TMS ONLINE | TMS PUBLICATIONS | SITE MAP An Article from the January 2002 JOM-e: A Web-Only Supplement to JOM

VISIT THE JOM COVER GALLERY JOM MENU

INTRODUCTION

Numerical simulation is widely used and accepted in manufacturing as a way to reduce hardware prototyping and to improve the parts design and manufacturing processes. The automotive industry and others are increasingly using computer simulation to attain their design objectives. For example, one important automotive design objective is to reduce the overall weight of the vehicle for better fuel economy and, thus, reduced CO₂ emissions. Weight reductions have been obtained by using different materials in place of cast iron for certain components. Aluminum alloys, for instance, have been substituted for cast iron because they are typically one third of the specific weight of cast iron. Additionally, the use of cast components exploits a relatively inexpensive near-net-shape manufacturing route for parts with complicated three-dimensional (3-D) shapes. This paper focuses specifically on the problem of replacing a cast-iron component with a lighter aluminum-alloy equivalent and modifying the component shape and methoding in order to optimize properties/casting quality for the reduced weight.

While the benefits of replacing cast iron with aluminum alloys are clear, a successful replacement requires the solution of several very different design problems. This is particularly true when dealing with structural components (i.e., parts that must play a load-bearing role within the vehicle). While the financial incentive for using aluminum alloys is great, predicting the final performance of cast aluminum-alloy parts is very difficult. This problem is often compounded by the uncoupled nature of the overall design procedure. The initial design of the component shape usually originates from the automotive company, whose designers will have worked from a stress-engineering viewpoint. Typically, finite-element (FE) stress analyses are carried out to model the behavior of the component in different loading scenarios. However, the component supplier, often a foundry organization, usually carries out the design of the casting, including casting orientation, location of gates, runner bars, and feeders, but from a very different viewpoint (e.g., heat transfer, solidification and pattern/die design). Design for performance and design for manufacture are being carried out independently.

This segregation of the design process inevitably leads to problems. For example, from a stress-engineering viewpoint, thickening up a section of a component will lead to increased load-bearing capacity at that location. However, during casting, a thicker region will solidify more slowly and, for aluminum alloys, a coarser microstructure will result in lower mechanical strength. (Foundry specialists say the material on an aluminum alloy casting that has the greatest strength is the unwanted "flash," which solidifies very rapidly.) Also, problems with feeding and shrinkage defects may arise in thicker sections. Of course, a component design that worked well for cast iron may be entirely unsuitable for an aluminum-alloy casting anyway. From experience, an uncoupled design procedure can result in an expensive cycle of prototypes and testing that does not always end successfully.

Bearing these points in mind, a project is underway to redesign a generic automotive component considering both the stress-related and solidification-related design problems simultaneously. Several different specialists from organizations within the extended car-making enterprise have been brought together. The aim of the work is to take a part that was originally made in cast iron and redesign it as a gravity-cast aluminum-alloy casting. The methodology has been to alter the geometry of the component to improve the results of stress analyses and, at the same time, reduce the size of cross sections where possible, thus minimizing weight and encouraging more rapid solidification and superior mechanical strength. Concurrent with this, the castability of the designs has been assessed via simulation in an attempt to ensure controlled mold-filling behavior and progressive freezing toward feeders. This will help reduce the cost of manufacture by enhancing right-first-time design with lower scrap rates and efficient use of raw material. (At a later date in the project, real sand-cast prototypes of the original design and an improved design will be mechanically tested to assess the effects of the coupled-design procedure.)

At this stage, only a subset of the total number of variables affecting the design is being considered. Thus, variables such as residual stresses, the results of post-heat treatment, gas porosity, and microstructural considerations in terms of oxide inclusions and impurity-related effects, are beyond the scope of the current project.
With a range of physical phenomena to consider, the use of computer visualization is essential. It also serves as a useful illustrator of different phenomena to the relevant specialists from different disciplines.

SIMULATIONS

The 3-D geometries of the casting and associated methoding (such as gates and feeders) were all imported from surface-stereo lithography (STL) files and converted by the MAVISFLOW software¹ into solid models. The MAVISFLOW package is an FDM method package; a regular structured grid of cubic elements was used and individual element edge lengths for all simulations were 2 mm. Three component designs were modeled, each with a different methoding approach. Animations 1, 2, and 3 show the three designs. Design 1 is the original component design (originally to be made in cast iron), with a methoding system appropriate to an aluminum-alloy casting. Design 2 has a modified component design (derived from numerous stress analyses) with a slightly different methoding system, and Design 3 is yet another modified component with a totally different methoding system. In all cases, the castings were assumed to be gravity sand castings with a filter incorporated into the runner system. Although the designs are presented here simultaneously, they were, in fact, developed as part of a progressive sequence.

Animation 1. A mold-filling sequence for design 1, the original component design with a methoding system appropriate to an aluminum-alloy casting.		Animation 2. A mold-filling sequence for design 2, a modified component design with a slightly different methoding system.		Animation 3. A mold-filling sequence for design 3, a modified component design with an entirely different methoding system.

First, the filling and solidification of the designs were simulated using MAVISFLOW. Details of the simulation parameters are given in Table I. Using a custom-written interface, data from these simulations can be exchanged with the FEMAP (pre- and post-processor)² and ANSYS (FE stress analysis)³ software systems. The output from these simulations is visualized using MAVISFOW and FEMAP to give:

• A visualization of the filling of the mold.
• An isochronal freezing map showing the progression of freezing (from experience, for the LM25/A356 alloy used, the time to 80% solid is selected).
• A predicted map of possible macro-shrinkage defects.
• A map of local freezing times (the time taken between start and end of primary aluminium phase solidification). From this map, a secondary dendrite arm distribution (SDAS) was predicted using a simple empirical relationship, Equation 1.⁴
  
  

  Table I. Parameters Used in Casting Simulations
  	Aluminum Alloy	Sand Mold
  Thermal conductivity (W/m/K)	160 (solid) & 60 (liquid)	0.75
  Specific heat (J/kg/K)	1,150	990
  Density (kg/m3)	2,685	1,493
  Kinematic viscosity (m2/s)	4.0 ´ 10-6	—
  Slip condition	—	0.7
  Latent heat of fusion (J/kg)	4.35 ´ 105	—
  Pouring temperature (K)	968	—
  Liquidus (K)	888	—
  Solidus (K)	840	—
  Freezing model	Scheil	—
  Heat transfer coefficient to atmosphere (W/m2/K)	25	25
  Heat transfer coefficient casting to mould (W/m2/K)	100	100
  Ambient temperature (K)	298	298
  Initial temperature of mold (K)	—	298

Using ANSYS, elastic-stress analyses of different component designs is carried out, where the mechanical properties of the component at any position is related to the SDAS via another empirical relationship, Equation 2, for A356 (LM25) alloy in the T6 precipitation-hardened condition.⁵

The loading case, which was one of several supplied by industrial collaborators (Land Rover), will be used for real mechanical tests at a later stage in the project. The stress analysis combines information from both the stress-related and casting-related (SDAS) design approaches. The overall stress distribution, especially where stress is concentrated, can be seen clearly.

RESULTS

Animation 1, Animation 2, and Animation 3 show the filling patterns predicted for design 1, design 2, and design 3, respectively. Figures 1a, 1b, and 1c show the predicted isochronal-freezing plots for the time to 80% solid, used to visualize the overall progression of freezing for the three designs. Figures 2a, 2b, and 2c and Figures 3a, 3b, and 3c show the local freezing times and predicted SDAS distributions, respectively, for the three designs. Figure 4 shows areas that are predicted to be more likely to contain macroscopic shrinkage defects. Figure 5 shows the predicted nominal strength distributions for each design, and Figure 6 shows the Von Mises stress as predicted from the elastic FE analysis for a load case supplied by Land Rover. In an attempt to visualize both manufacture-related and service-related predictions, Figure 7 shows contour plots of the Von Mises stress divided by the nominal strength, and Figure 8 shows a plot of maximum principal stress multiplied by local freezing times.

Figure 1. (a-left) The isochronal freezing pattern for time to 80% solid for design 1; (b-center) Isochronal freezing pattern for time to 80% solid for design 2; and (c-right) isochronal freezing pattern for time to 80% solid for design 3.
Figure 2. (a-left) A local freezing time map for design 1; (b-center) local freezing time map for design 2; and (c-right) local freezing time map for design 3.
Figure 3. (a-left) The predicted secondary dendrite arm spacing distribution for design 1; (b-center) predicted secondary dendrite arm spacing distribution for design 2; and (c-right) predicted secondary dendrite arm spacing distribution for design 3.

DISCUSSION

In terms of mold-filling, design 1 and design 2 (Animation 1 and Animation 2, respectively), show fairly controlled flow patterns. To minimize the chances for either air or oxide-film entrapment in the body of the casting, controlled and quiescent filling is the main design objective at this stage. Introducing more choking into the runner system could provide further improvements. Ideally, in design 2, the flow of metal around the lower ring would meet in the center at the top of the ring. This is very nearly the case, but is offset slightly by the fact that more metal appears to enter through the left side of the in-gate. Again, this could be improved by choking the runner system. From a foundry point of view, it must be noted that the in-gate and feeders are located on curved surfaces, and this may cause fettling problems. For design 3, after exiting the filter block, the liquid accelerates along the runner bar, trapping air at the far end. The feeder taken directly from the top of the runner fills very quickly, and the fluid entering through the gates appears broken and turbulent. Thus, the running and gating system present in design 3 would not be recommended.

Figures 1a, 1b, and 1c show that directional freezing toward feeders is attained for all designs, although the large feeder in design 3 means longer overall solidification times. Figures 2a, 2b, and 2c show the local freezing times where, again, the longest times are associated with design 3, and Figures 3a, 3b, and 3c show the predicted SDAS distributions. While the maximum values for local freezing times and predicted SDAS are both from design 3, these values are due to the presence of the large feeder. The distribution of SDAS in that part of the casting that represents the final component is most important. Taking this into account, design 2 and design 3 have the smallest SDAS values predicted, thus making them more attractive than design 1.

Figure 4 shows that design 2 and design 3 are less prone to macro-shrinkage within the component areas of the casting.

Figure 4. Macro-porosity predictions (assuming a 3% volume contraction on solidification), from left to right for designs 1, 2, and 3, respectively.

Figure 5 is illuminating in that it shows the predicted nominal strength for the three different designs. In design 2 and design 3, cross sections were reduced, where possible, to promote more rapid freezing and, hence, a finer microstructure. Design 3, and, particularly, design 2, would likely produce components with superior mechanical properties than the original design.

Figure 5. The predicted nominal strength, from left to right for designs 1, 2, and 3, respectively. (Linear scale from 100 {bottom} to 180 {top} MPa.)		Figure 6. The Von Mises stress under load, from left to right for designs 1, 2, and 3, respectively. (Linear scale from 0 {bottom} to 150 {top} MPa.)
Figure 7. A plot of Von Mises stress divided by nominal strength, from left to right for designs 1, 2, and 3, respectively. (Linear scale from 0 {bottom} to 1 {top})		Figure 8. A plot of maximum principal stress multiplied by local freezing time, from left to right for designs 1, 2, and 3, respectively. (Linear scale from 0 {bottom} to 40 {top} GPa.)

Figure 6 shows the predicted Von Mises stress distribution from an elastic FE analysis of one of the loading cases supplied by Land Rover. Each design has been subjected to the same loading condition. Areas of stress concentration can be seen clearly. It should be noted that the component shape in design 3 was devised from inspection of the stress-analysis results predicted from design 2. Figure 7 attempts to combine casting-related and stress-related information and shows a plot of the Von Mises stress under load divided by the nominal strength. Although the possibility of plastic deformation has not been included at this stage, the plot is interesting in that it highlights areas that are experiencing stresses close to or even surpassing the nominal strength. From this plot, design 3 appears to be superior even though it did not perform as well in terms of predicted nominal strength. The improved geometrical design has reduced overall stress levels at key points (Figure 6). Finally, Figure 8 is another attempt to combine casting and stress-related information. This is a plot of maximum principal stress under load multiplied by local freezing time. A large value of either of these terms is undesirable; their product emphasizes areas that are either at high stress or have inferior mechanical properties or, worst of all, both. It is interesting to note the similarities between Figure 7 and Figure 8.

CONCLUSION

In conclusion, the use of visualization techniques has facilitated a more complete design process for a generic automotive component. The consequences of design alterations, for both manufacture and in-service performance, can be considered simultaneously. As mentioned previously, this work is part of an ongoing project, the next phase of which will involve casting and testing of the original design and an optimized new design. From the data available to date, the improved design is likely to be the casting methoding system from design 2 (Figures 2a, 2b, and 2c) used to produce a component shape similar to that in design 3 (Figure 8, right side). The project is also serving another useful purpose in that automotive designers and casting designers are obtaining a better mutual understanding of each other's roles and challenges in the extended car-making enterprise.

ACKNOWLEDGEMENTS

This work is part of EPSRC "INTCAST" project GR/M43234. The help and cooperation of Land Rover and Alumasc-COPAL Ltd. is gratefully acknowledged.

References

1. www.alphacast-software.co.uk
2. FEMAP is produced by SDRC, 411 Eagleview Blvd., Suite 108, Exton, PA 19341; (610) 458-3660; fax (610) 458-3665.
3. ANSYS, Southpointe, 275 Technology Drive, Canonsburg, PA 15301; (724) 514-3304; fax (724) 514-9494.
4. M. Al-Meshhedani (PhD thesis, University of Wales, Swansea, U.K., 1992).
5. A. Wickberg, G. Gustafsson, and L.-E. Larsson (AB Volvo Technical Department), SAE Technical Paper Series 840121 (Paper presented at the SAE Int. Congress and Exposition, Detroit, MI, February 27-March 2, 1984).

For more information, contact S.G.R. Brown, University of Wales, Swansea, Materials Engineering, Singleton Park, Swansea, SA2 8PP, U.K.