Process modeling solves die design problems, saves more than $100,000. (Case History).

A job for a diecast impeller for an automobile transmission had bounced around several foundries over the last few years because nobody could produce it at an acceptable scrap rate. Too many of the production castings contained voids and porosity due to air trapped between two metal flow fronts.

Up to a challenge, Amcan Castings Ltd., Hamilton, Ontario, Canada, decided to step in and take the job.

The porosity specifications on the part were very demanding, primarily because voids could cause leakage from one chamber of the transmission to another or weaken the part in a way that might cause it to fail. Amcan's responsibility and challenge was to design a die that would consistently meet specifications.

The traditional approach would be to rely on engineers' experience and guesswork to design the die--trial and error. The problem is that the only way to determine whether the design works or not is to build the die and test it. Typically, the die needs to he modified a number of different times at a cost of perhaps $5,000-10,000 per iteration before quality reaches acceptable levels. Boris Lukezic, Amcan engineering manager, estimated that the company would have had 3% scrap due to porosity if it followed traditional trial and error methods.

In an effort to find a better way to validate their designs, Amcan engineers have been experimenting over the past several years with computational fluid dynamics (CFD) technology in an effort to evaluate design concepts on the computer before building a hardware prototype and obtain detailed information on why a proposed design didn't work so that they could move more quickly to a solution. CFD involves the solution of the governing equations for fluid flow and heat transfer at thousands of discrete points on a computational grid in the flow domain. When properly validated, a CFD analysis allows engineers to look inside the die and determine the exact position of the flow front at any point in time, as well as the temperature and pressure of the metal at any point in the die.

Amcan engineers selected Flow 3-D CFD software from Flow Science, Inc., Santa Fe, New Mexico, as its simulation tool. The software has the ability to predict flows with free surfaces during die filling. It uses the volume of fluid (VOF) method to predict free-surface fluid motions, surface tension and other flow complexities. In particular, this package provides algorithms that track sharp liquid interfaces through arbitrary deformations, applying the correct normal and tangential stress boundary conditions--an accuracy feature that distinguishes it from other CFD programs.

Amcan had never built an impeller before so its initial design concept for the 380 aluminum diecasting was based on guesstimates. The initial design had six runners oriented radially and spaced around the periphery of the part. The runners split off from a main branch about halfway between the molten metal source and the die.

Looking at the filling patterns in the simulation results, Lukezic saw two flow fronts converging with air pockets trapped in the middle with nowhere to escape. This indicated that porosity would have been a major problem if the initial design had been built. Instead, he modified the gate position and feeder orientation in an effort to obtain a more favorable flow pattern as indicated by the analysis. The four gates were concentrated in one section of the part and feeders were all pointed in the same tangential direction in an effort to initiate flow through each gate simultaneously, creating a circulating flow pattern that would fill the entire part with a single flow front. This iteration showed a significant improvement, indicating that he was moving in the right direction.

Lukezic kept tinkering with the design and on the 13th iteration, he was satisfied that he had achieved a design that would be nearly or completely porosity free. The company built the die according to the simulation results and found that from the very first sample the porosity of the parts produced was virtually zero.

"There's no doubt that if we had used traditional trial and error methods, we would have had to settle for a scrap rate due to porosity of at least 3%," Lukezic said. "Over the life of this part, that would have cost us hundreds of thousands of dollars. Instead, in a matter of weeks we were able to optimize the design of the part and virtually eliminate porosity. Based on these results, we came to the conclusion that every die that we build should be simulated in advance."

Select No. 035 at www.moderncasting.com/info