The most obvious verification of a simulation is to compare the flow-front position from analysis results to actual short shots. The widespread expectation is that if the short shots and the flow fronts in the analysis don’t match up, then the analysis is wrong. But wait a minute . . . were the simulation and the short shots conducted under the same conditions? More often than not, many of the "bad" analysis results come from overlooking some critical software model assumption or from assuming process conditions. As Giorgio Bertacchi, president and founder of Plastics & Computer International, put it, "Your analysis was perfect, but your perception of reality leaves a lot to be desired."

Poor resin data is often cited as the reason why an analysis did not match reality. While this is certainly a factor, its importance is often exaggerated because it is poorly understood. Remember that resin varies lot to lot and the variation increases if it is colored, contains regrind, or has other additives. Moisture content can further modify the rheological properties. The laboratory data from which the analysis data is generated comes from analyzing a small sample from one batch of resin and may or may not be a representative batch of that resin.

A much more important factor is engineering changes. All too often engineering changes are made and not reflected in the CAD database provided for the analysis. Changes in thickness, which can have a tremendous effect on the flow of material in the part, are frequently slight and certainly not readily visible in the CAD database. While obvious, make sure your analysis is representative of the mold used during verification.

Case Study #1

The most challenging variable is to understand the effect of plant floor practices, capabilities, and human factors on production. In one situation, all the engineers and plant managers met to review the analysis results before the first mold tryout, but they did not include the setup man or the machine operator. After the meeting, they asked the setup man and operator to mount the mold and set up the process per the analysis recommendations. The first parts were terrible and the cycle time had to be extended significantly because it was determined that the parts were not sufficiently cooled at ejection time. Even with the longer cooling time there were still problems. The engineer who did the analysis kept looking at all the process settings at the press to try to understand why his predictions were so far off the mark. The injection times and speeds were correct, the pressure settings were correct, and the chiller temperature was set correctly. So why was there such a difference?

After several hours, the engineer happened to notice that one of the coolant line valves from the chiller to the mold was turned off. He asked the operator to open the valve and all of a sudden, they started getting good parts and were able to reduce the cooling time to the predicted value.

Case Study #2

Researching the reason behind a discrepancy between an analysis and mold trial is a valuable tool to better understand the true production process capabilities and what is happening in the tool, as well as to find a solution.

For example, a large part was being molded with six hot drops. The cycle time was much longer than predicted and there were problems meeting tolerances in some areas of the tool. By verifying the assumptions used in the analysis, it was determined that one of the cooling lines had been left out of the tool. The analysis was conducted assuming a uniform temperature mold, but there was actually one area (the one with tolerance problems) that upon inspection was almost 100 deg F hotter than the rest of the mold. This area also logically caused the need for a longer cycle time.

Case Study #3

In another case, a molder was having problems filling a part with a thick frame and thin center section. The LCP resin was race-tracking around the outside frame and not filling the thinner center section. The filling analysis, conducted under "ideal" processing conditions, showed very little race-tracking and part filling. The first reason suggested for the difference between reality and the analysis was that the mold was poorly vented (filling analysis packages assume ideal venting). This theory was dismissed by the moldmaker, who had verified that there were more than sufficient vents.

Additional iterations, under different conditions, demonstrated that the analysis would show severe race-tracking or short parts similar to what the molder found if the mold was too cold and the filling time too long. In reviewing these results with the molder, it was discovered that he was not molding the parts on a machine capable of filling the mold fast enough and controlling the switchover to holding, and that he was using water rather than hot oil cooling. The latter resulted in a much colder mold than desired. This theory was further verified by the fact that the few times that he was able to fill the mold, the thin center section was so overpacked that he had a lot of trouble getting the parts out of the mold.