trans. B, 2008, 39(2): p. 298-303. 6. Raiszadeh, R. and Griffiths, W.D., The behaviour of double oxide film defects in liquid Al alloys under atmospheric and reduced pressures. J. Alloys and Compounds, 2010. 491(1-2): p. 575-580. 7. Anyalebechi, P.N., Analysis of the effects of alloying elements on hydrogen solubility in liquid aluminum alloys. Scripta Met. et Mat., 1995. 33(8): p. 1209-1216. 8. Nyahumwa, C , Green N. R., and Campbell J., Effect of Mold-Filling Turbulence on Fatigue Properties of Cast Aluminum Alloys. AFS Trans., 1998. 58: p. 215-223.
156
Shape Casting: The 4lh International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
WEIBULL ANALYSIS OF THIN A356 PLATES CAST WITH THE ELECTROMAGNETIC PUMP GREEN SAND PROCESS Ratessiea Lett1, Sergio Felicelli1, Rafael Cuesta2, John Berry1, J. Antonio Maroto2, Ruth San José2 'Department of Mechanical Engineering and Center for Advanced Vehicular Systems, Mississippi State University, Mississippi State, MS, 39762, USA foundation for the Research and Development in Transport and Energy (CIDAUT), Parque Tecnologico de Boecillo,Valladolid, SPAIN, 47151 Keywords: Aluminum A356, Weibull, four point bend, electromagnetic pump Abstract Four different gating systems were used to produce plates by means of an electromagnetic pump linked up with a green sand molding machine, promoting a counter-gravity of the mould. Three of the systems were multiple gated, whereas one was single gated. Eight cast plates were examined, two from each gating design. The method of four point bend testing was used to obtain information about the mechanical properties of the castings, as this method produces a uniform distribution of the bending stress within the central span of the test specimens. From these results, a Weibull statistical analysis was performed in order to quantify specimen failure rate for each of the configurations. The specimen cross sections were then examined using optical microscopy. This project is funded by the National Science Foundation (NSF) under the International Research and Education in Engineering (IREE) program. Introduction During conventional casting processes, liquid metal is often subjected to agitation and movement at high velocities, producing a final melt with a considerable amount of inclusions and defects which affect mechanical properties of the material, thus resulting in premature failure[l-3]. As aluminum alloys are highly susceptible to oxidation during casting, there exists the opportunity for the formation of a thin film-like structure to form on the surface of the metal. These structures, referred to as oxide films, may result in an unsuccessful joining or "weld" of opposing streams of metal in the mold cavity, producing a region known as a confluence weld when the streams meet[l]. In previous works [4-5] the production of confluence welds amongst multipleingate configurations of varying geometries and their effect on the casting structure was studied. The focus of this work is the determination of mechanical properties for various bottom-filled gated systems (referred to as Al, A2, B, and BO). Experimental Procedures Two A356 10 mm-thick casting plates for each configuration were produced by the Electromagnetic Pump Green Sand (EPGS) process in the foundry Aleaciones Ligeras Aplicadas S.L. (Applied Light Alloys), in Valladolid, Spain. The EPGS process, which was developed by CIDAUT, employs the DISAMATIC to produce high speed vertically parted green sand molds, linked up with an electromagnetic pump for controlled dosing of the molten metal inside the
157
mold. A more quiescent filling of the casting is produced in this manner, thus minimizing the introduction of oxides into the melt and preventing defect formation during solidification in comparison with other casting processes [6]. The inlet is located on the side of the mold near the bottom. Cuesta, Martin, and Maroto [7] provide a thorough explanation of the development of the process and corresponding parameters. From the cast plates, information about mechanical properties was gathered to gain insight on the quality of the castings in relation to the different gating geometries (see Figure 1).
(a) (b) (c) (d) Figure I. The four gating configurations usedfor this research: (a) Configuration Al; (b) Configuration A2; (c) Configuration B; (d) Configuration BO. Configurations Al and A2 both have tapered bottom ingates while configuration B has tapered side ingates. For the BO configuration, the same geometry is used as that of the B configuration, however, only one of the ingates is utilized, whereas in the B design metal flow is permitted through both side ingates. These are the major geometrical differences amongst the molds. Each consists of a runner connected to side ducts which serve as filling channels for the Bconfiguration. At the top of all plate cavities feeders have been incorporated, in order to assist the feeding of the cavity during solidification. Four Point Bend Tests In order to characterize the flexural properties of simply supported beams, it is common practice to conduct three- or four-point bend tests. For this research, the method of four point bend testing was chosen over that of the three point, as the maximum load is evenly distributed between the top forces (referred to as the center span hereafter), whereas in the three-point test the maximum load is concentrated at the point of application of the top force [8-9]. With the even distribution of the load, a more accurate analysis of the cause of failure can be performed on the specimen, as fracture will generally not occur at the same location for each test. For this study, a pouring temperature of 730C was used. Sixteen bend test specimens were cut from each plate, each having dimensions of about 90 mm long, 20 mm wide, and 10 mm thick. The bend tests were conducted on an Instron 5869 testing apparatus with a default loading rate of .001in/s (,0254mm/s). Figure 2 shows the bend test specimen locations for all plate configurations. All specimens were tested in the as-cast condition.
158
(a) (b) Figure 2. Locations of the bend test specimens for each of the plates (not drawn to scale): (a) Entire plate with overall dimensions and reference to outer region excluded for edge effects; (b) Specimen area with dimensions and location of the 16 test specimens. For all samples, two ASTM Standards for testing of flexural properties of non-metallic materials were used (see Figure 3) [8-9], as there are currently no standards for the testing of metallic specimens.
Figure 3. Bend test configurations as described in the ASTM Standards [8-9] with central span length divided into sections of length L/3, where the load is represented as F. Weibull Analysis A Weibull statistical analysis was used in order to show the distribution of mechanical properties about mean values in relation to failure for each of the gating configurations. One form of the two parameter Weibull distribution is shown in the following Equation [10]: P = l-exp
f
σ σ„
(1)
where P is the probability of failure at the variable being measured (ultimate bending strength σ ), m is the Weibull modulus, and σ o is the scale parameter, which is a characteristic stress value at which approximately 63.2 percent of the specimens failed [10-11]. After taking the natural logarithm of both sides of Equation 1 twice, the linear regression equation is given as [11-12]:
159
F = l n lIn n
(
1
il= mlna-m\na -'JJ
1-
0
(2)
The Weibull modulus, m, is the slope of the regression line between Y and the natural logarithm of the ultimate bending strength (UBS), namely X for simplicity, and -mlnfa o) is the Y-intercept in the X-Y plane. The probability of failure, Pf„, for each value of UBS ranked in ascending order, can be evaluated using various methods [11-12]. The method chosen for this work is given by the equation:
n Where /' is the ranking of the result when UBS values are arranged in increasing order, and n is the total number of results. Results and Discussions As expected, test samples fractured within the central span. Most of the specimens tended to do so in the center of the span, confirming the soundness of the procedure.
Figure 4. Bend test of one of the specimens taken from an Al plate prior to complete fracture in the Instron 5869 load cell. Microscopy After bend tests were conducted, the cross sections of the samples were examined using Optical Microscopy. Eight samples were submitted to observation. Following the code [plate type_plate number-specimen number], these samples were: A l l - 3 , A12-12, A21-16, A 2 2 - 7 , B l - 5 , B2-10, B01-14, and B 0 2 - 1 . Samples were chosen from these locations in order to have a comparison of the microstructures throughout the specimen area from the regions closest to the feeder and to the bottom of the plate.
160
The surfaces of the samples tended to have similar dendritic features and apparent areas of porosity, however, preliminary inspection of the fine pore sizes would generally indicate a relatively quiescent filling compared with typical gravity systems [1]. Figure 5 illustrates characteristic microstructural images of the samples analyzed. Additional analysis will be performed on the samples in order to characterize mechanical properties in relation to porosity grain size and microstructure.
(c) (Φ Figure 5. Characteristic optical microscopy images ofspecimens: (a) A1J2-12; (bj A21-16; (c) B2-10; (d) B02-1 showing a largely free of porosity dendritic structure. Weibull Distributions Table I summarizes the results obtained from the four point bend tests. As is shown, values ranging from approximately 226MPa to 327MPa were obtained. The high magnitude of these figures for the as-cast A356 material used reinforces the idea that, in general terms, no major entrainment events (macroscopic bubbles) took place during the filling of the plates. Table I. Characteristic UBS Values (MPa) for each of the Gating Systems System Minimum Maximum Average Al 264.42 252.89 226.49 A2 253.44 269.04 285.81 296.69 B 259.14 327.43 323.84 286.22 BO 232.78
161
The results of the Weibull analysis (see Table II and Figure 6) allows us to make comparisons between the different gating configurations. As it can be seen, the Weibull moduli of the bottom gated plates (Al and A2) are widely higher, (and thus the span of the corresponding UBS values is widely smaller) than those filled laterally. This generally agrees with the specification given by Campbell of limiting the horizontal flow to a minimum extent in order to avoid the incorporation of the oxide film into the bulk melt [13]. Congruently, some indications of turbulent fold-in of the oxide upon the joining of the two metal streams inside the B plate (featuring relatively long horizontal flow paths) have been assessed by computer simulation [5]. In relation to this, it is also interesting to note that the maximum value of Weibull modulus was found for the plates for which the horizontal flow at the early stage of their filling is theoretically of least importance (A2) even when the fracture area of the test specimens lies most precisely in the region located in between the ingates, and therefore it is most likely to contain oxide films at the end of the casting process.
Configuration Al A2 B BO
Table II. Weibull Parameters for All Plate Configurations Weibull Distribution Weibull Modulus Scale Parameter Equation y = 35.46x- 196.72 35.46 256.83 y = 41.04x- 230.17 41.04 272.68 304.85 y = 19.56X- 111.85 19.56 y =19.25x-109.41 294.22 19.25
3
2 1 0 ? -1 Έ £
-J
-2
-3
-4
5.2
5,3
5,4
5,5
5,6
5,7
5,8
Ln (UBS)
Figure 6. Weibull plot of UBS distribution for all gating configurations.
162
5,9
6
Conclusions • • •
A series of A356 plates were produced by means of the EPGS process using several bottom and side filling systems prone to produce oxide entrainment into the bulk melt. The Weibull analysis of the ultimate bending strength of multiple test specimens preliminary indicates that oxide entrainment is basically related with the extent of horizontal flow at an early stage of the filling of the plates. Future work is planned in order to assess the effect of flow velocities, porosity, microstructure and grain size over mechanical properties. Acknowledgements
This work was funded by the National Science Foundation through Grant Number CTS-0553570 and supplemental funding by the International Research and Education in Engineering (IREE) program. The authors are thankful to Aleaciones Ligeras Aplicadas S.L. of Valladolid (Spain) for providing the castings used for this research. The assistance of Prof. E. William Jones with the Weibull analysis and Jacob Coleman and Hunter Cole with Optical Microscopy and porosity analysis is gratefully acknowledged. References 1
1. John Campbell, Castings 2" Edition - The New Metallurgy of Cast Metals fButterworthHeinemann, Oxford, UK, 2003). 2. M. Barkhudarov and C.W., Hirt, Paper presented at the Proc. Materials Solutions Conference on Aluminum Casting Technology, Illinois, October 1998. 3. J.T. Berry and R. Luck, "Porosity criteria functions revisited", Paper presented at the 2006 World Foundry Congress, Harrogate, United Kingdom, 4-7 June 2006. 4. R.L. Lett, S.D. Felicelli, R. Cuesta, J.T. Berry, A. Rivas, and M.E. Alcalde , "Confluence welds in aluminum castings - Part Two", AFS Transactions, 118 (2010), 29-38 5. R.L. Lett, S.D. Felicelli, R. Cuesta, J.T. Berry, and D. Losua, "Confluence welds in aluminum castings", AFS Transactions, 117 (2009), 131-138. 6. C. Nyahumwa, N.R. Green, and J. Campbell, "Effect of mold-filling turbulence on fatigue properties of cast aluminum alloys", AFS Transactions, 106 (1998), 215-223. 7. R. Cuesta, J. Martin, and J.A. Maroto, "New casting process: the EPGS process: the science and technology of casting production", Foundry Trade Journal, (March 2008), 66-70. 8. "Standard Test Method for Flexural Properties of Unreinforced and Reinforced Plastics and Electrical Insulating Materials by Four-Point Bending," ASTM International, 08.03, D627202(2008), 513-518. 9. "Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperature." ASTM International, 15.01, Cl 161-02c (2008), 217-226. 10. J. Mi, R.A. Harding, and J. Campbell, "Effects of the Entrained Surface Film on the Reliability of Castings," Metallurgical and Materials Transactions A, 35 (2004), 2893-2902. ll.W.D. Griffiths and N.W. Lai, "Double Oxide Film Defects in Cast Magnesium Alloy," Metallurgical and materials transactions A, 38 (2007), 190-196. 12. D. Wu, Y. Li, J. Zhang, L. Chang, D. Wu, Z. Fang, and Y. Shi, "Effects of the Number of Testing Specimens on the Weibull Parameters of Solid Catalysts", Chemical Engineering Science, 56 (2001), 7035-7044.
163
13. Campbell J. "The 10 Casting Rules Guidelines for the Reliable Production of Reliable Castings", First International Conference on Gating, Filling and Feeding of Aluminum Castings, October 1999.
164
Shape Casting: The 4"1 International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
GUIDELINES FOR 2-PARAMETER WEIBULL ANALYSIS FOR CASTINGS Murat Tiryakioglu School of Engineering University of North Florida Jacksonville, FL 32224 USA e-mail: m.tirvakiogluiS?.unf.edu David Hudak Department of Mathematics Robert Morris University Moon Township, PA 15108 USA e-mail: hiidakiS!rmu.edu Keywords: Weibull modulus, confidence interval, hypothesis testing Abstract 2-parameter Weibull statistics used in the analysis of mechanical data from castings are reviewed. Guidelines to estimate Weibull parameters by the linear regression technique are provided. Moreover goodness-of-fit tests for Weibull fits and calculating confidence intervals for the estimated Weibull modulus are discussed. A new hypothesis test for comparing two estimated Weibull moduli is introduced. The use of these guidelines is demonstrated by using data from the literature. Introduction Wallodi Weibull [1,2] introduced an empirical distribution based on the "weakest link", using the theory developed by Pierce [3], which has since been widely applied to the fracture-related mechanical properties of ceramics and metals. The cumulative probability function of the Weibull distribution is expressed as: — (1) σ0) where P is the probability of failure at a given stress (strain, fatigue life, etc.), σ, or lower. The term, σο, is the scale parameter, and m is the shape parameter, alternatively referred to as the Weibull modulus. Green and Campbell [4] showed that the tensile strength (ST) of A356 castings alloys follows a 2-parameter Weibull distribution and that the filling system design has a strong effect on the Weibull modulus. Hence the reliability of castings could be measured with the Weibull modulus. Since the results of Green and Campbell, the 2-parameter Weibull distribution has been used extensively in the casting literature to characterize fracture-related mechanical properties such as tensile strength (ST) [5], elongation and fatigue life (Nf) [6,7,8], According to Campbell [9], m is often between 1 and 10 for pressure die castings, and between 10 and 30 for many gravity-filled castings. For good quality aerospace castings, m is expected to be between 50 and 100. Hence Weibull modulus has been used as a measure of the casting quality.
165
Recently, Tiryakioglu and Campbell [10] provided guidelines for interpreting Weibull probability plots including the 3-parameter Weibull distribution and Weibull mixtures. They stated that the 2-parameter Weibull distribution is applicable only when castings have defects with large sizes that come from the same, single distribution, i.e., one source of damage during melt preparation and mold filling. The present study is intended to provide additional guidance when the 2-parameter analysis is warranted and provides a step-by-step procedure for Weibull analysis with the linear regression technique. Procedure for 2-Parameter Weibull Analysis Step 1. Assign probability to each data point: There are a number of probability estimators (alternatively referred to as plotting position formulae) in the literature. These formulae can all be written in the form (2)
n+b
where ; is the rank in ascending order, n is the sample size and a and b are empirical constants. Montecarlo simulation studies [11,12,13,14] showed that all probability estimators produce biased estimates of the Weibull modulus, i.e., the average of estimated Weibull moduli is different from the true Weibull modulus used in these studies. The magnitude of the bias depends on the values of a and b as well as the sample size. The authors [15] determined the values of a and b that yield unbiased estimates of σο and m, which are listed in Table 1 for sample sizes (n) between 5 and 100. It is strongly recommended that the values of a and b listed in Table 1 be used in lieu of other plotting position formulas in the literature. Table 1. The values of a and b for Equation 2 that yield unbiased probability estimators for the two Weibull parameters. n a b n a b 5 6 7 8 9 10 11 12 13 14 15 17 20 22
0.173
0.5
0.243
0.39
25
0.28
0.31
0.309
0.251
0.322
0.21
0.348
0.19
0.367
0.16
0.371
0.13
0.382
0.11
0.388
0.1
0.394
0.08
0.407
0.05
0.417
0.03
0.43
0
27 30 32 35 40 45 50 55 60 65 70 75 80 90
0.443
0
100
0.448
0.518
0 0 0 0 0 0 0 0 0 0 0 0 0 0
0.519
0
0.455 0.46 0.465 0.472 0.481 0.486 0.499 0.503 0.509 0.518 0.522 0.516
166
Step 2. Obtain linear regression fit to 1η(σ) versus ln(-lnd-P)) One of the most commonly used methods of presenting the Weibull fits to data is the Weibull probability plot. After rearranging, Equation 1 can be written as 1η[-1η(ΐ-Ρ)] = »ι1η(σ)-»ι1η(σ 0 )
(3)
Note that Equation 3 has a linear form when the left-hand side of the Equation is plotted versus 1η(σ) with a slope of m and an intercept of-m 1η(σο). Hence the best fit line to this probability plot represents the linear regression method for estimating the two Weibull parameters. Step 3. Conduct Goodness-of-Fit test: Usually the trend of the data on the Weibull probability plot is used to judge whether the data indeed come from a Weibull distribution. The use of a probability plot, however, is subjective and insufficient and therefore it is strongly recommended that probability plots be always augmented by other goodness-of-fit tests [16]. The coefficient of determination, R2, has been commonly used as a measure of the goodness-offit, especially in the metal casting literature. The higher the value of R2, the higher the confidence that data follow the distribution being tested. However clear guidelines for using R2 as a goodness-of-fit test have been recently developed by Tiryakioglu et al. [17] who ran Monte Carlo simulations to determine the critical points of R2 at a=0.05 (R2o 05) and reported that the following formula can be used for. sample sizes between 5 and 100: R6.05 = 1 . 0 6 3 7 — ^ 3 -
n°
(4)
The hypothesis that the dataset follows the tested distribution is rejected when p-value for calculated R2 is less than a specified value for Type I error (a), which is typically prescribed as 0.05. If the calculated R2 is higher than R2oos, then it can be concluded that the data indeed come from a 2-parameter Weibull distribution. The authors recommend that Equation 4 be used to evaluate goodness of fit to the Weibull distribution with R2 instead of other general guidelines provided in literature. It should be noted that the 2-parameter Weibull analysis is not valid if R2 < R2o.05, in which case steps 5 and 6 should not be taken. Step 4. Calculate Confidence Interval for the estimated Weibull Parameters It is important to realize that estimated Weibull parameters have their own statistical distributions. Consequently, it is necessary to calculate confidence intervals for the two Weibull parameters, especially m. The distribution of the estimated Weibull modulus was found [14] not to follow any formal distribution. Therefore the use of percentage points is necessary to calculate the confidence intervals. Percentage points for the distribution of estimated m when unbiased probability estimators in Table 1 are used are presented in Table 2 for sample sizes between 5 and 100. For instance if the sample size is 30, then two-tailed 95% confidence intervals are calculated by first finding the percentage points of 0.025 and 0.975 for n=30 (95% will be
167
included between percentage points of 0.025 and 0.975). In Table 2, these points are 0.712 and 1.499, respectively. Finally, the 95% confidence limits are found by multiplying the estimated Weibull modulus by 0.712 and 1.499. Table 2. Percentage points of the distribution of estimated m (after normalization) obtained by using the unbiased probability estimators in Table 1. n
0.005
0.01
0.025
0.05
0.1
0.9
0.95
0.975
0.99
0.995
S
0.274
0.316
0.407
0.497
0.613
1.984
2.304
2.625
3.077
3.425
6
0.328
0.371
0.457
0.540
0.643
1.821
2.105
2.392
2.793
3.125
7
0.363
0.406
0.492
0.572
0.666
1.727
1.996
2.247
2.558
2.809
8
0.388
0.436
0.513
0.587
0.676
1.661
1.887
2.128
2.445
2.674
9
0.422
0.470
0.545
0.606
0.693
1.618
1.835
2.049
2.342
2.551
10
0.435
0.488
0.558
0.623
0.705
1.567
1.776
1.984
2.237
2.475
11
0.456
0.499
0.573
0.635
0.715
1.534
1.718
1.890
2.151
2.331
12
0.490
0.529
0.592
0.648
0.725
1.511
1.686
1.855
2.066
2.294
13
0.505
0.544
0.608
0.662
0.734
1.477
1.639
1.818
2.049
2.222
14
0.518
0.558
0.613
0.670
0.741
1.456
1.618
1.786
1.988
2.179
15
0.526
0.565
0.624
0.683
0.750
1.439
1.600
1.761
1.988
2.165
17
0.538
0.576
0.635
0.690
0.755
1.412
1.555
1.684
1.894
2.037
20
0.568
0.608
0.661
0.708
0.770
1.374
1.506
1.637
1.795
1.912
22
0.583
0.621
0.676
0.723
0.780
1.340
1.464
1.575
1.739
1.866
25
0.609
0.638
0.691
0.734
0.790
1.326
1.435
1.543
1.675
1.770
27
0.620
0.650
0.699
0.746
0.797
1.312
1.425
1.534
1.664
1.789
30
0.639
0.670
0.712
0.753
0.806
1.294
1.397
1.499
1.637
1.736
32
0.651
0.676
0.723
0.763
0.810
1.280
1.372
1.462
1.580
1.689
35
0.657
0.688
0.730
0.769
0.816
1.266
1.355
1.449
1.550
1.634
40
0.673
0.702
0.745
0.784
0.829
1.253
1.335
1.412
1.511
1.582
45
0.694
0.720
0.758
0.794
0.836
1.230
1.309
1.383
1.493
1.580
50
0.704
0.728
0.767
0.800
0.840
1.220
1.294
1.362
1.462
1.529
55
0.711
0.738
0.775
0.807
0.845
1.205
1.274
1.337
1.429
1.493
60
0.725
0.751
0.783
0.814
0.852
1.200
1.266
1.335
1.410
1.479
65
0.735
0.757
0.790
0.821
0.856
1.189
1.253
1.312
1.389
1.447
70
0.742
0.762
0.794
0.826
0.863
1.183
1.244
1.300
1.368
1.422
75
0.742
0.767
0.799
0.829
0.864
1.176
1.235
1.289
1.361
1.420
80
0.754
0.776
0.808
0.835
0.870
1.170
1.224
1.282
1.351
1.403
90
0.767
0.786
0.818
0.845
0.876
1.160
1.214
1.266
1.325
1.383
100
0.777
0.800
0.826
0.853
0.883
1.149
1.199
1.242
1.300
1.348
In contrast to the distribution of the modulus m, the distribution of estimated scale parameter is normal [15]. There is no need therefore to use percentage points tables. The standard deviation of the estimated scale parameter (after normalization), Soo, calculated by using the probability estimators listed in Table 1 is found by
168
\
=_
0.359
_
r~
(5)
Therefore confidence intervals for the true scale parameter ( σ0,ιηιί ) can be found by 0 359
1.000 + Z, _, ^ 4 ^ · ^ n
σ 0|<me
V«
Taking o=0.05 (95% confidence), Z
'Ά
Q 35Q
οΐ,™, ^ 1.000 + Z/ 2a / ττττκ ^X- (6) V"
, and Z
7
Δ
are 1.96 and -1.96, respectively.
Step 5. Comparing two Weibull Moduli Because Weibull modulus has been used as a measure of reliability of castings, a formal procedure is necessary to compare the Weibull moduli from two sets of castings. Such a hypothesis test, to the authors' knowledge, is not available in the literature. The authors [18] have recently provided the results of their Monte Carlo simulations for comparing two Weibull moduli for sample sizes between 10 and 100. The results are provided in Table 3, which shows the 2.5 and 97.5 percentile of the distribution of mi/ni2 where ni>n2. The values in Table 3 can be used to test the hypothesis that the two Weibull moduli are equal at a=0.05. Table 3. The 2.5 and 97.5 percentiles for the distribution of the ratio of two estimated Weibull moduli.
Application to Datasets Three datasets were considered from the study by Green and Campbell [19] who showed that the tensile strength (ST) of cast Al-7%Si-Mg alloys is affected to a great extent during the mold filling stage. If the mold is filled quiescently, tensile strength is not only higher but also has less variability, i.e., higher reliability. Conversely, tensile strength has a lower average and higher variability when the mold filling occurs turbulently. The datasets represent these two types of mold filling: top-filled (TF) which is quite turbulent, and bottom-filled (BF), which is quiescent.
169
Also included in the analysis are the castings which were modified by Sr followed by quiescent filling (BFmod). The sample size is 45, 36 and 37 for TF, BF and BFmod, respectively. For the plotting position formula (Equation 2), a has a value of 0.481, 0.466 and 0.468 for TF, BF and BFmod, respectively, and b=0. The Weibull probability plot for the three datasets is presented in Figure 1 and the estimated parameters as well as goodness of fit measures are given in Table 4. For TF, R2 < R2o 05, indicating that the 2-parameter Weibull fit has to be rejected. Note in Figure 1 that the slope for the lowest five points seems to be less than that for the rest of the data. This change in slope is indicative of the presence of multiple defect distributions and therefore the data may come from a mixture of two Weibull distributions [10,20], For BF and BFmod, R2 values are in excess of R2oo5, indicating that the Weibull fit is acceptable. The 95% confidence limits for the two parameters were established for BF and BFmod castings, as presented in Table 4. Note that the values to establish the confidence limits for m for n=36 and 37 are not listed in Table 2 and therefore they were found by interpolation.
ln(ST(MPa)) Figure 1. Weibull probability plot for the tensile strength of Al-7%Si-Mg alloy castings filled turbulently (TF) and quiescently (BF), as reported by Green and Campbell [4].
170
Table 4. Statistical results on the Weibull fits to the three datasets.
If one wishes to determine whether the reliability of BF and BFmod castings is different, i.e., whether Sr modification has an effect on reliability, the estimated Weibull moduli of the datasets can be compared. The ratio of the estimated Weibull moduli is 1.321 (50.71/38.40). Note that the Weibull modulus of BFmod is written in the numerator because the sample size of that dataset is larger than that of BF. This ratio of 1.321 falls within the values listed in Table 3 for sample sizes of 30 and 40. Hence we do not have enough evidence at this confidence level to support the hypothesis (at a=0.05) that Sr modification increases reliability. Conclusions •
A step-by-step procedure was introduced for the 2-parameter Weibull analysis of casting data by the linear regression method, in cases where such analysis is warranted.
•
Goodness-of-fit tests should be conducted on the Weibull fits to determine whether the data do come from a 2-parameter Weibull distribution.
•
If the distribution is indeed Weibull, confidence limits on the two parameters should be established.
•
A new hypothesis test for the comparison of estimated Weibull moduli has been introduced.
References 1. Weibull, W. (1939). A statistical theory of the strength of materials. Proc. The RoyalSwedish Institute for Engineering Research. Nr. 151. 2. W. Weibull, The phenomenon of rupture in solids, Royal Swedish Institute of Engineering Research (Ingenioersvetenskaps Akad. Handl.), Stockholm, Vol. 153, 1-55, 1939. 3. F. T. Pierce: J. Textile Inst. 1926 vol. 17, pp. T355-T368. 4. N. R. Green, J. Campbell: Mater. Sei. Eng. A, 1993, vol. A137, pp. 261-266. 5. G. E. Byczynski, J. Campbell: In Advances in Aluminum Casting Technology II. Edited by M. Tiryakioglu and J. Campbell; ASM International, 2002, pp. 65-74.
171
6. C. Nyahuniwa, N. R. Green, J. Campbell: Metall. Mater. Trans. A, 2001, vol. 32, pp. 349358. 7. D. Casellas, R. Pérez, J. M. Prado: Mater. Sei. Eng. A, 2005, vol. A398, pp. 171-179. 8. Q. G. Wang, D. Apelian, D. A. Lados: J. Light Metals, 2001, vol. 1, pp. 73-84. 9. J. Campbell: "Castings", 2nd Edition, p. 303, London, Elsevier, 2003. 10. M. Tiryakioglu, J. Campbell: Metall. Mater. Trans A., in press. H.A. Khalili, K. Kromp, J. Mater. Sci. 26 (1991) 6741. 12. R. Langlois, J. Mater. Sci. Lett., 10 (1991) 1049. 13. K. Trustrum, A. de S. Jayatilaka, J. Mater. Sci. 14 (1979) 1080. 14. M. Tiryakioglu, D. Hudak: J. Mater. Sci. (2007)42:10173-10179 15. M. Tiryakioglu, D. Hudak: J. Mater. Sci. (2008)43:1914-1919 16. S.S. Shapiro, C.W. Brain, in "Statistical Distributions in Scientific Work", v.5, eds. C. Taillie, G.P. Patii, B.A. Baldessari, D. Reidel Publishing, 1981, p.l. 17. M. Tiryakioglu, D. Hudak, G. Ökten: Mater. Sei. Eng. A, 2009, vol. 527, pp. 397-399. 18. D. Hudak, M. Tiryakioglu, submitted to Mater. Sei. Eng. A. 19. N.R. Green, J. Campbell: AFS Trans. 102 (1994) 341. 20. C.A. Johnson: J. Frac. Mech. Ceramics 5 (1983) 365.
172
Shape Casting: The 4th International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals h International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
Advanced Methoding Concepts for the Gravity Casting of Steel Alloys Bob Puhakka Alloy Casting Industries, New Hamburg, Ontario Canada Keywords: steel castings, casting defects, casting methoding, bifilms Abstract Wlodawer and Chvorinov provided invaluable early guidelines for the feeding of steel castings, but technology has moved on. This paper describes a new set of concepts that apply to the filling conditions for the entire family of steel casting alloys - from plain carbon to super duplex stainless steels. Application of the concepts is achieving new standards of quality together with reduced costs. The Current State of the Art The casting industry (especially the steel castings industry) has been fixated on fulfilling the known rules for castings derived from the work of Wlodawer and Chvorinov. These rules are, of course, useful since they relate to the feeding of the solidification shrinkage. However, the industry has completely neglected the problems of filling to avoid reoxidation and surface turbulence. As a consequence our current methods cannot produce castings reliably with any sort of reproducibility. This serious problem has unfortunately led to pervasive low expectations for steel gravity sand castings. All previous work on this matter has been relatively ineffectual. A fresh conceptual break from tradition is required. In an effort to better understand the inclusion presence in steel castings a study was carried out inspecting castings from several casting facilities. The nature of the inclusion content was tabulated and presented. [4] The results indicated that in excess of 83% of the defects present in steel castings were reoxidation inclusions. Reoxidation inclusions result when molten steel and air combine. They are solid particles and agglomerates of oxides such as AI2O3, MnO, CaO, T1O2 and FeO which mainly initiate as films on the surface of the liquid metal. Castings that possess a large quantity of reoxidation inclusions will require costly rework to make saleable; and even then they may never be fully reliable. Such castings are prime candidates for leaks, reduced mechanical properties, poor machining and poor weldability. A recently published article on the nature of casting defects states, "Inclusions are generally associated with the flow of liquid metal into the mold during pouring. However, modeling and verification trials in foundries have failed to indicate how gating systems may be universally improved." [5] In fact, the conclusions of almost every research paper on the topic of castings defects in steel have come to nearly identical conclusions. [6] [7] The fill system design logic currently employed in the industry is relatively arbitrary and unscientific. In fact it is common for the filling system to be assembled from pre-formed refractory sleeves as a result of the widespread belief that sand molds will not withstand contact with liquid steel. The parallel channels dictated by the use of this system ensure a maximum of air entrainment, with consequential problems for the casting. This paper will explain in some detail how it is possible to design a fill system that prohibits the formation of reoxidation
241
inclusions. Before we begin, however, it is important to review the current industry fill system design features in detail. The Pouring Cup
F,gure 1
An illustration of the aspiration mechanism of a conical pouring cup
, .
e
. ,
Tnfir.hanwin or a conical nonnno cnn
The traditional conical pouring cup is perhaps the worst single feature of traditional filling systems. The 'design' is for the sole purpose of creating an enlarged, targeted entrance to the downsprue. However, a detailed review of the system reveals that it is in fact an efficient aspiration pump, taking down at least as much air as metal into the running system during the pour[12]. The quantity of entrained air explains the so-called 'Discharge Coefficient' value of only about 0.5. A reasonable input for the system should achieve a value 1.0 otherwise defects are certain to be created. Figure 1 illustrates the aspiration of air that takes places within the conical basin. Usage hints such as "pour to the back of the cup to avoid vortex generation" is irrelevant. Such , . . .,
,■
·
,
, ■
.,
operational details are a diversion, overlooking the major r
damage caused by this feature.
'
σ
J
The Non-Tapered /Reverse Tapered Sprue The taper of the sprue is not only to allow for drawing the sprue pattern from the mold! Parallel refractory runner sleeves or reverse-tapered sprues are widely used despite the well known fact that gravitational acceleration narrows the falling stream predictably, predicting necessarily that parallel and reverse tapered sprues will not constrain the melt, and will therefore allow surface turbulence to create oxide damage. Clearly, the sprue taper must be accurately calculated and moulded for each and every casting. The Sprue-Runner Transition Studies have shown [2] that it is not possible to have an ideal flow behavior when transitioning from a round downsprue into a square or rectangular runner. Thus the continuity of form of the flow channels is critical to success. The Sprue Well
Figure 2- the sprue weil
essentially serves as a meial-air bl
,r
Traditionalists view the well as a 'cushion' for the melt at the sprue exit. Its widespread use appears to have become dogma, somehow supposed to reduce flow velocity and promote laminar flow. Unfortunately however, the well once again removes the constraint on the flowing stream, allowing churning, mixing and combining air with the molten metal. There are many formulae used to calculate sprue well size for a particular gating system but recent research has emphatically confirmed that the ideal volume for a sprue well, contrary to current practice, is actually zero. It is known ^, ^.,
>,
·.
L
c
t
,« , , ι ^
that the well provides a stream of entrained bubbles for some
242
seconds, whereas a simple radiussed turn can redirect the flow without the generation of a single bubble [2]. The Runner Choke
Figure 3 - the choke produces severe localized turbulence resulting in the entrainment of surface films and air bubbles
An industry publication recently sent out a newsletter with the following advice, "Practical Tip: Increase Choke Size When Having Issues with Flow Rates"[8]. The advice overlooks the fact that the provision of any choke accelerates the melt, causing it to jet, resulting in the generation of copious quantities of reoxidation inclusions. Figure 5 shows a screen grab from a fill simulation of a 'textbook' AFS 1:4:4 ratio system with a runner choke. The velocity vector behavior immediately following the choke indicates the location where volumes of air bubbles and reoxidation inclusions will be generated during the pouring of the casting. [10]
The Runner Expansion Methods The practice of reducing the velocity of the advancing fluid front by increasing the cross-sectional area of the runner over the sprue exit area can have disastrous consequences. It has been demonstrated by X-Ray radiographie video for real metal, and many times in computer simulations, that the advancing melt will not expand to fill an enlarged channel. An unfilled runner is subsequently responsible for the entrainment of air, oxide films and mold material - depositing much of this into the casting cavity. There is no 'expansion ratio' capable of producing premium results. The use of traditional expansion ratios must be avoided.
Figure 4 - runner expansion techniques result in unfilled runners producing severe surface turbulence, surface film and air entrainment
The Runner Extension
Figure 5 — the runner extension provides a location of unrestrained flow resulting in the aspiration of air into the casting cavity.
During a gating course early in my career I was told, "Thou shalt not allow the first metal poured to enter the casting cavity." This advice, of course, was an emphasis for the need to include a runner extension into the fill system. At the time this was probably good advice because the traditional filling systems created large quantities of damaged melt. The extension was intended to collect the debris and oxides generated in the running system by the first-
243
arriving metal. In fact, in the naturally pressurised systems described in this report this is no longer the case since the melt arrives at the gates in relatively good condition. What fraction of metal does initially bypass the ingates is nowadays potentially counter-productive. It can often be seen to become part of a rolling-back wave, rolling in air and oxides that will find their way into the casting cavity via the last ingate. [9] [11] The New Methodology To avoid oxides it seemed rational to adopt a system that would assist with the elimination of air from the flowing stream. For this reason, the naturally pressurized filling system design was adopted [2]. This is a system in which the areas of the filling channels are calculated by finding the velocity, V, at each fall distance, h, from the melt level in the pouring basin, assuming no friction. The approach is therefore a simple balance between potential energy, mgh, and kinetic energy, mV2/2. The approach is well known to casting method engineers.
Figure 6-an ideally designed spruedual runner transition. No well, no choke, no runner expansion
The difference in this situation is to accept these areas and provide the filling system with only these calculated areas at every point throughout the downsprue and runners. Only the gates would be increased in size to reduce the velocity of entry to the mold to the critical 0.5 m/s if possible (on occasions this would be raised to 1.0 m/s if necessary, but not beyond this already 'stretched' limit to the Rule). A typical 'sprue exit/runner/gate' ratio for such a system might vary from 1:1:4 to 1:1:20. It must be stated however that the preselection of such a 'ratio' has no part to play in the design of a proper fill system; the ratios simply happen, occurring as a result of the design process.
It is worth repeating that the standard technique of increasing the area of the runner to reduce the velocity of the flow, using for instance a ratio 1:2:4 or other expanding ratios, does not work. It merely provides an unfilled runner in which turbulence can be generated to damage the flow [2], The system requires a specially designed pouring basin of a deltashaped, offset stepped type [2], to reduce so far as possible the ingress of air into the entrance to the filling system. The benefits of such a design were recently excellently illustrated with a video of a water model by workers at CANMET [3] [13]. (Campbell [2] describes a contact pour technique using bottom-teemed ladles to exclude air at the sprue entrance but our lip pour techniques suited to our size of casting did not require this particular solution). The sprue requires a correctly calculated taper. Although Campbell recommends a curved hyperbolic taper [2] it was judged that our castings were not sufficiently tall to benefit greatly from such sophistication. Thus so far we have calculated only the sprue entrance and the sprue exit and connected the two with a straight taper It clearly works well with the limited size of casting we pour (encompassing the range 1 to 2,500 kg).
244
Figure 7 - delta-shaped offset step pouring basin
The other major feature of this approach is that only bottom gating into the mold cavity can be permitted; otherwise the melt falls inside the mold cavity, exceeding the critical fall distance of a few millimeters in which gravity accelerates it to above its critical velocity, so that it starts to jump and splash, creating surface turbulence and entraining defects such as bubbles and bifilms (and sand inclusions, which are an excellent indication of a turbulent filling system - not an indication of poor molding sand). Thus gating at the mold joint (for traditional horizontally parted molds) has become a feature of the past - a luxury that can be no longer risked or afforded. Along with blind risers [14], conical pouring basins and runner chokes; parting line gating serves as a corner-cutting, "cost-saving" practice that produces damaged castings A final discipline is the modeling of every casting to check that the methoding is complete and effective. Thus the action of the filling system is carefully studied to ensure that no pockets of air remain after the first pass of liquid. Such pockets allow turbulence and the entrainment of air and oxides. The system needs to be seen to fill and to pressurize gently against the mold walls. This is the natural pressurization concept, arising from the friction generated by the flow against the channel walls. The channels clearly need to stay full during the filling process. Having filled successfully with only metal (no air in the form of bubbles), the feeding system is then checked to ensure that there is no danger of shrinkage. For some castings this exercise is
vigun
8
- the use of process simulation is absolutely
essential. The unique solidification profile of each and y individual geometry must be evaluated and
ever
not always straightforward. The provision of manipulated. a bottom gate and high level feeders (top if possible) sometimes requires turning the casting through 180 degrees before a workable solution can be found. Occasionally, much use of heavy chills is required to generate favorable temperature gradients. It is at this step that one notices the vast difference in approach compared to the traditional methoding practices. The traditional approach selected the casting orientation based on the 'easiest' way to feed the casting, with the fill system being only an after-thought. The new approach begins with assessing how to fill the mold as perfectly as possible; and then finds a way to meet the feeding requirements. The fill system takes precedence. Only when both filling of the mold cavity is seen to be tranquil, and the action of the feeders is seen to be adequate to all locations of the casting, is the tooling built, the mold made and finally poured. It has become interesting to note that the pouring of the mold is no longer viewed as being in the lap of the gods, but is now viewed as an expected confirmation of the methoding technique. That is not to say that failures have not been experienced. These have occasionally occurred as a result, for instance, of forgetfulness or oversight of some key aspect of the design prior to pouring. Such mistakes illustrate the fallibility of a regime in which single person is totally
245
responsible, but at the same time working under pressure in a production environment. Clearly, like all professional engineering designs, at least one other competent person should be available to check and sign off the design prior to manufacture. The naturally pressurized fill system is now used exclusively at the author's facility. Castings are produced from the entire family of ferrous engineering alloys in sizes ranging from 1 to 2500 kg. The naturally pressurised concept is applied without exception, and found to apply successfully across the complete range of casting sizes. On Filtration The use of reticulated ceramic foam filters is very common in the steel foundry. Almost universally, the designs use the filter as a damage mitigation device even though it is found to be largely unsuccessful if simply introduced into a fill system designed with traditional features. The filter does not right several wrongs! For instance, an unconstrained front-end with a conical pouring basin and parallel sprue will generate reoxidation inclusions that will plug the filter. In addition, an unconstrained, ratio-based runner design will generate reoxidation inclusions postfilter that will now be deposited into the casting cavity[15]. In contrast, the introduction of a ceramic foam filter into a naturally pressurized fill system as a velocity reduction device yields surprising additional benefits. Without the generation of reoxidation inclusions the filtration capacity of the filter now appears to be essentially infinite. This brings significant savings, as systems can be designed with fewer and smaller filters. In fact, recent experience confirms the new fill system designs allows up to ten times more metal through the filter than the manufacturer's recommended filtration capacity; there has been no experience of a filter blocking prematurely. On Fill Rate There continues to be great debate on the determination of target fill rate for the pouring of steel castings. Traditional formulae include considerations such as fluid life and cooling rate etc. There is also the questionable approach of simply taking the square root of the casting weight (in lbs) and using that value as the target pour time (in seconds). All such techniques seem to be largely irrelevant. Furthermore, it is adherence to these largely arbitrary calculation sets that appears to have prolonged the frustrating attempts toward defect elimination for decades. A fill system must not be designed that exceeds the volumetric output of the ladle. If the system is too large to be kept full by the ladle, dramatic and severe aspiration necessarily takes place damaging the casting. A fill system needs to be designed that pours continuously, as quickly as possible, and fills the casting cavity as slowly as possible- all at the same time. 'As quickly as possible' is very simply the maximum volumetric output of your ladle. 'As slowly as possible' is simply maintaining a maximum limit ingate velocity of approximately 0.5 m/s. An arithmetic manipulation of these two competing design targets is the fill system design process. The challenge for the design engineer includes: 1. Know the maximum volumetric output for each and every ladle used in the facility. An intelligent designer will design a fill system customized for the very ladle to be used to pour the casting.
246
2. An empirically determined correction factor for friction and velocity losses. Specific to the particular molding materials, sprue/ runner geometry and number of fill system branches - the design engineer must know this value. A direct comparison of actual fill times vs. ideally calculated (excluding loss compensation) will provide a very specific correction factor to be used. As afirstapproximation africtionalcorrection to thefillrate a value of 1.3 (corresponding to a 30% loss of speed) is a useful starting point. 3. A specifically designed pouring basin for the ladle to be used. The pouring basin is the critical step in the transfer of the molten metalfromthe ladle to the sprue entrance - the basin must be designed for the ladle. It is not unreasonable, then, to imagine that a casting facility will have a basin design for each and every ladle used. Confirmation by Results Steel castings are notorious for their need for excessive dressing, rework and cosmetic repair. The majority of the imperfections visible on the surface are clearly reoxidation inclusions [4] resultingfromthe entrainment of air during thefillingof the mold. In addition, the surface finish is often poor. The carbon steel (ASTM A216 WCB) pump casing shown in Figure 9 is an example of the new approach. It has been shaken out and put through the coarse shot blaster to remove the remaining adhering mold material. This is a casting that is required to be fully pressure tight and therefore subjected to full magnetic particle inspection (MPI), liquid penetrant inspection (LPI) and radiographie testing (RT) to insure this condition. The casting is seen to be clean and devoid of surface inclusions, shrinkage and tearing. In addition, it is perfectly leak-tight.
Figure 9 - a carbon steel pump casing immediately following shakeout and shot blasting. There are no inclusions and the surface finish is excellent.
Conclusions 1. The majority of casting defects (including much gas porosity, perceived microporosity, hot tears, leakers and reoxidation inclusion) are the result of defects (mainly double oxide films and/or air bubbles) entrained by thefillingprocess. The problems arisefromthe use of traditional filling systems (including conical pouring basins, ill-tapered sprue using refractory sleeved running systems, choked runners, parting line gated molds etc.). 2. A naturally pressurizedfillingsystem design can practically eliminate entrained defects, giving essentially perfect castings. Castings are characterized by good surface finish, freedomfromleaks, surface inclusions, porosity, hot tears and cracks. Absence of
247
upgrading work, dressing and welding, has brought significant economies. Additionally, the absence of rework avoids the unintentional masking of initially unseen defects that may contribute to subsequent service failures. 3. The author has validated the success of these techniques on a broad range of alloys in the steel, white/grey/ductile iron, brass, bronze, Ni-base superalloys and high-conductivity (pure Copper) alloy families. It seems reasonable to suppose that the casting manufacturing systems presented here would apply to metals and alloy of all types. Acknowledgements I am grateful to John Campbell, author of 'Castings' for his continued advice and encouragement. John has become a close, personal colleague without whom I would never have been able to accomplish the conversion as quickly and completely. Our near-daily correspondence allows for me to dialogue with another individual in the same headspace regarding the nature of casting defects. Sincere thanks are also due to my co-workers at Alloy Casting Industries in New Hamburg whose dedication and hard work made our revolution possible. References 1. 2. 3. 4. 5.
J Campbell "Castings" 2003 Elsevier. J Campbell "Casting Practice" 2004 Elsevier S Kuyucak; 68th World Foundry Congress, Chennai, India 2008 (February) pp 483-48 J M Svoboda, Monroe R W, Bates C E, Griffin J; Trans Amer Foundry Soc 1987 95 187-202. Predicting the Occurrence and Effects of Defects in Castings; Malcolm Blair, Raymond Monroe, Christoph Beckermann, Richard Hardin, Kent Carlson, and Charles Monroe, JOM 2005 6. J.A. Griffin and CE. Bates, Ladle Treating, Pouring and Gating for the Production of Clean Steel Castings, SFSA Research Report No. 104 (Crystal Lake, I: Steel Founders' Society of America, 1991) 7. P. Scarcer, Jr., CE. Bates, and J.A. Griffin, "Using Gating Design to Minimize and Localize Reoxidation" (Paper presented at the 56th Steel Founders' Society of America National Technical & Operating Conference, Chicago, Illinois, 7-9 November 2002) 8. AFS Education Connections Newsletter, April 20, 2010 9. http://bobpuhakka.blogspot. com/2010/06/not-to-beat-dead-horsebut.html 10. http://bobpuhakka.blogspot.com/2010/04/sacred-cows-of-metal-casting-episode-2.html 11. http://bobpuhakka.blogspot.com/2010/01/left-to-end-of-runner.html 12. http://www.youtube.eom/user/bobpuhakka#p/u/4/i5vWOrQUVLI 13. http://www.youtube.eom/user/bobpuhakka#p/u/57/5Zkh4wrnsAk 14. http://www.youtube.eom/user/bobpuhakka#p/u/46/uwS3-l_rUlg 15. http://www.youtube.eom/user/bobpuhakka#p/u/44/JxnNKiCz8Jk
248
Shape Casting: The 4"1 International Symposium Edited by: Mura! Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
ADVANCED CASTING MOLD DESIGN TECHNOLOGY OF THE LCS WATERJET INLET TUNNEL ENTRY EDGE COMPONENTS LaurentiuNastac1 and John Romanelli2 'Concurrent Technologies Corporation, Pittsburgh, PA 15219, email:
[email protected] Concurrent Technologies Corporation, Johnstown, PA 15904, email:
[email protected] Approved for public release; distribution is unlimited. This material is submitted with the understanding that right of reproduction for government purposes is reserved for the Office of Naval Research, Arlington, Virginia 22203-1995. Keywords: LCS waterjet entry edge components, casting modeling, mold design, sand mold printing, ASTM A757 C1Q steel, prediction of macro-shrinkage, porosities, hot tears, and cracks Abstract In order to reduce cost, increase performance and ensure quality, this Navy Metalworking Center (NMC) project utilized an advanced casting simulation-based optimization approach to assist in the improvement of the mold design of LCS Waterjet Inlet Tunnel (WjIT) entry edge components. This approach helped to minimize mold filling and solidification-related defects (misruns, coldshuts, shrinkage and porosity and hot tears), as well as post-solidification-related defects (hot and cold cracks, distortion and residual stresses). The results of this optimization were used to more readily achieve first-time quality on the geometrically challenging WjIT components. The melt chemical composition and the thermo-physical and mechanical properties of ASTM A 757 C1Q steel material as a function of temperature were used in the simulation study. To accomplish this work, NMC used Nova Flow&Solid™ and NovaStress™ software. Introduction The water-jet inlet tunnel entry edge (WjIT) for the Lockheed Martin Freedom class Littoral Combat Ship (LCS) was previously manufactured from 13 ASTM A131-DH36 25-mm thick plate segments by welding and grinding. The hydrodynamic profile of the WjIT entry edge is a critical operational factor (see Figure 1). The objective of this Navy Metalworking Center (NMC) manufacturing technology (ManTech) project was to validate that a casting solution is indeed the most promising alternative manufacturing method for WjIT [1], The ManTech project also sought to demonstrate that a cast steel (ASTM A 757 C1Q) provides comparable tensile and impact properties to ASTM A131-DH36 rolled plate steel. There are four WjITs entry edges on the LCS and each is slightly different. The total length of each WjIT entry edge is about 4m and its width is about 1.5 m. The thickness of each WjIT is 25 mm. The casting solid models developed in this project were divided into three segments due to the size, casting, heat treating and handling issues. The segments were designated as Part A, B and C. These segments will be assembled by welding. This paper describes the process modeling approach used in the development of the mold design technology used to produce the WjIT parts.
Figure 1. WjIT geometry (Part B).
249
Objectives and LCS WjIT Casting Requirements The objectives of this work were as follows: (1) apply modern mold design technologies to assist in the manufacturing of the LCS entry edge casting components using: (i) ProMetal Rapid Casting Technology (RCT) [2] and (ii) NovaCast software for mold filling, solidification and stress modeling [3], and (II) achieve first-time quality on the geometrically challenging WjIT components made of ASTM A 757 C1Q steel material with the following requirements: (i) nondestructive testing (NDT) inspection (magnet particle, die penetrant, and radiography inspection); (ii) mechanical properties; (iii) dimensional accuracy; and (iv) surface tolerances. The fabricated WjIT component was previously made from ASTM A131 DH36 steel plate [4], A cast steel alloy that can provide similar tensile and impact properties is ASTM A757 C1Q steel [5]. Chemical composition maximum limits of ASTM A 757 C1Q steel are presented in Table 1. Tensile and impact minimum requirements for ASTM A 757 C IQ and ASTM A 131 DH36 are shown in Table 2. Table 1 Chemical composition of ASTM A 757 C1Q Carbon
Manganese
Phosphorus
Sulfur
Silicon
Nickel
Molybdenum
0.25
1.2
0.025
0.025
0.6
1.5/2.0
0.15/0.30
Table 2 Tensile and impact requirements for ASTM A 757 C IQ and ASTM A 131 DH36 UTS (MPa) 515
YS (MPa) 380
El
(%) 22
RA
CVN (J) 20@-46°C
(%) 35
355 22 49024@-20°C in T direction n/a 620 34@-20°C in L direction T- Transverse direction, L- longitudinal direction
Specification ASTMA757C1Q ASTMA131 DH36
Requirements for NDT were selected in accordance to NAVSEA Technical Publication S9074-ARGIB-010/278 [6] and ABS Guide for Building and Classing Naval Vessels 2004 [7]. Thus, the castings have to be subjected to dimensional, visual, magnet particle (MT), die penetrant (PT) and radiography (RT) inspections. Each casting has to be 100% visually inspected for any surface defect. Criticality level 2 requires 50% minimum RT coverage. Level 4 is required for shrinkage, porosity and inclusion in accordance to ASTM E 446 [8]. Mold Design Technology Development: Optimized Rigging System for WjIT Part B ProMetal RCT designed the rigging system (gating and risering), and NMC performed the simulation work to improve the rigging system design (see Figure 2). The process and material parameters used in the simulation are shown in Table 3. The chemical composition, thermophysical and mechanical properties of C1Q steel material used in the simulations were functions of temperature. Also, the mold material, chills and exothermic neck-down sleeves used in the simulations were also functions of temperature. 10 ppi ceramic foam filters were used on each runner for each component. NMC used Nova Flow & Solid™ and NovaStress™ software [3] for these simulations. These casting simulation software tools were previously validated by CTC [9-11]. For part B, the best casting position was determined to be at 15 degree orientation (e.g., tilting angle). However, due to some casting and mold printing potential issues, the 0 degree casting position was used. Then, the bottom poured gating system was hand calculated and added to the part and a second analysis was performed with mold filling to
250
determine the best location of risers and chills. After 3-5 iterations for each part, an optimized rigging system was developed. The predicted mold filling time was about 10 seconds and the predicted total solidification time of the casting with rig is about 75 minutes.
Figure 2. Optimized casting mold rig design system and mold package for Part B. Table 3. Simulation parameters Simulation parameters
Material type /Value
Mold Mold thickness
Silica Sand/Furan binder Min. 50 mm
Initial Mold and Ambient Temperature
20 °C -25 s
Pouring time (ladle pouring-over lip) Pouring Temperature ASTMA757C1Q
1560°C T L = 1489 °C,Ts= 1437 °C
The casting with rigging weighed 661.5 Kg. No misruns and coldshuts were predicted. To avoid significant turbulence in the in-gates and mold erosion, the sprue was choked and seven 10 ppi ceramic foam filters were used, one for each runner. The predicted flow rate was relatively constant during filling demonstrating that the gating system is properly calculated. To avoid mold erosion, mold wash was applied to the bottom of the sprue well. The predicted (recommended) mold shakedown time was about 24 hours. Figure 3 illustrates the temperature distribution at the end of solidification and the solidification time profile. Figure 4 shows the predictions for CIQ Part B in terms macro- and micro-shrinkage (e.g., shrinkage porosities) profiles. It can be seen from Figures 3 and 4 that the exothermic sleeves and the chills performed very well since the risers are still very hot and all the shrinkage was directed toward the risers. The predicted pressure (e.g., tensile and compressive stresses) and the yield strength of the Part B casting at a maximum temperature of 1400 °C are shown in Figure 5. The predicted pressure and the yield strength of the Part B casting during cooling are shown in Figure 6. The thermal tensile stresses shown in Figure 5 are relatively high especially in the nose area at temperatures near Solidus. The tensile stresses are slightly below the yield stress of the CIQ steel casting at 1400 °C. Therefore, slightly higher hot tearing and cracking tendency would be expected for CIQ component under the current rig design and casting conditions. Mold Design Technology Development: Optimized Rigging System for WjIT Parts A and C A similar methodology was applied to design the molds for parts A and C. Figure 7 shows the predictions for CIQ Parts A and C in terms macro-shrinkage and Figure 8 shows the micro-shrinkage (e.g., shrinkage porosities) profiles for both parts.
251
Figure 3. Predicted temperature distribution at the end of solidification and the solidification time profile for Part B.
Figure 4. Predicted macro-shrinkage and porosity profiles for Part B.
Figure 5. Stress analysis results: Pressure and yield strength distributions at the end of solidification (time ~ 75 min).
Figure 6. Stress analysis results: Pressure and yield strength distributions during cooling (time 200 min).
252
Figure 7. Macro-shrinkage (a) and shrinkage porosities (b) predictions for Part A.
Figure 8. Macro-shrinkage ((a) and (b)) and shrinkage porosities ((c) and (d)) predictions for PartC. Manufacturing of WjIT Entry Edge Parts The assembled mold for Part B of the WjIT is presented in Figure 9. In addition to the WjIT segments, several plate shape samples were cast separately. The plate shape sample was used to develop the appropriate heat treatment parameters.
Figure 9. Assembled cores (mold) for part B of WjIT. Castings were shot blasted to remove the oxide from the surface and were then inspected which consisted of visual and dimensional inspection, magnet particle and radiographie inspection after the risers were removed by flame cut and grinding. Magnetic particle examination indicated that (i) Parts A and C are free of defects and (ii) the presence of three cracks in the nose area of Part B of the WjIT. The cracks were excavated and the crack area was
253
reexamined with the die penetrant. The die penetrant indicated no presence of cracks after excavation. Part B of WjIT was then sent to heat treat. Heat treatment consisted of normalizing at 898°C for 4 hours and then air cooling. The part was again die penetrant examined to make sure there are no cracks before weld repair. Part B was preheated to 186° C; weld repaired using 8018-C3 8 mm diameter electrode and stress relieved at 537° C. After stress relieve, Part B was again magnetic particle inspected and no indication was detected. The shape of the Part B WjIT casting after mold removal in shakeout and after die penetrant and X ray inspection is presented in Figure 10a and 10b, respectively.
Figure 10. WjIT after Mold Shakeout (a) and after Die Penetrant and X-ray Inspection (b). Comparison of Simulation Results with Actual Test Results A comparison between predictions and experimental measurements for porosities are illustrated in Figures 4, 7, 8, and 11-14. The locations for radiographie inspection are also shown in Figures 11, 13, and 14. The predictions compare well with the radiography results in terms of shrinkage amount and location. Figure 12 show the cracks developed after riser removal prior to any weld repairs. The tensile stresses at the end of solidification and during cooling are relatively high in the nose area (see also predictions in Figure 8) and can create hot tears if associated with segregation [12, 13] and porosities that act as nucleation sites for hot tears [14]. Hot tears will act as nucleation sites for cracks that may propagate later on during cooling and further processing (e.g., riser removal and/or welding) [14]. Carbon positive segregation toward the risers can also significantly contribute to the increase of the cracking tendency [14] in steel alloys, especially in the regions beneath the riser necks, where cracks are typically found after plasma riser removal. The casting trials confirm the predictions that Part B has some potential for hot tearing and cracking development and additional steps have been taken to minimize these defects. The steps included mold shakeout after minimum 24 hours, stress relieving immediately after casting and before riser removal, and additional improvements in the mold design minimize the thermal gradients in the casting during solidification and cooling.
Figure 11. Actual shrinkage porosities for Part B (visual inspection and radiography)
254
Figure 12. Part B nose casting showing cracks 1,2 and 3 prior to any weld repair.
Figure 13. Comparison of Shrinkage Porosities (Visual Inspection and Radiography) and Predictions for Part C.
Figure 14. Comparison of Shrinkage Porosities (Visual Inspection and Radiography) and Predictions for Part C. Concluding Remarks A comprehensive simulation was performed to assist in the mold design developments for the LCS WjlT cast steel components to minimize mold filling and solidification-related defects (misruns, coldshuts, shrinkage, porosities and hot tears), as well as post-solidification related defects such as hot and cold cracks, distortion and residual stresses. It was demonstrated that the model predictions compare well with the experiments in terms of shrinkage amount and location.
255
From this simulation study, NMC established that it is, indeed, feasible to produce the LCS WjIT casting components using the specified material and ProMetal RCT advanced molding processing technology [2]. NMC also determined that the main factors that influence the integrity and mechanical properties of these complex castings, and therefore, their quality, are the melt chemistry, the foundry practice and the solidification characteristics [1]. Several additional improvements were made on the mold rig design of all WjIT components under the existing foundry practice to further minimize the amount of shrinkage, porosities, and deformation, as well as the tendency for hot tearing and cracking. Prototype castings were manufactured and successfully passed NDT inspection and met all the mechanical properties and dimensional requirements. The actual results of radiography inspection compared well with the predicted shrinkage and micro-porosity. Acknowledgments This work was conducted by the Navy Metalworking Center, operated by Concurrent Technologies Corporation under Contract No. N00014-06-D-0048 to the Office of Naval Research (ONR) as part of the U.S. Navy Manufacturing Technology Program. Approved for public release; distribution is unlimited. The authors would like to thank ProMetal RCT personnel for successfully manufacturing the cast LCS WjIT entry edge components. References 1. J. Romanelli et al, "LCS Waterjet Inlet Tunnel Manufacturing Improvement S2279: Task 1: Entry Edge Manufacturing Improvement Report" (NMC Final Report, TR No. 10-131,2010). 2. ProMetal RCT (www.prometal-rct.com). 3. Novaflow&Solid™ and NovaStress™ Software (Novacast AB, Sweden, www.novacast.se). 4. ASTM Standard A 131/A 131M-04a "Standard Specification for Structural Steel for Ships" (ASTM West Conshohocken, PA, 2004). 5. ASTM Standard A757/757M-00 "Standard Specification for Steel Castings" (ASTM West Conshohocken, PA, 2000). 6. 9074-AR-GIB-010/278 NAVSEA Technical Publication:"Requirements for Fabrication Welding and Inspection, and Casting Inspection and Repair for Machinery, Piping, and Pressure Vessels (Naval Sea System Command, 1995). 7. ABS Guide for Building and Classing Naval Vessels 2004. 8. ASTM Standard E 466/A 466 "Reference Radiographs for Steel Casting up to 2 inch [51 mm] in Thickness" (ASTM West Conshohocken, PA 2000). 9. L. Nastac et al, "Advances and Challenges in Investment Casting of Ti-6-4 Alloys," International Journal of Cast Metals Research, UK, 19 (2), (2006). 10. L. Nastac, J. Valencia and K. Stefanick, "Stainless Steel Investment Casting Evaluation Rapid Response" (NMC Final Report, TRNo. 05-006 2005). I L L . Kramer et al, "Implementation of Steel Castings to Enhance Reliability and Decrease Cost for the M777 Lightweight Howitzer" (NMC Final Report, TRNo. 07-02,2007). 12. L. Nastac, Modeling and Simulation of Microstructure Evolution in Solidifying Alloys (Springer Verlag, 2004, 305 pages). 13. L. Nastac, A. Patel, and G. Maurer, "Modeling of Macro-segregation in a permanent mold casting" (Proceedings of the International Symposium on Liquid Metal Processing and Casting, Santa Fe NM, September 20-23,2009, TMS, 2009). 14. L. Nastac, "Estimation of Hot Tears in Castings" (CCAT-AFRL meeting, Springfield, OH, November 9-10,2009).
256
Shape Casting: The 4>h International Symposium Edited by: Mural Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
E V A L U A T I O N O F T H E D I S T O R T I O N OF A H Y D R O TURBINE BLADE DURING HEAT TREATMENT PROCESS Jinwu Kang, Xiaokun Hao, Gang Nie, Haimin Long, Hailiang Yu and Tianyou Huang Key Laboratory for Advanced Materials Processing Technology, Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China Keywords: Turbine blade casting, Heat treatment, Distortion, Curvature Abstract Hydro turbine blade castings are susceptible to distortion during heat treatment process due to their thin and curved shapes. By the numerical simulation method, the displacement results of the castings can be acquired. However, the displacement consists of distortion and contraction or expansion and depends on the selection of reference points which are hard to be exactly selected. In this paper, a distortion evaluation method is presented, in which the curvature variation of local areas with respect to the original shape is utilized. By adding the displacement results, the finite element model at each step is converted into STL format files. Then, based on the STL files, the curvature around each vertex is calculated and the curvature variation of each step relative the original shape is acquired for the description of distortion. The distortion degree of the whole casting is evaluated by the local variation of curvature, which is independent on references points and also suitable for the evaluation of distortion during heat treating process. Reference points can be selected in the areas with smallest distortion, which can be used for the displacement evaluation. 1. Introduction Distortion is one of often found problems of castings, especially large castings with thin sections such as frame shape or those with curved surfaces. To predict distortion, it is necessary to perform thermal and stress analysis of castings because distortion is the result of non-uniform cooling and closely related to the behavior of casting material. Distortion prediction is usually designated by displacement results directly obtained from stress analysis. And the comparison of the final shape and that of the beginning is used as direct illustration of distortion. [1-6]. However, the displacement results depend on the selection of constraints, as the constraints are changed, the displacement will be also different. However, this has not yet attracted researchers' attention. Castings undergo expansion, contraction and distortion during heat treatment processes; the former two are caused by cooling and phase transformation, the latter by uneven cooling of different positions of the casting. Contraction or expansion is just uniform reduction or enlargement of size, which is different from
257
distortion. Therefore, distortion must be separated from displacement and it should be independent of constraint. The author has proposed three methods, the net distortion (separated from displacement), local surface normal variations and machining allowance [7]. However, the distortion evaluation still depends on the selection of reference points. Chen also proposed to evaluate a qualified impeller casting by the final distribution of uniform machining allowance [8]. In this paper, a curvature method is presented to depict distortion without reference points. And it is used to analyze the distortion of a heavy hydro turbine blade casting. 2. Distortion evaluation by curvature variation Casting process starts from part design which is provided by product design department. Foundry engineers add machining allowance to the part design and enlarge it by contraction rate and design rigging and gating system. For numerical simulation, the casting design has to be enmeshed to finite difference (FD) or finite element (FE) models for thermal and stress analysis, and stress analysis is usually performed using finite element modeling. Numerical simulation provides displacement of casting as the final results. The displacement results of each step are added to the original shape, then, the shape of each step is obtained. The shape in finite element model is converted into STL file which are a group of discretized triangles. Curvature of each vertex of the triangle is calculated. The flowchart of this method is shown in Figure 1. Part design Add machining allowance, contraction rate Casting design (FE model)
Stress analysis of heat treatment process
Displacement results Casting design (Surface STL file)
Simulated casting shape (Surface STL file)
Curvature variation
Figure 1 Flowchart of the distortion calculation methods
258
2.1 Conversion of finite element model to STL format As the stress analysis is finished, the coordinates of each node is modified by adding the displacement result. Then, the finite element model of each step is obtained. Select the surface elements and surface nodes, and then judge the surface of each surface element; if the element type is tetrahedral, each surface is a triangle, which can be directly transformed into the format of STL triangles. As it is hexahedral elements, each quadrilateral shape can be separated into two triangles. 2.2 Calculation of local curvature The curvature of a vertex of a triangle in STL format is calculated as follows. m
1ι,=2π-Σθ„
(1)
Where #,· is the angle ; in the triangle j , m is the number of triangles at node /', as shown in Figure 2. The curvature of each vertex is smoothed by averaging its value and that of its surrounding vertices. The variation of curvature at vertex / is M/=*,'-*,"
(2)
Where M / is the relative variation of curvature at the time t, kt andA:,' are the curvature at the beginning and time t. The variation of curvature is used to depict the distortion.
Figure 2 Discretized surface triangles of casting (STL format) 3. Distortion of a heavy turbine blade during casting process The hydro turbine blade for Three Gorges Project is made of Cr-Ni stainless steel weighing 18 tons after machining. The casting undergoes normalizing, first tempering and second tempering. During normalizing process, the casting is heated to 1050°C
259
and held for a period of time and then cooled by forced air flow. As the casting cools to the martensite transformation start temperature, Ms (276°C), martensitic transformation starts and finishes at Mf (78°C). The casting is simulated by Deform-3D for the normalizing process, with the bottom constrained. Based on the finite element model, the STL file is formed, as shown in Figure 3. Based on the above method, the curvature is calculated, as shown in Figure 3(c). It can be seen along the diagonal line from the bottom right corner to the upper left corner is of big curvature. The negative value of curvature is because of the spatially twisted shape.
(a) finite element model (b) STL format (c) initial curvature distribution Figure 3 The format conversion from finite element model to STL The heat treatment thermal schedule and the heating and cooling curves of the highest and lowest temperature of the blade are shown in Figure 4. There are two small temperature difference peaks during heating and a big peak during fast cooling, representing the uneven temperature distribution which results distortion. The temperature, displacement and the curvature variation results are illustrated in Figure 5. During heat treating process, the two upper corners exhibit relatively significant displacement which depends on the selection of reference points. However, the curvature variation of each step illustrates the actual distortion which is independent of displacement. Displacement is not exactly distortion, while, the curvature variation can directly describe it. That means the area with big displacement may be mainly the result of distortion, or the result of other area which serves as a distortion source. For example, a plate is bent at the center line, the two ends will be greatly displaced, but the two ends remain flat. Under this circumstance, the displacements of the two ends are not caused by the distortion of itself, the real reason is the distortion of the center line, thus, the center line behaves as a distortion source. The bottom region of this blade casting is neglected because of constraints. It can be seen the area in green color is of big curvature variation at 5.5h which serves as a distortion source and caused the distortion at the upper corners. At 22.Ih, temperature distribution is uniform, so there is small and uniform distortion without a significant distortion source. During the beginning of cooling process, the red area serves as the distortion source because of the uneven cooling. As cooling progresses, this region expends and becomes more significant. As the casting is cooled below the start point
260
of martensite transformation Ms (32.8h), the distortion source becomes weaker as shown in Figure 5 (c). Therefore, as big temperature difference exists in casting, there is significant distortion resource. As the casting is in relative uniform cooling or heating conditions, there is no significant distortion resource. From Figure 5, it can also be seen that the displacement results are not consistent with the actual distortion. The distortion is actually the variation of displacement with spatial location, which can be depicted by the curvature variation. 1000 800 V ^600
I
£400 6 ω
H
200 0 0
10
20
30
40
50
60
time (h)
Figure 4 Thermal schedule and heating, cooling curves of the blade
(c) Curvature variation Figure 5 Temperature, displacement and curvature variation of the turbine blade during heat treatment process
261
The final displacement and curvature variation are shown in Figure 6. Finally the maximum displacement mainly exists at the two upper corners. The regions close to the two corners are of little bigger curvature variation, i.e., the distortion finally happens at the two corners. Therefore, the final displacement complies with distortion. The curvature variation of the center region of casting is little. Thus, the reference points for the final distortion comparison in traditional way in production can be selected in the area with no or less distortion, i.e., the dark blue center region of casting shown in Figure 6. The calculated displacement result of point A (as shown in Figure 5 (a)) at the upper right corner of this blade casting is compared with that of the measured of a similar blade casting, as shown in Figure 7. It can be seen that they are basically in agreement with a peak of about 60mm appeared during cooling. However, the direct validation of the curvature variation is hard for real blade castings. Therefore, it is necessary to do experiments with specially designed specimen for further validation.
(a) Final displacement (mm) (b) final curvature variation (radian) Figure 6 Final displacement and curvature variation of the blade
Figure 7 Comparison of calculated and measured displacement results 4. Conclusions Distortion prediction and control is of great significance during production of castings. Based on displacement results and the finite element model obtained from thermal stress analysis of castings, the surface discretization model STL format is acquired. The curvature and curvature variation of local surface area is calculated,
262
which is used to depict the actual distortion, the area with big curvature variation serves as distortion resource during heat treatment process. This method was applied into a heavy hydro turbine blade casting. The results show that the displacement is not consistent with distortion during heat treatment process; however, the final displacement is consistent with distortion and mainly occurred at the area close to the two corners of the blade. Curvature variation is proved to be a useful method for the depiction of distortion independent of reference points. Acknowledgement The project is funded by the project 2007BAF02B02 of the National Science & Technology Program in the Eleventh Five-year Plan Period and Major National Sci-Tech Project of China No 2009ZX04014-082.
REFERENCES 1.
Y. Chen, J. W. Kang and B. C. Liu: Proceedings of the 8th International Conference on 'Modeling of Casting, Welding and Advanced Solidification Processes', June, 1998, San Diego, USA,TMS, 771-778
2.
J.W. Kang, X G. Liu, Y.B. Duan and et. al.: 5th Decennial International Conference on 'Solidification Processing' , July, 2007 , Sheffield, UK .University of Sheffield, 554-557
3.
B.C. Liu, J.W. Kang and S.M. Xiong: Sci. Technol. Adv. Mater., 2001, 2: 157-164.
4.
Y. Song, Y. Yan, R. Zhang and et al : Proc Inst Mech Eng Part B J Eng Manuf., 2002, 216: 1123-1134
5.
Yool-Kwon Oh and Je-Se Choi: Advanced Materials Research Adv. Mater. Res., 2008,47-50: 1043-1046
6.
Sung-Mo Lee, Won-Jae Lee: J Mater Eng Perform., 2005, 14: 388-394
7.
Jinwu Kang, Haimin Long, Tianjiao Wang, Tianyou Huang and Baicheng Liu: Proceedings of The 8th Pacific Rim International Conference on 'Modeling of Casting and Solidification Processes', April 12-15, 2010, Incheon, Korea, 223-228
8.
Lugui Chen, Yong Ling, Xiuhong Kang, Lijun Xia, Dianzhong Li: J Mater Sci Technol,, 2008, 24: 364-368
263
Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
The capability enhancement of aluminium casting process by application of the novel CRIMSON method Xiaojun Dai1, Mark Jolly2, Binxu Zeng3 '•2'3School of Mechanical Engineering, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK Key words: aluminium casting process, CRIMSON, melting, oxide film, Abstract The conventional foundry not only frequently use the batch melting where the aluminium alloys are melted and held in a furnace for long time, but also use the gravity sand casting process where the melted aluminium alloys are transferred using a ladle from furnace to pour station and are poured into a mould. During filling a mould the turbulent filling behaviour due to gravity is easily to make the oxide films on the surface of the liquid cracked and trapped into the liquid. Also the long exposing time of the liquid surface with the around air during melting and filling will increase the level of hydrogen absorption. All of the abovementioned factors are often the main reasons for casting defect generation. In this paper, a novel CRIMSON aluminium casting method is introduced which has a number of advantages. Instead of gravity filling method, it uses the single shot up-casting method to realize the rapid melting and rapid counter-gravity-filling mould operations which reduce the contact time between the melt and environment thus reducing the possibility of defect generation. Another advantage is the drastic reduction of energy consumption due to shortened melting and filling time. A simulation software, FLOW-3D, is used to compare the CRIMSON method with the conventional gravity casting process. A tensile bar case is used as a sample to simulate the filling process. Introduction Casting industry has been driven by the requirement of improving product quality and minimisation of the production costs. To satisfy these requirements both casting production process and energy efficiency play vital roles in the foundry. Selection of the proper casting process, facility and minimisation of the energy consumption will be the key factors for the casting enterprise to successfully compete against rivals in the tough market. In conventional foundries, the capacity of a typical aluminium alloy melt furnace usually is between the range of 100 kg and several tonnes. The liquid metal is held at about 700 °C in a holding furnace before it is transferred to a ladle and poured into a casting mould at pour station. It can take several hours or even several days for the liquid metal in a batch to be used up and any leftover metal is poured off to be re-used or scrapped for re-melting or refining in a secondary processing plant [1].
265
Quality issues can arise when the liquid metal reacts with hydrogen, oxygen and water in the atmosphere. An oxide surface layer is created when the melted aluminium alloy is exposed to the air. During filling a casting mould the turbulent filling behaviour of the liquid metal due to gravity is easy to make the oxide films on the surface of the liquid cracked and trapped into the liquid. Also the long exposing time of the liquid surface with the around air during melting, transferring and filling will increase the thickness of the oxide film on the liquid surface and the level of hydrogen absorption. All of these will result in layers of cracked oxide films, porosity and shrinkage which damage the integration of the micro-structure of the alloy, leading to degraded mechanical properties of the end product [2. 3]. In traditional casting industry the energy efficiency of a casting facility depends largely on the efficiency of its melting and heat treating performance. In association with the two performances, over 60 % of the total process energy costs are represented in a typical casting facility [4] where there are huge opportunities for metal casting industry to adopt the best energy practices which will provide the great energy saving potential. To ameliorate the current processes for increasing energy efficiency will have an important effect on reducing the production costs and promoting the competitiveness. For instance, by implementing some state of the art technologies such as the CRIMSON method in aluminium alloy casting will make use of such opportunities.
Figure 1 Schematic plan of the new casting process facility The researchers and engineers from University of Birmingham and a local company, - N-Tec LTD have co-invented a patent CRIMSON (Constrained Rapid Induction Melting Single Shot Up-Casting) method. The objectives to develop this method are to decrease the energy consumption and to meliorate the casting quality within light-metal casting industry. The methodology of the new method is that foundries, using an induction furnace, need only to melt the quantity of metal required to fill a single mould in a closed crucible rather than large batches that use unnecessary energy and create more rejects. As shown in Figure 1. the closed
266
crucible in the induction furnace, then, is transferred to a pour station and the melted metal is pushed up using a computer controlled anti-gravity filling method to fill the mould. Due to a characteristic of rapid melting, transfer and filling in the new method, the holding time of melted metal is minimised, the possibility of hydrogen absorption and formation of surface oxide film are decreased largely and in the mean time a huge amount of energy saving is achieved [5]. In this paper, the tensile test bar was used as a sample and a CFD simulation software (Flow3D) was used to simulate the casting filling process. Sand moulds for the new up-casting method and conventional gravity sand casting were designed to compare their filling behaviours using the numerical method. In addition the traditional melting process from one local company was investigated and it was compared with the new up-casting method. We mainly focus on the issues of quality and energy saving, other issues will be investigated later. The simulation results of filling the two different sand moulds were compared to see how the new method can avoid the turbulent behaviour during filling process. The calculation and analysis of energy consumption were completed to see what the difference between the current melting processes and the novel method. Thus, the quality issue and the potential energy saving for the new method can be found. Runner system design and simulation software Runner system design of CRIMSON method and gravity sand casting method A runner system for CRIMSON method is shown in Figure 2(a) where the system consists of runner, ingate, riser, feeder and six tensile bars. A runner system of gravity sand casting is shown in Figure 2(b) where the system consists of basin, down sprue, runner, ingate, riser, feeder and six tensile bars.
(a) CRIMSON
(b) Gravity sand casting
Figure 2 Structure of the tensile bar with different runner systems. Simulation software and conditions In order to simulate the liquid metal flow in different runner systems, a commercial Flow-3D CFD simulation software was used. A 'velocity magnitude' method of the Flow-3D code was
267
used to predict the flow behaviour of liquid metal during filling a mould. The simulation was implemented using a Workstation with 16.00GB RAM and eight 2.66GHz CPUs. For the simulation of the CRIMSON runner system, Finite Difference Method (FDM) was used to generate the mesh which includes about 170,000 control volumes (cells). The filling flow rate of 0.25 L.s"1 [6] and a pressure of 9 kPa were used. For the runner system of gravity sand casting, the mesh has about 240,000 control volumes (cells). Same main conditions for the simulation of both runner systems are: pouring temperature is 700 °C and the around atmosphere pressure is 1 atm (1.013xl05 kPa). Test facilities and casting sample for energy saving Test facilities CRIMSONfacility: the structure and layout of the novel casting process facility is indicated in Figure 1 .Where its functions and features are: • High power Induction furnace (275 KW): it is used to quickly heat and melt the metal to the required temperature. Each time a billet with required size and weight is put in; • Up-caster: when the crucible with the melted metal inside is ready, it is moved and cramped in the right position in Up-caster and a mould is located on the top of pouring position, a piston in the Up-caster will raise and push the melted metal in the crucible into the mould; • Computer-controlled operation board: the movement of the piston in Up-caster is automatically controlled by the pre-programmed computer program; • Mould transfer stop: after pouring, cooling down and solidification, the mould can be moved to the transfer stop, waiting for lifting and cleaning;
Figure 3 Schematic of the aluminium meltingfurnace in G&W LTD.
268
Melting facility of G&W LTD: Grainger & Worrall (G&W) LTD. is currently using one type of melting furnace (4 tonne, Figure 3) with combining melting process where the primary melting area functioning like a stack melter and gas is used to preheat and melt aluminium ingot, then the melted aluminium alloy flowing along an inclined channel to a refining area where an electric resistance furnace is used. The refined liquid aluminium alloy is held in the electric resistance furnace. The holding time for the furnace is up to 4-5 days. The overheating temperature of A354 aluminium alloy is 760 °C and the pouring temperature is 700 °C. Casting sample and mould selected for comparing the energy consumption Half of sand mould of the tensile bar is shown in Figure 4 which has also been selected to use novel method and conventional casting process from G&W to examine the difference of energy consumption. The mould with a runner system has a profile of 530 mm length x 390 mm width x 100 mm height and has a weight of 4 kg for filling metal [7].
Figure 4 Sand mould of the tensile bar Results and discussion Simulation For the simulation of runner system of CRIMSON method, the filling time is 6.08 seconds. The simulation took about 30 minutes. The velocity magnitude of the liquid metal during filling is depicted in Figure 5 where the filling velocity of liquid metal can be observed using the velocity scale.
Figure 5 Numerical simulation of runner system of CRIMSON method.
269
For the simulation of runner system of gravity sand casting, the filling time is 3.67 seconds. The simulation took about 20 minutes. The velocity magnitude of the liquid metal during filling a mould is described in Figure 6 where the velocity of liquid flow can be judged using the velocity scale.
Figure 6 Numerical simulation of runner system of gravity sand casting. From Figure 5, it is found that the maximum velocity of liquid metal flow during filling is 0.4 m.s"1 which is less than 0.5 m.s"'. The quiet flow behaviour of liquid metal in the runner system during filling was approved proper for avoiding the generation of trapped oxide films, porosity and other casting defects [8]. In addition the filling method using against gravity decreases the exposing time to the air which will reduce the opportunity of generating oxide films. From Figure 6, the maximum velocities of liquid metal flow in downsprue and in runner during filling are more than 1.0 m.s"1, respectively. These phenomena mean that the violent and turbulent follow flow behaviour will easily to crack the oxide films on the liquid and make them trapped into the liquid. Although the filter existed in the runner system can leach the coarse trapped oxide films from the liquid which depends on the size of the holes and effectiveness of the filter, the fine oxide films will still pass through the filter and remain in the liquid. After solidification, these remained fine oxide films will generate the defects such as porosity, shrinkage etc [2, 3]. In the mean time, the turbulent flow behaviour of liquid metal in the downsprue and runner will readily make the air or hydrogen entrapped into the liquid where the porosity or bubble will be formed which will impair the mechanical properties of the casting [9]. In addition the long transferring time from furnace to ladle and the usage of the traditional gravity filling method make the melted liquid metal exposed to the air for long time, which increases the opportunity of generating oxide films on the surface of the liquid metal. Based on the abovementioned discussion, it is found that the up-casting filling process that the new method used can drastically reduce the opportunity of generating oxide film on the surface of the liquid alloy and the potential time for hydrogen absorption. Therefore, the quality of the casting can be secured accordingly. In opposition to the new process, the filling process of the conventional gravity sand casting process has the violent flow behaviour which will easily make the oxide films on the liquid metal and the air entrapped into the liquid
270
which will generate different type of casting defects and damage the mechanical properties of casting. Energy saving The energy consumption of the two types of melting processes was investigated and the results and the calculated energy efficiency were recorded in Table 1. Melting process
Energy consumption
Energy efficiency
G&W (Gas+Ele)
27.86 MJ.kg-1 (Gas: 17.78 MJ.kg-' Eie: 10.08 MJ.kg1) 1.98 MJ.kg'1
Gas: 5.65% Eie: 1.70%
Normal energy efficiency of furnace Gas: 7-19% Eie: 59-76%
57.8%
Eie: 59-76%
CRIMSON Induction furnace
Table I Energy consumption and energy efficiency of the two different meltingfacilities As shown in Table 1, the thermal efficiency of the melt furnace at G&W for gas is 5.65 % and for electricity is 1.70%. The former is near the normal thermal efficiency of crucible furnace using gas (7-19 %). The later is far more less than the normal thermal efficiency (59-76%) of an induction furnace using electricity. This means that there is lot of energy loss for the current melting process at G&W due to the long holding time. Therefore, it is assumed that if the current long melting and holding process at G&W could be replaced by the new single shot melting method, 14 times more energy can be saved. It is estimated that 26 GJ.tonne"1 (7.2 MWh.tonne1) can be saved for producing every tonne of A354 casting alloys when using the new process and thus the melting cost will be significantly decreased. Conclusions and future works From the simulation results, the maximum flow velocity of the new CRIMSON method during filling a mould is less than 0.5 m.s"1. The total flow behaviour is quiet and stable which reduces the possibility of generation of oxide films and other casting defects. In opposition to the new method, the conventional gravity sand casting process has turbulent flow behaviour during filling a casting mould where the maximum velocities of liquid flow in the downsprue and runner are more than 1.0 m.s"1. The violent flow behaviour will not only easily make the oxide films on the liquid metal cracked and entrapped into the liquid but also make the air trapped into the liquid. These entrapped oxide films and air will generate different types of casting defects such as porosity, shrinkage and bubble etc which will damage the mechanical properties of the final casting. The investigation on examining the energy consumption and the melting efficiency of both methods has showed that the CRIMSON process is a novel method for reducing energy consumption. If the traditional foundries could use the new melting method instead of their traditional melting process, the estimated energy savings could be of the order of
271
26 GJ.tonne-1 (7.2 MWh.tonne"1) in this case. This could hugely reduce the production cost by about £546 pounds.tonne"1 (7.6 p.kWh"1) and would tremendously increase the company's competitiveness. To validate the simulation results, the experiment on comparing the mechanical properties of the tensile bar using different filling processes will be implemented in next stage. The other issues of the energy efficiency for the foundry will be considered too where not only the melting process is included, other relevant processes should be considered. Acknowledgement This research project is sponsored by the Engineering and Physical Sciences Research Council (EPSRC) of the UK under the grant of EP/G060096/1. Many thanks to the University of Birmingham, N-Tec LTD. and Grainger & Worrall Ltd for providing the experiment equipment and data. References [1] Casting method saves energy, 25, Mar 2009, Professional Engineering Magazine. http://www.profeng.com/areliive/2009/2206/22060076.htm [2] X. Dai, X. Yang, J. Campbell and J. Wood, 2003, Effects of runner system design on the mechanical strength of Al-7Si-Mg alloy castings, Journal of Materials Science and Engineering. A354. Pg 315-325. [3] X. Dai, X. Yang, J. Campbell & J. Wood, "The Influence of Oxide Film Defects generated in Filling on the Mechanical Strength of Aluminium alloy Castings", Materials Science and Technology, 2004, 20(4), 505-513. [4] R. Eppich, R.D. Naranjo., Implementation of Metal Casting Best Practices, Report of U.S. Department of Energy, 2007. [5] M. Jolly, Energy Saving in the Foundry Industry by Using the "CRIMSON" Single Shot UPCasting Process, 2010 TMS Annual Meeting & Exhibition, Febuary 14-18, 2010, Seattle, WA. [6] J. C. Gebelin, M. Lovis & M.R. Jolly, SIMULATION OF TENSILE TEST BARS: DOES THE FILLING METHOD MATTER?, Symposium on Simulation of Aluminum Shape Casting Processing, TMS2006 March 2006, Warrendale, PA, Eds Q. Wang, M.J.M Krane and P.D Lee. [7] X. Dai, M. Jolly, Potential energy savings by application of the novel CRIMSON aluminium casting process, Applied Energy (Accepted for publication) [8] J. Campbell, (1991) Casting, Butterworth-Heinemann, Oxford. [9] M. Divandari, 2001, PhD Thesis, 'Mechanisms of bubble damage in castings'. University of Birmingham.
272
Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
Optimization of the process parameters and tooling improvement for the rheocasting of high quality aluminum components using the SEED process Chang Qing Zheng, Ehab Samuel, Florentin Laplume National Research Council of Canada, Aluminum Technology Centre (ATC-NRC), 501 boulevard de l'Université est, Chicoutimi (QC), G7H 8C3 Keywords: rheocasting, aluminum, SEED process, semi-solid, semi-solid casting, HPDC Abstract The automotive industry has leaned greatly towards the use of aluminum alloys by virtue of their strength and low density. Given this, the potential for aluminum use in the fabrication of vehicle parts has greatly increased. However, there are limited studies devoted to the improvement of the casting process. In the present work, the SEED (Swirled Enthalpy Equilibrium Device) rheocasting method, as developed by Rio Tinto Alcan in collaboration with the Aluminium Technology Center of NRC Canada (ATC-NRC), was analyzed by the authors in an attempt to optimize operating parameters (e.g. proper mold filling, slurry temperature, injection speed, etc.), which affect the final cast part quality. In many of the existing semi-solid casting processes which use billets as feedstock, for example, it is often found that the outer surface of the billets is contaminated. During the injection phase, a billet's external skin comes into contact with air and lubricant, and, as a result, becomes contaminated. The use of such a contaminated billet can often result in an increased rejection rate of cast parts. The SEED process, which uses heat extraction of the liquid aluminum alloy via mechanical agitation (swirling) in a confined cylinder to form the semi-solid billet on site, has already proven successful in producing sound aluminum castings having an excellent combination of strength and ductility. The resulting semi-solid billet, having a microstructure consisting of a-Al globules surrounded by the eutectic phase, is then injected into the cold chamber of an HPDC machine. Introduction Conventional shape casting processes such as pressure die casting, for example, offer both low cost and high productivity. The downside, however, is that pressure die-cast parts are generally prone to fabrication defects and limited mechanical properties [1]. One possible remedy is semisolid pressure die casting, which can yield quality parts comparable to those made with wrought processes; the cost of casting is comparable to using conventional pressure die casting. Indeed, semi-solid pressure die casting can eliminate casting defects and allow for additional improvement, such as the use of heat treatments, to the production process [2, 3]. Within the framework of a collaborative effort between Rio Tinto Alcan and ATC-NRC, studies were carried out to assess the performance of a 357 alloy designed for semi-solid pressure die casting and improve casting tooling design which, in turn, helps to improve mechanical properties; these studies are based on a new design of experiment (DOE) method that builds on experience, i.e. the Expectation-Maximization (EM) Method. The semi-solid billets were prepared using the SEED process [4], as shown in Figure 1.
273
Figure 1. SEED process of Rio Tinto Alcan |3]. This patented method [5] employs the heat extraction of the liquid aluminum alloy via swirling, in a confined cylinder. The resulting semi-solid billet is injected into an HPDC press to yield quality parts. A proper set of working parameters can lead to a uniform and globular microstructure which, in turn, can decrease slurry viscosity and improve die filling; likewise, an improper use of these parameters can result in defects, and reduced mechanical properties [6, 7]. Methodology In the DOE, the EM Method was used to keep the number of tests to a minimum. This consists of: (i) defining parameters (factors and responses) and objectives, (ii) conducting experimental tests and building operational domains, (iii) regressions and building models, (iv) validation and (v) multi-criteria optimization, which consists in maximizing a desirability function that expresses the importance of combined criteria in relation with targeted goals. A direct research method is then applied to find optimal solutions [8, 9]. The operational domain, in a multidimensional space, is modeled as an ellipsoid (Figure 2) and defined with continuous values, which discriminate between accepted tests (inside) and rejected tests (outside). This is not to say that the test points which do not fit inside the operational domain are discarded from the work, rather they are not included in the ellipsoid. However, they are still used in the overall model. Test points have coordinates (a, b, c, d, e...), which change with the number of operational parameters (i.e. factors) being considered.
Figure 2. Simplified view of the multidimensional operational domain, modeled as an ellipsoid.
274
Because this ellipsoid is a crude representation of the operational domain, perfect discrimination between all the accepted and rejected points is not always possible. Usually, limitations and constraints on the factors truncate the ellipsoid, contributing to a more realistic shape of the operational domain. For this work, the ΈΜ Optimization' software by EM Optimization International Inc. was used. Semi-solid 357 aluminum alloy wedge plate parts (Figure 3) were cast using the SEED process highlighted in Figure 1, with a 530-tonne HPDC press. These plates were subsequently subjected to a T6 heat treatment consisting of (i) a solution heat treatment of 540°C for 6 hours, (ii) a 20°C water quench and (iii) an aging step of 170°C for 6 hours. Following this, ASTM B557 round bar tensile samples, with a gage length of 25 mm and a gage diameter of 6.35 mm, were machined at pre-assigned locations along the plate length (Figure 4).
Figure 3. Schematic of the cast wedge plate.
Figure 4. Position of tensile samples and hardness test points on the cast wedge plate. For simplicity, we will only consider the results obtained for section A (Figure 4). In other words, the mechanical properties such as yield strength, ultimate tensile strength, etc. will all be measured from tensile samples taken at position A. Standard liquid metal treatment and control steps were applied to the aluminum in the furnace, e.g. chemical composition check, rotary fluxing, degassing, etc. Results and Discussion In this work, 13 parameters (seven factors and six responses) were considered. These are listed in Tables 1 and 2 for factors and responses, respectively. Figure 5 illustrates the scanned area used for measuring the average a-Al globule size. In this work, 55 castings were produced. Of these, 34 met the pre-established ranges. The distribution of these 55 test points is relatively uniform inside the operational domain (Figure 6). Regression models, based on the tests carried out, were established to quantify the effects of the working parameters on the responses.
275
Table 1. Factors and their typical range.
Table 2. Responses and their typical range.
Figure 5. Scanned section tensile sample for average width value (μιη) of o-AI globules.
Figure 6. Operational domain and test distribution (total of 55 tests).
276
Figure 7 demonstrates the degree of influence of the factors on a given response - in this case, the yield strength at section A. These effects are normalized, and shown in terms of being either positive (moving to the right) or negative (moving to the left). The greater a factor's influence (either positive or negative) on a response, the longer the red bar is in the histogram of Figure 7.
Figure 7. Factors and their relative impact on the yield strength (at section A). According to Figure 7, Sr concentration (factor 'F') has the most influence on the yield strength, while filling speed ('Β') has the least. Die temperature (factor 'D'), on the other hand, has no effect, as it is not shown in the histogram. There is an interaction between filling speed ('Β') and intensification pressure ('C'), as well as pre-filling speed ('Α') and drained mass ('G'). Figure 8 illustrates the behaviour of the yield strength response in more detail. The interactions demonstrate that the effect of factors C and A, for instance, are only significant when they are combined with B and A, respectively. In other words, the relationship is proportional, rather than linear (as in the case of Sr concentration) or squared (as in the case of the metal pouring temperature).
Figure 8. Yield strength at section A relative to filling speed, metal pouring temperature, drained mass, Sr concentration and die temperature. It can be seen from Figure 8 that the yield strength: (i) increases from 277 to 282 MPa as the filling speed increases from 0.2 to 1.4 m/s, (ii) decreases, but then increases, as the metal pouring temperature increases (note that the curve is not linear like the others, owing to the E2 relation shown in Figure 7), (iii) decreases from 292 to 269 MPa as the Sr concentration increases from 30 to 100 ppm, (iv) decreases from 287 to 270 MPa as the drained mass increases from 80 to 268 g, and (v) remains constant at all die temperatures tested. These observations can likewise be
277
made by considering the behaviour of any of the other responses (ultimate tensile strength, elongation, etc.) with respect to the effects of the working parameters. Histograms similar to Figure 7 are presented in Figure 9, for the remaining responses. It can be seen here, for example, that the Sr concentration greatly affects the percent elongation, a-Al globule size and density. On the other hand, the die temperature has little effect on the percent elongation or the hardness.
Figure 9. Relative impact of operational parameters (factors) on responses. As can be seen from Figures 7 to 9, the regression models derived from the multi-dimensional analysis show the degree of influence of the operating parameters (factors) on the alloy's mechanical properties, microstructure (by measurement of primary a-Al phase size) and porosity defects (by density measurement). The overall influence of the operational parameters on the responses studied in this work is fully summarized in Table 3. Table 3. Effects of operating parameters (factors) on responses.
According to Table 3, is can be observed that: (i) the prefilling speed has no significant effect on the majority of the responses, (ii) the filling speed has a significant effect on hardness, (iii) the intensification pressure has a significant effect on ultimate tensile strength and hardness, (iv) the die temperature has a significant effect on percent elongation and a-Al globule size, (v) the metal pouring temperature has a significant effect on yield strength, ultimate tensile strength and density, (vi) the strontium concentration has a significant effect on yield strength, percent elongation, a-Al globule size and density, and (vii) the drained mass has a significant effect on all of the responses, except hardness.
278
Multi-criteria optimization was used to maximize the mechanical properties (yield strength, ultimate tensile strength and elongation) all at once, in order to improve the performance of the parts being cast. One best solution point inside the feasible domain was found using the EM Optimization software, based on the models (Figure 10).
Figure 10. Optimization by EM software and the best solution found. The end point (i.e. best solution) has coordinates (A, B, C, D, E, F, G), which correspond to the optimized values of the working parameters. Subsequent to the maximization of the mechanical properties, 50 wedge plate parts were produced wiüi these optimized parameters, in order to provide the best possible solution (i.e. the best quality parts). In order to ascertain the quality of these optimized plates, once again, ASTM B557 tensile samples were machined and testing resumed. The results obtained for this optimized set of tensile samples are presented in Table 4. Table 4. Optimized parameters and results.
As can be seen from Table 4, the predicted and observed values of the mechanical properties (yield strength, ultimate tensile strength and percent elongation) are in good agreement. Moreover, it is clear from these results that the alloy demonstrates very favourable mechanical property values, in terms of strength and ductility. Conclusions 1. After having tabulated the effects of the working parameters on the responses, it was found that the yield strength is mainly influenced by pouring temperature, Sr concentration and drained mass. Moreover, the ultimate tensile strength is mainly influenced by pouring temperature, intensification pressure and drained mass, and the elongation is mainly influenced by die temperature, Sr concentration and drained mass.
279
2. After having carried out 55 tests, the optimum operating parameters for the 357 alloy in this work were obtained. With this information in hand, a new batch of 50 optimized parts were cast resulting in mechanical (tensile) property values which were in good agreement with the values predicted by the EM Method. 3. The average mechanical (tensile) property values (yield strength = 297±4 MPa, ultimate tensile strength = 350±3 MPa and percent elongation = 8.3±0.8%) obtained in this study demonstrate a very favourable combination of strength and ductility. These properties are attributed to the uniformly globular microstructures obtained using the SEED process, which result in high integrity castings and exceptional values of mechanical properties. Acknowledgements The authors acknowledge National Research Council Canada for the support and permission to publish. The authors wish to express special thanks to Mr. Dany Drolet, Ms. Geneviève Simard and Ms. Marie-Eve Larouche at the Aluminum Technology Centre of NRC Canada. Thanks also to our collaborators at Rio Tinto Alcan R&D in the SEED development Mr. Alain Lemieux. References 1. Major, J.F. and Richman, D.: Aluminum Automotive Castings - An Ever Expanding Role in an Increasingly Competitive Market. Proceedings of the International Symposium on Recent Metallurgical Advances in Light Metals Industries, Canada, Aug. 1995, 25-42. 2. Jorstad, J.L.: Semi-Solid Metal Processing; The High Integrity Die Casting Process. Die Casting Engineer, Jan. 2004, 48(1), 42-48. 3. Yurko, J., Fleming, M., and Martinez, A.: Semi-Solid Rheocasting (SSR™) - Increasing the Capabilities of Die Casting. Die Casting Engineer, Jan. 2004,48(1) 50-52. 4. Doutre, D., Langlais, J. and Roy, N.: The SEED Process for Semi-Solid Forming, in Proceedings of the 8th International Conference on Semi-Solid Processing of Alloys and Composites, Limassol, Cyprus, pp. 397-408 (2004). 5. Doutre, D., Hay, G. and Wales P.: U.S. Patent No. 6,428,636 Aug. 6, 2002. 6. Zheng, C.Q. and Simard, A.: Optimization of Casting Parameters on an Improved AA6061 Aluminum Alloy for Semi-Solid Die Casting, Journal: Advances In Light Weight Materials Aluminum, Casting Materials, and Magnesium Technologies, USA, Apr. 2010, SP-2294. 7. Pineau, F. and Simard, G.: Investigation of the Primary Phase Segregation during the Filling of an Industrial Mould with Semi-solid A357 Aluminium, S2P 2008 International Conference, 2008, 141(43), 635-640. 8. Galopin, M., Dao, T.M., Zheng, C.Q. and Hansquine, S.: A New Approach to Machinability Testing, Seminar and Applications Forum on a Systems Approach to Machining, Institute of Advanced Manufacturing Sciences Inc. (IAMS), Cincinnati, Ohio, USA, May 1993, 4-5. 9. Galopin, M., Dao, T.M., Zheng, C.Q. and Hansquine, S.: A Problem Solving Tool for Optimization of Welding Processes, 5th International Conference on Computerization of Welding Information, Golden, Colorado, Aug. 1994, 9-12.
280
Shape Casting: The 4th International Symposium Edited by: Murat Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
SHAPED CASTINGS AND MACHINING John E. Wyatt1 & John T. Berry2 'Mississippi State University Instructional Systems & Workforce Development, Box 9730, MSU, MS, 39762, USA Mississippi State University Mechanical Engineering, Box 9552, MSU, MS, 39762, USA Keywords: Machinability, Machining Sequencing, Residual Stresses Abstract This paper will discuss three aspects of the machining of shaped castings: a) machinability, b) machining sequence, and c) residual stresses, the principal thrust being that of machinability. Currently there is a growing movement to reconsider the whole process of machining from the standpoint of fracture mechanics. This has resulted from the groundbreaking work of Atkins at Reading University (UK), which is relevant to cast alloys which invariably are multiphasic materials. Although secondary and tertiary phases have been blamed for tool wear, they are in fact intrinsically tied into the cutting process. The premise of Atkins is that ductile rupture takes place near the tool tip. These ductile tears manifest as dimples associated with the secondary and tertiary phases. Although some secondary phases will cleave (silicon particles in AISi type alloys) some will decohere from the matrix leaving dimples. The authors will describe some experiments confirming the theories of Atkins. Introduction Ever since F.W. Taylor wrote his treatise "On the Art of Cutting Metals" in 1907 [1], researchers have sought ways of creating chips more efficiently, as well as of modeling the chip formation process. There have been many worthy contributions, including the studies of Ernst, Merchant, and Shaw for example [2, 3, 4]. However, throughout the past century insufficient fundamental thought has been given to the surface that is generated by the machining process. The surface seems to have been regarded as a by-product of the machining process with the chip being the principal product. The reality of this is that the chips are in fact the by-product and the surface generated is the major product of machining. The Current Understanding of the Chip Separation Process With reference to cast and weld-repaired components, the actual mechanism regarding the process of metal removal, especially the act of separation, has not been at the center of metal cutting research in the past. However, recent work has emerged that has shed new light on the surface generation basics. Atkins [5, 6] has suggested an alternative approach than heretofore regarding the way in which metal cutting relates to fracture mechanics. Using Scanning Electron Microscopy (SEM) based approach experimental evidence has been presented by Subbiali and Melkote in the microcutting of 2024-T3 aluminum alloy [7], Melkote et alia[8, 9] have justified this new approach and have established that with small uncut chip thickness values (15 - 105 um) a ductile tear is initiated ahead of the tool nose which then propagates into either the chip or the workpiece, as illustrated in Figure 1. Furthermore, it is likely that this tear could possibly be affected by the tool geometry and other
281
cutting variables. Further discussion with the proposer of this new concept [6], suggests that this tear is present in all forms of machining. This has only recently been confirmed by work at MSU examining larger uncut chip thicknesses (250-500um). Figures 2 (a), (b), (c), and (d) present evidence for the existence of ductile tears, confirmed by the presence of classical "dimpling" phenomenon associated with secondary and/or tertiary phases. Engle and Klingele demonstrated that voids could be formed through ductile tearing that produce dimples associated with constituent phases or inclusions. This is illustrated in Figure 3 [10]. As the nascent chip forms, aided by the ductile tears referred to above, the top half of each tear will be incorporated into the underside of the chip. The lower surface of these tears will pass under the nose radius of the tool, forming part of the tertiary shear zone [7] as illustrated in Figure 4. Assuming that a "dimple" escapes with the chip, its associated nucleation feature (e.g. inclusions or constituent particles) will in turn be incorporated into the newly generated surface. This latter feature is likely to become a nucleation point for fatigue cracking or abnormal component wear in contact with subsequent mating surfaces. It should be noted that the tool tip is the last part of the tool that is in contact with the surface of the workpiece and thus plays the final part in the generation of that surface, very possibly burnishing over the remains of the ductile tear, see Figure 2(d).
Figure 1. A depiction of the presence of a ductile tear ahead of the tool nose. The chip formation process is causing the tear to open. The upper part of the torn surface generated by the cutting operation escapes partly with the chip while the lower portion enters the tertiary shear zone and is then burnished over into the final generated surface.
(a)
(b)
282
(e) (d) Figure 2. SEM Photomicrographs showing dimpling and cracking, (a) Shows cracks appearing on the underside of the chip and at the interface, (b) Shows the dimples inside the crack on the underside of the chip illustrating that these 'dimples' appear throughout the removed material, (c) Shows the interior of the crack at the tool nose interface, (d) Shows the possible remnants of a ductile tear after being burnished into the machined surface. The material is wrought AL6061 in the T6 condition
Figure 3. Engel and Klingele's depiction of ductile tearing in ductile materials by void formation [10].
Figure 4. A depiction of the shear zones in metal cutting This leads to the possibility that the surfaces generated by machining are affected by a combination of the above ductile tear, or possibly a brittle crack, produced by the chip formation process, and the burnishing effects of the tool tip at the tertiary shear zone. These potential effects are poorly understood, especially in as cast materials and weld-repaired surfaces. Properties likely to be affected would be those associated with the superficial state of stress and
283
characteristics of the refurbished layer. Such characteristics include fatigue, corrosion and oxidation resistance. Finally, these effects are especially germane to as cast and weld-repaired surfaces which are often known to contain features which would encourage ductile fracture (intermetallics and non-metallic inclusions, as well as constituent particles in age hardened materials). Thus surfaces so generated are likely to be heavily populated by what were once 'dimple' rupture zones especially in materials containing a high fraction volume of injurious inclusions. Consequently, the subsequent mechanical behavior of the machined surface will be strongly affected by this abundance, along with the stress-state obtained there. As mentioned above, this will be especially relevant to fatigue, corrosion, and oxidation resistance. In view of the changing picture of the manner in which chip formation is initiated and the special role of local ductile failure in the tool nose region of the workpiece, it is highly desirable that parallel research is undertaken in cast light alloys. Since the majority of such alloys are multiphasic, the effects of second phase population and distribution upon both chip formation and subsequent effects of this upon fatigue crack inititiation on the generated surface are ripe for study. These effects may be of special importance where weld-repair is permitted Post Machining States of Stress Residual stress in machined components is critical in determining both the wear and the fatigue characteristics ofthat component [11, 12]. It is not just the levels of stress that are critical but whether the stress is tensile or compressive. A machined surface that has tensile residual stress induced by the machining process would be prone to fatigue cracking, whereas a machined surface with compressive residual stress would be resistant to fatigue cracking [13]. It is widely accepted that the initiation stage of fatigue cracking is facilitated if the residual stresses present at the machined surface are of a tensile nature. Consequently, post-machining operations such as shot peening have been employed to provide a compressive stress field at the surface. The experimental study of residual stress associated with conventional machined surfaces goes back at least 50 years [14]. Much of the early work was concerned with turning operations in an attempt to simplify the subsequent analysis of the mechanical state obtained. Typical milling operations used in finishing castings will generate similar residual stress patterns to those observed when turning, as shown in Figure 5 [14]. The sense (sign) of the superficial residual stresses determined in this study appeared to depend upon the composition and thus hardenability aspects of the various steels examined. Figure 8 shows that the residual stress at the surface is also dependent on tool sharpness which gives rise to the varying residual stress patterns.
Figure 5. The effect of wearland on residual stress in face milling
284
Unfortunately, such stress pattern determination is often expensive and involved. However, a new low cost, readily executed method of estimating superficial residual stress directly has recently been investigated at MSU. The technique involves the change in microhardness indentation size and spacing associated with the removal of residual stresses [15, 16]. Turning to the question of predicting such residual stress patterns by computational techniques, in many previous model experiments some authors have assumed an initial state of zero stress and strain in the elements concerned. The initial state of residual stress in the as-received material or component must also be taken into account in subsequent machining operations. Although the machining operation might well remove layers affecting residual stresses which can cause component distortion, it could later superimpose its own residual stress profile on the surface that the machining process has generated. Consequently, the historical progression of mechanical states must be taken into account if effects of subsequent behavior in monotonie or repeated loading are to be predicted. Much can be learned of the effects of residual stresses upon fatigue life. Work on wrought aluminum alloys by SAAB revealed the effect of high speed machining on the fatigue resistance of an aluminum alloy workpiece under pocket milling operations [17]. Fatigue test data showed that first a decrease in fatigue resistance was observed when the cutting speed increased above the conventional speed level, 100 m/min, but then an increase occurred when the cutting speed was raised towards 3,000 m/min. The minimum resistance appeared to be in the speed range of 500 to 1000 m/min. The fatigue life of the specimens was also dependent upon the cutting mode employed. In reviewing the SAAB investigation, the Swedish Defense Research Agency [18], presented an S-N curve which indicated a serious reduction in notch fatigue life when utilizing high speed milling. Although the curves were of similar shape the high speed machining curve at a cutting speed of 1800 m/min was significantly lower than that of the conventionally machined test piece. The tests were conducted on notched specimens excised from plate (7010-T7451). The exact mechanisms and cause-effect relations that produce residual stresses and surface material properties arising in machining are thus not well documented. Certainly, windows of opportunity exist in which high speed machining can be exploited to provide excellent fatigue lives but these areas have not yet been mapped [17, 18]. Such effects can be expected to be present in weld-repaired surfaces of the type described earlier. Using MSU's low cost residual stress technique, the authors original results as seen in Figure 6 show the corresponding relationship between the fatigue life results for face milling with superficial residual stress measurements on aluminum alloy bar machined over a range of cutting speeds. The residual stress measurements were made using the newly developed method at MSU. The correspondence is indeed remarkable and lays the background for this part of the proposed work [15].
285
Cutting Spc»ei (m/minì
-2000 -1050
8»
1Q0O
Î5Q0
/»DO
2900
2030 i
-V
4000 i :
SQ30 ·
0
506
1000
1500
200O
250C
Cutting Spe«4 (mimili)
(a) Measured Superficial Residual Stress [17]
((,) Fatigue life curve
Figure 6. Correlation between residual stress and fatigue life for face milled [15] Whether these effects involve local reprecipitation or overaging phenomena or features associated with burnished-over remains of ductile tears, is also a moot point, and requires further investigation. In summary, it is usually agreed that compressive surface residual stress improves fatigue life and tensile residual stress decreases it. The literature reveals that residual stress has very much more complex effects on fatigue compared to the general belief. There is thus much to be done before residual stress affects on component life are fully understood. This is especially true for cast components as their initial state of stress needs to be known as well as the microstructure of these layers before a machining strategy can be formulated. In summary, work on the high speed machining of wrought alloys, especially those of aluminum, has established that the superficial residual stress pattern is at least in part responsible for significant variation in their fatigue resistance. The relative importance on tool geometry and cutting conditions, in particular cutting speed should be determined for cast light alloy components where extensive machining is undertaken and where these components are subject to repeated loading. The time is ripe for such investigation. Conclusions Although the primary direction of research interest in the science and engineering of metal cutting over the last half century has been centered upon either the efficiency of removing chips from the workpiece or upon the problems of tool wear improvement, both especially at high cutting speeds, precious little attention has been directed to the nature of the surface so generated. This is important for both wrought and cast materials. One exception has been the investigation of the state of residual stresses at the generated surface, something which is closely linked to fatigue crack initiation. However, much of the literature concerns wrought rather than cast light alloys. Additionally the overall pattern of residual stress, which affects component distortion, of special interest in the assembly of cast components, has not received the attention of investigators in the area of metal cutting. Recent research by Atkins, Melkote and the writers has suggested that the presence of second (and no doubt tertiary) particles affect both the initiation of chip formation as well as the initiation of fatigue cracking in the generated surface. The latter, of course, will also be
286
profoundly affected by the magnitude and sense of the superficial residual stresses implanted or modified by machining. The writers point out that although the machinability of cast alloys of all types is at last commanding appropriate attention, research is badly needed on the specific mechanisms of chip formation and its impact on the nature and properties of the surfaces generated in these materials. Acknowledgements The writers would like to acknowledge the support received from the Office of Research and Development of Mississippi State University, and from the Coleman endowment, for the support of their work. Helpful discussions with Dr. Schreyes Melkote of the Georgia Institute of Technology are also gratefully acknowledged. Finally, the contributions of several senior undergraduates must be recognized. References 1. TAYLOR F.W., 1907, On The Art of Cutting Metals, Transactions of ASME, Vol. 28, USA 2. ERNST H., 1938, Physics of Metal Cutting, Cincinnati Milling Machines L™ 3. MERCHANT M.E., 1945, Mechanics of the Metal Cutting Process, Journal of Applied Physics, Vol. 16, pp.267 - 275, May 4. SHAW M.C., 1984, Metal Cutting Principles, Oxford University Press (U.K.), ISBN 0-19859020-2 5. ATKINS A.G., 1974, Fracture Toughness and Cutting, Int. J. Production Research, Vol. 12 (2), pp. 263-274 6. ATKINS A.G., 2003, Modelling Metal Cutting Using Modern Ductile Fracture Mechanics: Quantitative Explanations for some Longstanding Problems, International Journal of Mechanical Sciences, Vol. 45, pp. 373-396. 7. SUBBIAH S., MELKOTE S.N., 2007, Evidence of Ductile Tearing Ahead of the Cutting Tool and Modeling the Energy Consumed in Material Separation in Micro-Cutting, Journal of Engineering Materials and Technology, Vol. 129, pp. 321-331, April 8. MELKOTE S.N., 2008, Private Communication 9. ALTINTAS Y., 2008, Private Communication 10. ENGEL L„ KLINGELE H., 1981, An Atlas of Metal Damage, Prentice-Hall, Englewood Cliffs, NJ, p. 41 11. LIU C.R., LIN Z.C., BARASH M.N., 1984, Thermal and Mechanical Stresses in the Workpiece During Machining, High Speed Machining, Presented at the Winter Annual Meeting of the ASME, USA
287
12. G. ARNDT, 1971, On the Study of Metal Cutting and Deformation at Ultra-High-Speeds, Proceedings of the Conference of Production Science Industry, Vol. 30, pp. 30-41 13. KALPAKJIAN S., SCHMID S.R. 2001, Manufacturing Engineering and Technology - Third Edition, Addison - Wesley Publishing Company (U.S.A.), ISBN 0 - 201 - 36131 - 0 14. FIELD, M., KAHLES, J.F., 1964, Surface Integrity of Machined and Ground High Strength Steels,OM\C Report 210. 15. WYATT J. E. & BERRY J. T., 2006, A New Technique for the Determination of Superficial Residual Stresses Associated with Machining and other Manufacturing Processes, Journal of Materials Processing Technology, Vol. 171, pp. 132-140 16. BERRY J.T., WYATT J.E., 2005, A Low Cost Method for Determining Superficial Residual Stresses as Applied to Machined Surfaces, US Patent No. 6,934,642, Issued August 23 r i 2005 17. ANSELL H., 1999, Fatigue and Damage Tolerance Aspects of High Speed Machined Airframe Parts, Meeting of the International Committee on Aeronautical Fatigue, July, Bellevue, WA. 18. BLOM A.F., PALMBERG B., 2001, A Review of Aeronautical Fatigue Investigations in Sweden During the Period June 1999 to May 2001, Swedish Defence Research Agency, FOI, The Aeronautics Division, FFA, Sweden, FOI-R-0138-SE.
288
Shape Casting: The 4lh International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
THE ESTIMABLE VALUE OF 'CLEVER' EXPERIMENTS John T. Berry Mississippi State University Mechanical Engineering, Box 9552, MSU, MS, 39762, USA Keywords: Casting Shape, Chilling Power, Experimental Validation Abstract The inexorable growth of computational modeling in materials science and engineering has been associated with a serious decline in the number of 'clever', or at least genuinely useful experiments Experimental evidence, if provided, is often second-hand and merely placed there to validate a model, thus not stimulating further study. Many established concepts in materials science have been overturned as a result of careful experiments. These experiments have often nucleated new concepts and have started a sequence of experiment pacing theory and viceversa. In his long career, the writer has witnessed many examples of this pacing effect, some of which he lists. Several of these resulted in viable industrial processes. He describes unpublished work concerning the heat extractive capacity of molds which is relevant to current processing developments. The experiments concerned spawned further analytical and computational work, providing examples of the said pacing effect. Introduction As in many areas of science and engineering, progress in materials processing research, especially in the solidification area, has been governed by experiment pacing theory and viceversa. Often this progress has been conditioned by what Albert Einstein termed 'the delicacy of observation' (1916). In the past, countless seemingly well accepted theoretical concepts have been questioned and subsequently rejected through the results of well-planned and conducted experiments that were properly interpreted. Certain of these experiments have in turn led to the formation of new theories, which have been proven correct, often modified or even themselves rejected by further experimental work. A few have perhaps resulted in a new process or product. Perhaps a prime example of this 'pacing' effect is one which involves the concept of the dislocation 'pile-up' in low-carbon ferritic steels and its connection to the yield discontinuity. Subsequent research employing the transmission electron microscope (TEM) indicated that dislocation pile-up would only occur in low stacking-fault energy (SFE) materials. Austenitic stainless steels thus exhibit such features, whereas low-carbon ferritic steels, the subject of much previous analysis, do not! Consequently, other theories based on blocked slip-bands and dislocation multiplication became favored as the basis for alternative explanations of yielding and brittle fracture in mild steels.
289
The majority of the major contributions of the last century, and indeed those of the present one, contain examples of experiment and theory pacing each other. During his professional career the writer has witnessed many such examples : • •
The efforts of US and Canadian scholars to enhance understanding and knowledge of the morphology of the freezing front (Winegard 1964) The evolution of the single crystal gas-turbine blade (Kear at alia 1969) The development of controlled-rolling and the rise of the versatile microalloy steels (Microalloying 1975) The confirmation of the damaging effects of entrained oxide bifilms during melting and casting of light alloys (Campbell 2003) The recognition of the contribution of fracture mechanics and the role of second-phase particles in the metal-cutting process (Atkins 2009 together with confirmatory experimental work by Subbiali and Melkote (2008) - see also this conference (Wyatt and Berry 2011))
The writer's initial contact with solidification related research occurred towards the end of a period where continuing experimental work, often performed by practitioners, questioned the pioneer contributions of Nicolai Chvorinov on the solidification times of shaped castings (1938, 1940, 1951, 1953). In this instance that author had performed both analytical and experimental work which was questioned by experiment. At the time of Chvorinov'sl953 publication, much progress in interpreting this work and furthering understanding had taken place in the US (Pellini, 1952,1953) The sequence of the pacing phenomenon which followed is the principal subject of this paper. Certain aspects of this sequence have considerable bearing in certain contemporary developments in the technology of shaped casting production. Clever Experiments And Their Effect On Scientific Progress Although the celebrated Chvorinov rule (1938, etc) relating casting solidification time to the square of the volume to surface area ratio was apparently anticipated by Farquar (1920), as pointed out by Ruddle (1971), it has proved to be of considerable utility to practicing steel foundry personnel, and even of help to modelers in the preliminary design of feeding systems. The volume to surface area ratio or feeding modulus, has also proved useful in other applications, both in the avoidance of solid state embrittling reactions in heavy section steel castings (Monroe and Huff, 2010) and in the control of nodule and cell count in ductile irons (Fras and Lopez, 2010) The original experiments of Chvorinov, first reported in English in 1938 and then later detailed in Czech (1951-1953), wisely covered a wide range of sand-casting sizes in low-carbon steel. They included plates ('Deska') covering sizes from 10x400x400 to 200x1800x2400 cm and included a large very large casting ('Sabota o vaze') of 65 tonnes ! Also included were the results of Briggs for some spherically shaped castings, as well as cylinders ('Valec'). After the original publication of Chvorinov's work, a number of objections, mainly by practitioners, arose saying that the index of two in the proposed rule just did not work in practice. Thus in 1952 the writer was tasked with looking into this proposed relationship, in particular into the extent the mold material and casting shape affected matters. This research took place under the guidance of his doctoral supervisor, Dr. Voya Kondic and a Technical Sub-Committee of the
290
Institute of British Foundrymen (Now the Institute of Cast Metals Engineers, University of Birmingham, UK.
ICME)
at
the
Interestingly, a prominent member of the Sub-Committee (TS46) was Mr. Ronald Ruddle, author of the classic text 'The Solidification of Castings' (1957). The study study was both experimental and to some extent analytical. Several of the findings were reported in TrAFS in 1959 (Berry, Kondic and Martin). Although the experiments performed do not fall into the 'clever' category, they can be regarded as distinctly useful and also of relevance to contemporary developments in certain aspects of mold chilling power, both as far as cooling fins and ablative casting are concerned. In addition to examining mold chilling power, the study looked at casting size and end-effects. The aspects studied, especially end-effects, had probably led to some of the objections raised by practitioners to the original Chvorinov relationship. Much of this work is to be found in Berry, Kondic and Martin, 1959 The interaction of casting size and mold moisture content was of particular interest, both a drying-out effect seen at heavier casting thicknesses and an incredible enhancement of chilling power at small thicknesses. Perhaps a review of some unpublished experiments and just how they led to the development of new ideas, thus illustrating the pacing effect referred to earlier, would be appropriate. Some Useful Experiments And Their Consequences In the abstract to this paper the writer touched upon the importance of validating experiments to justify computation based conclusions and the pacing effect this has on theoretical and computational progress. It is worthwhile to point out that two of the earliest large scale investigations of computer applications to solidification problems - the NSF funded CADCAST program - led by Prof. Robert Pehlke and the writer, together with the previously conducted AFS funded program at the University of Michigan (UM) contained experimental validation at every stage, either from the thermal analysis of castings poured at UM, or by members of the AFS monitoring committee. (See progress reports contained in Trans. AFS) This type of Tnhouse' validation, which represents an ideal situation, is hardly seen today. Returning to the question of the early industry based objections to the Chvorinov rule, it appeared to the writer that many of them clearly arose from end, or corner-effects. It will be recalled that the derivation of the rule was based on the assumption of semi-infinite planar heat flow occurring during the solidification of pure metals or short freezing range alloys. Ruddle had also recognized their importance and had provided a partly experimental, partly analytical method of accounting for effects on solidification time (Ruddle and Skinner, 1951). The writer subsequently used this 'corner-correction' to indicate why the rule was so often refuted, itself an example of experiments pacing theory. Figures 1 to 4 show the results of experiments employing the Ruddle/Skinner correction to the surface area for two eutectic alloys ( 11% Si-Al and 8.5% Al-Cu). This in turn inspired later work of an analytical or computational nature in the nineteen eighties. The PhD dissertation of C.S. Wei (Georgia Tech, 1982) examined how a combination of thermal
291
properties, as well as the included angle of the corner affected the strength of any such correction. One important conclusion was that corner effects were not as strong in low -carbon steel castings as for those in pure aluminum. A possible ramification of this is the relatively small deviation of experiment from prediction in the original Chvorinov plot. A subsequent development, in this case where theory was pacing experiment, led to the design of a test casting to demonstrate the effect of included corner angle. That test casting has since been used in teaching many 'generations' of undergraduates by the writer (See H. Huang, PhD dissertation, University of Alabama, 1982) Yet a further spin-off of this pacing sequence was the interesting concept of replacing the enmeshment of the mold with an interfacial heat-flux map which could be applied to the casting surface. Eisuke Niyama introduced this concept in 1977 applying this approach to a onedimensional semi-infinite mold. In 1981 the writer suggested that this might be applied to more complex geometrical features.(Berry, 1981) Subsequently, Hansen, Berry and Wei, in 1983 described their proposals for the Q-Method. Franklin, and Moosbrugger both provided experimental work elaborating on this method and its potential application (Master theses at Georgia Tech, respectively 1983 and 1985) It was envisaged that this approach of eliminating the thousands of nodes involved in the discretization of the mold would be of great benefit in telescoping computational time. Dantzig and his students at the University of Illinois neatly packaged this concept by designing software (SPIDER) that literally climbed around casting features determining their curvature and ascribing a suitable heat flux expression thereby. They also looked at the question of variation of mold chilling power (Dantzig et alia, 1985) Alas, progress in this general area of interfacial heat flux descriptors was 'Leap-frogged' by the rapid growth in the automated and virtually instantaneous enmeshment software together with the astonishing advances in computational speed. However, the extent of corner and curvature effects, which are driven by the local variation in the interfacial temperature gradient, are fascinating in terms of thermophysical properties, metal superheat, as well as geometry. Figure 5 shows the effect of time on the magnitude of the interfacial temperature gradient at a rightangled three dimensional corner as measured by Franklin for an aluminum alloy sand casting. (1982). The extent of deviations from the Chvorinov planar solution were also shown schematically, Figure 6. Moosbrugger, Berry and Wei, 1986 summarized much of the work in this area. Thus the question of corner and shape effects on the Chvorinov rule are now reasonably well understood. However, a more recent attempt to combine effects of both volume and shape more compactly into one simple equation are especially noteworthy and may well eventually displace the use of the original rule in approximate calculations of solidification time (Tiryakioglu, 1995). Turning to the question of mold chilling capacity and early predictions of solidification times, the sparsity of thermal properties at that time (1950s) did undoubtedly present problems. Chilling capacity was one aspect which the IBF sub-committee cited earlier expressed a particular interest.
292
Three methods were then in use to distinguish between thermal effects of molding media: (a)
(b) (c)
The Chvorinov Curve - Fit method, where the dimensionless temperature in the mold at a specific location is plotted against the quotient of the distance from the mold-metal interface divided by the square root of twice the elapsed time. A set of master curves was then used to assign a value of the mean temperature diffusivity . Figures 7 and 8 present data for a dried coarse-grained clay-bonded sand compacted to two distinct ramming densities. The ordinate of the plot is the dimensionless mold temperature (See Berry, 1954) An entirely analytical method (Russell-Gittus) where the thermal conductivity of the sand mold was calculated from the properties and fineness of its components. (See Berry, 1954) A mold-calorimetric method where measurement of the solidification time of a slab of known geometry was used, along with known thermal properties of the casting medium , to calculate the mean heat diffusivity of the mold. (See Berry, 1954)
Table 1 contains results for a variety of sands, in one case for two different compaction densities. The data are presented as mean values of heat diffusivity. Although the agreement between the results employing each method varied, the data points to the fact that compaction (i.e. ramming density) and grain coarseness are important parameters. These methods might well be reexamined in terms of current practice as well as a means of obtaining accurate temperature dependent of thermal properties for use in modeling. A further aspect of mold chilling and its effect on deviations from the Chvorinov rule which was of interest to the investigating committee was the effects of moisture and/or combustible material present in sand molds. It is worthwhile re-examining these observations in the light of the use of chill-fins, as well as of ablative casting and lost foam casting developments. Perhaps the most dramatic of the experimental results which surfaced at that time was the combined effects of section size and moisture content on chilling power. Figure 9 provides evidence of the power of the presence of moisture in enhancing chilling power of both a coarse grained sand (14-28) and one of a fine grained nature (H). The average fineness levels were 25 and 150 BSS respectively. This is especially true for the case of the small section-size casting (6mm) poured in the moist fine sand. The casting medium was again the 8.5% Al-Cu alloy. (Berry, 1954) This is of special relevance to the use of chill fins and to a lesser extent to ablative casting. Although water and molten metals do not mix, clearly under certain controlled circumstances the results are highly beneficial rather than explosive! At the other end of the scale, where heavy sectioned castings are involved, especially thicknesses greater than 50mm, a distinct 'drying-out' effect was observed. Figure 10 indicates how for castings solidifying at temperatures associated with cast irons and steels this effect becomes manifest. (Berry,1954) Clearly, the permeability of the molding medium would need to be taken into account in attempts to model this phenomenon. In addition to moisture driven effects, the writer was asked to look into the combined effects of moisture and combustible mold additions. At that time coal-dust additions were common in molding practice in the UK. On this side of the Atlantic the literature contains reference to what
293
were termed 'sea-coal' additions. This was a term which had crossed to the US in previous centuries but had virtually been forgotten in the home country where it had originated. The principal reasons were for such additions were related to the improvement of surface quality in iron castings. The exact mechanism involved was much debated at the time. Other combustible additions, such as wood flour were also common at this time. The general observation was that such additions controlled the appearance of the Veining" defect. The results of parallel tests embracing dry sand, green sand and a 'black', sand containing coal-dust each based upon use of the same molding and casting media were somewhat inconclusive. However, there was some indication that the chilling power enhancement brought about by the presence of moisture may have been more effective than convective effects of combustion of the coal-dust. (Berry, 1954) Later work employing an organically bonded silica sand into which aluminum A356 was poured, confirmed this suspicion that combustible materials may, in fact, provide an insulating effect. Figure 11 shows the interfacial temperature gradient (which drives mold chilling power) plotted against the reciprocal of root time, as in figures 6 and 7. Current simulations generally employ standard data bases of thermophysical properties, for example those values quoted by Fras and Lopez (2010). However, it is perhaps appropriate to ask whether the building in of data-shifts reflecting some of the above effects, especially where chill-fins are applied in green sand molding, might be undertaken. Discussion The opening statement of this presentation referring to the decline in the number of truly 'clever' experiments, or even useful ones, may have surprised members of the materials community, especially if one turns to TMS or AFS publications. However, there are now many members of our group as it exists today who were not trained in schools of materials science and engineering where experimental assignments are still commonplace. It is the writer's belief that our colleagues from other disciplines could benefit by emulating this tradition, if only in part. It has been pointed out that several pioneer investigations of the various factors influencing the progress of solidification and of heat flow into the mold were greatly enhanced by the pacing effect of experiment on theory and vice-versa. If we examine the trends in various university curricula we see that many engineering programs have de-emphasized laboratory activity and industry contact. Indeed, it has been alleged that the graduates of certain PhD programs 'don't even know how things are made' (Wyatt, Altan and Berry, 2009). Despite the praiseworthy efforts by Professor Ashby and colleagues in advocating a 'top-down' approach to how materials and shaping processes are intimately connected with the design process (Ashby, 2011), many young PhDs entering the ranks of academia have had little experience of designing and building components, or even of conducting experiments, quite apart from exposure to industry. Graduates of co-operative programs, in addition to those of certain technology curricula, are perhaps exceptions. This is not to deny the significant contributions to modeling of the casting process by computer scientists, mechanicians and graduates of similar disciplines. Indeed, this very series of symposia has been greatly strengthened by their presence and has further enhanced mutual understanding.
294
There is no reason why a similar process of bringing together individuals as undergraduates drawn from the various disciplines should not be encouraged. There is little doubt that the interuniversity teams involved in motor sports and human-power type projects benefit immensely in this aspect - the majority of our students invariably crave the 'Hands-on' approach to learning. This presentation has attempted to show how experiment and analysis, and hence computation, have paced each other and have resulted significant progress in materials processing and even actual product development. Conclusions There are many examples in the broad area of materials processing and eventual product performance, where the mutual stimulation exhibited in both experiment and analysis have proven valuable. A number of specific examples have been cited in the area relating to shape casting production. In particular the influence of casting shape and thermophysical property variation on freezing progress have been exampled. For this process to continue and grow it is essential that engineers and technologists are made aware early in their education and training of the important role played by 'clever', or at least 'useful' experiments. Acknowledgements At the end of a long career it is impossible to cite the very many individuals, mentors, colleagues, industry friends, funding agencies and students who have influenced that career, especially since there is a great likelihood of leaving out and thus disappointing them. 1 would, however, like to acknowledge the individual who not only assisted in putting together my dissertation, but also gave unflinching support and contributed boundless common sense and sound judgment throughout that career, my wife. Nomenclature A
Surface area
V
Volume
t
Time
T
Temperature
x
Distance
R
Volume/Surface area
R'
Volume/Effective surface area
b
Mean heat diffusivity, Vkpc
295
k
Thermal conductivity
p
Density
c
Specific heat
Subscripts s
Solidification
x
Location with respect to mold/metal interface, i
o
Initial, with respect to mold temperature
References Ashby, M. F., "Materials Selection in Mechanical Design," 4th Edn. (2011) Elsevier Atkins, A.,
" The Science of Engineering of Cutting," (2009) Butterworth-Heinemann
Berry, J. T. PhD Thesis, University of Birmingham, UK, 1954 Berry, J. T, Kondic,V. and Martin, G., Trans. Am. Foundry Soc, 67 (1959) 449-476 Berry, J.T., Private communication, NSF CADCAST Project (1981) Campbell, J., Castings 2 nd Edn. (2003) Butterworth-Heinemann Chvorinov, N., Proc. Inst. Brit. Foundrymen, 132 (1938-1939) 29-40, also Giesserei, 27 (1940) 177-186, 201-208,222-225 and Hutnicke Listy, 6(1951) 11-20 Dantzig, J.A. and Lu,S. C , Met. Trans. 16B (1985) 195-202 and Dantzig, J.A. and Wiese, J.W., Met. Trans. 16B (1985) 203-209 Einstein, A., Relativity, The Special and General Theory, (1916) Methuen and Co. Ltd. Farquar, R.B., Trans. Am. Foundry Soc, 29 (1920) 171- 201 Franklin, PH., M.S. Thesis, Georgia Inst. Of Tech. 1982 Fras, E. and Lopez, H., Int. Jnl. Metalcasting, 4 (2010) 35- 58 Huang, H., PhD Dissertation, The University of Alabama, 1994 Manganon, PL. and Heitman, W.E., in 'Microalloying '75', 1977, Union Carbide, NY Monroe.C. and Huff, R., Int. Jnl. Metalcasting, 4 (2010) 27-33
296
Kear, B.H., Leverant, G.R.and Oblak, J.M., Trans. Am. Soc. Metals, 62 (1969) 639 Subbiah, S.and Melkote, S„ Materials Science and Engineering, 474 (2008) 283-300 Moosbrugger, J., M.S. Thesis, Georgia Inst. of Tech., 1985 Moosbrugger, J., Berry, J.T. And Wei, C.-S., ASME Paper 86-WA/HT-95 (1986) Pellini, W.S., Am. Foundryman, 24 (November, 1953 and December, 1953) 58-61 and 62-71 Ruddle, R.W., 'The Solidification of Castings', 2nd. Edn. (1957) Inst. of Metals, London. Ruddle, R.W. and Skinner, B.F., Jnl. Inst. of Metals 79(1951) 35-56 Tiryakioglu, M., Private Communication to J.T. Berry, 1995 Wei, C.-S., PhD Dissertation , Georgia Inst. Tech.., 1982 Wei, C.-S., Hansen, P.N. and Berry, J. T., in ' Numerical Methods in Heat Transfer', Vol. 2, 1983 (Lewis.R., Morgan, K. and Schrefler, B.A., Eds.) 461-471 Wiley Winegard, W.C., 'An Introduction to the Solidification of Metals', (1964) Inst. of Metals, London Wyatt, J.E. and Berry, J. T. (This Symposium) Wyatt, J.E, Altan, T. and Berry, J.T, Paper presented at 2009 SE Regional Meeting of Am. Soc. of Engineering Education
297
Figures
Fig. 1 Analytical Solution (Solid line) With Experimental Results Solidification Time v. Volume/Surface Area,ll% Si-Al Plates in Dried Silica Sand Molds (Berry, 1954)
298
Fig. 2 As Fig. 1 but Solidification Time as a Function of Volume/Effective Surface Area, R'. (Berry, 1954)
299
Fig 3 Analytical Solution (Solid Lines) With Experimental Results Solidification Time v. Volume/Surface Area, 8.5% Al-Cu Plates in Dried Silica Sand Molds of Coarse (14/28), Medium (Bromsgrove, Natural) and Fine (H) Grain Size (Berry,1954)
300
Fig. 4 As Fig. 3 but Solidification Time as a Function of Volume/Effective Surface Area, R'. (Berry, 1954)
301
Fig. 5 Interfacial Temperature Gradient v. Reciprocal of Sq. Root of Elapsed Time for ThreeDimensional Wedge (All Right Angles). Circles are Experimental Values for Pure Aluminum in Dried Silica Sand . Solid Line is Chvorinov Solution For Planar Solidification. (Franklin, 1982)
302
Fig. 6 Schematic of Experimental Relationship of Figure 5, Summarizing Effects of Mold Geometry and Casting Superheat. Linear Relation is Chvorinov Solution For Planar Solidification. (Moosbrugger, Berry and Wei, 1986)
303
Fig. 7 Experimental Determination of Chilling Power Using Chvorinov Curve Fit Method, for Coarse Silica Sand(14/28) Packed at High Density (1.82 gm/cc) (Berry, 1954) Ordinate is Dimensionless Mold Temperature
304
Fig. 8 Similar Data to That of Figure 7, but for Same Sand With Low Density Packing ( 1.34 gm/cc) (Berry, 1954)
305
Fig. 9 Experimental Results Solidification Time v. Volume/Surface Area for 8.5% Al-Cu in Dry and Green Silica Sands of Coarse (14/28) and Fine (H) Grained Nature. Note the Dramatic Chilling Enhancement Present for Light Sectioned Castings. (Berry, 1954)
306
Fig. 10 Experimental Results for Solidification Time v. Volume/Effective Surface Area for Gray Iron Plates (CE 4.2%) in Green and Dry Silica Sand Molds. Note the Drying-Out Effect for Heavy Sectioned Castings. (Berry, 1954)
307
Fig. 11 Interfacial Temperature Gradient Data for a Three-Dimensional Corner (all RightAngles) for Pure Aluminum in Three Different Type Silica Sand Molds, Dry and Green ClayBonded, and Organically Bonded. Note Enhanced Chilling Effect of Moisture Presence and Decreased Chilling Effect of Presence of Organic Bond. (Berry, unpublished work)
308
Tables Table 1. Companson of Mean Heat Difrusivity Values for Various Molding sands as Determined by Three Separate Methods (Berry, 1954) Sand 14/28 (Coarse) 14/28 (Coarse) H (Fine) T (Fine, Natural) Notes:
Compacted Density (gm/cc) 1.82 1.35 1.40 1.19
1 31.0 20.8 18.3 15.8
Mean Heat Difrusivity Method 2 27.4 20.9 13.9 13.0
3 25.2 21.6 17.2 15.8
Figures given as originally converted to CGS units x 1000 Methods of Determination are :
1. 2. 3.
309
Chvorinov Curve-Fit Method Russell-Gittus Method of Calculation Mold-Calorimetric Method
Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
AUTHOR INDEX Shape Casting IV A
Antonio Maroto, J Ares, A Au, D
B
Berry, J Bozorgi, S Browne, D Byczynski, G
c
Campbell, J Ceschini, L Cockcroft, S Connolley, T Cuesta, R
D
Dai, X Dispinar, D Druschitz,A Duan,J
E
El-Sayed,M
F
Felicelli, S Flender, E
G
Gassa, L Green, N Griffin, J Griffiths, W Grupke,C Gueijman, S Guo,J
H
Haberl, K Hamilton, R Han,Z Heisser, C Hsu, F Hudak, D
157 207 21
157, 215, 281, 289 113 129 191
J
Johnson, M Jolly, M
K
71, 139, 181 121 21 87 157
Kandeil, A KangJ Kendig, K Khajeh, E Kneissl, C
L
265 173 199 21
Laplume, F Lee, P Leonard, C Lett, R Leung, A Li, J Lin, H Liu,B
149
M
29, 53, 157, 215 3
Mackay, R Maijer, D Mirihanage, W Morri, A
207 13 199 149, 225 233 207 103
N
Nastac, L Nath, R Nguyen, K Nordmark, A Nyahumwa, C
311
113 87 61 3 45 165
233 13, 265
149 257 199 37 113
273 87 233 157, 215 87 61 45 61
191 21, 37 129 121
249 233 21 173 139
o
Onsoien, M.
P
Pabel.T Phillion, A Puhakka, B
R
Reilly, C Rockett, P Romanelli, J
S
Sajja, U Salem, H Samuel, E San José, R Schumacher, P Schvezov, C Scott, S Senkov, 0 Senkova, S Siavashi, K Squatrito, R Stern, J Syvertsen, F
T
Tiryakioglu, M Todaro,! Tomesani, L Topping, C
w
z
.95
Zeng, B Zheng, C Zhu,J
113 87 79,241
13, 21 87 249
53 149 273 157 113 207 103 199 199 225 121 3 173
139, 165 121 121 225
Wang, L Wyatt,J
29, 215 281
Yang, W Yin, H
61 29
3 12
265 273 103
Shape Casting: The 4,h International Symposium Edited by: Murai Tiryakioglu, John Campbell, and Paul N. Crepeau TMS (The Minerals, Metals & Materials Society), 2011
SUBJECT INDEX Shape Casting IV A
A356 Wheel Al-Cu Al-Cu Alloys Aluminium Aluminium Alloys Aluminium Casting Process Aluminum Aluminum A356 Aluminum Alloy Aluminum Gravity Casting ASTM A757 C1Q Steel Autonomous Optimization AZ91
B
Bifilms Brittle
E
21 37 207 87, 149 225 265 29,273 157 61 45 249 3 215
Electromagnetic Pump Entrainment Eutectic Experimental Validation
F
Facets Fatigue Fatigue Potential Fatigue Specimens Filling Flow-Related Defect Four Point Bend Fractional Step Method Fracture Freckles
139, 173, 233, 241 181
Grain Transport Griffith Crack
Cast Aluminum Alloy 199 Casting 13, 103, 149,215 Casting Defects 139, 241 Casting Methoding 241 Casting Modeling 249 Casting Shape 289 Cellular Automaton 37 Ceramic Foam Filter 45 CET 129 Chilling Power 289 Classic Nondestructive Testing 233 Columnar-to-Equiaxed Transition 207 Confidence Interval 165 Control Arm 215 Cracking Susceptibility Coefficient 113 Cracks 71,249 CRIMSON 265 Critical Gating Velocity 45 Curvature 257
Defects Degassing Dendrite Growth Distortion Double Oxide Film Defects Ductile Ductile Cast Iron
139 181 139 121 3 21 157 53 181 53
G
C
D
157 13 37 289
H
HAZ Heat Treatment High Pressure Die Cast Hot Cracking Index Hot Cracking Susceptibility Hot Tearing Hot Tears HPDC Hydrogen Hypothesis Testing
129 181
71 3, 257 191 113 113 103 249 273 121, 173 165
I
In-Mold Melt Treatment In-Mold Thermal Analysis
L
13, 181 173 29 3, 257 149 181 95
LCS Waterjet Entry Edge Components Lost Foam Casting LPDC
3 13
95 95
249 225 21
M
Machinability Machining Sequencing Macrosegregation Magnesium Mathematical Modeling Mechanical Properties Melting Mesh Adaptation Microporosity Microstructure Modeling Modeling and Simulation Modelling Mold Design Molecular Weight Mould Filling
N
Naturally Pressurized Fill System Ni Superalloys
o
Oxide Bifilms Oxide Film Oxides
P
Permeability Porosities Porosity Precision Sand Prediction of Macro-Shrinkage Process Compensated Resonant Inspection
R
Residual Stresses Rheocasting Runner System
S
Sand Mold Printing SEED Process Semi-Solid Semi-Solid Casting Simulation
Single Crystals Solidification Solidification Under Pressure Squeeze Casting Steel Castings Stress Structural and Corrosion Parameters Super Duplex Stainless Steel
281 281 53 215 21 149, 225 265 53 61, 121 3, 95, 215 3, 29, 103 61 13 249 225 21
T
Tensile Properties Terminal Freezing Range Thermophysical Properties Turbine Blade Casting
u
Ultra-High Strength..
79 71
w
Water Analogue Experiment Weibull Weibull Modulus Weibull Statistics Welds
79 265 13, 233
X
X-Ray Microtomography X-Ray Tomography
37 249 87, 173 191 249 233
281 273 45
249 273 273 273 3
314
71 3, 129 199 61 241 3 207 79
191 113 3 257
199
21 157 165 191 71
37 87