Editor's Note: Model accuracy must improve

In the December/January issue, I published an interview with Doug Marinaro, VP of MTS's software and consulting business. During our conversation, the most surprising thing Doug said was, "Based on our discussions with automotive engineering organizations, we estimate that CAE models simulate only 15 to 20 percent of the behavior observed on the proving ground and on the road."

This is incredible to me. The economic pressure to reduce prices—which is in turn forcing manufacturers to reduce development and test costs—is pushing manufacturers to do more simulation and virtual testing. But if the models can only accurately simulate 15–20% of a component's or vehicle's behavior, aren't we putting too much faith in these simulations?

Obviously, something has to change. We must either slow down this move to virtual testing or work harder on improving the models. Since the former is unlikely to happen, given the economic pressures, we must find ways to improve the models.

The most obvious way to do this is to make better use of the physical test data that a typical test program generates, but too often this data never makes it back to the design department. The first company to come up with a way to do this quickly and economically will make a lot of money. If companies find it worthwhile to simulate and run virtual tests even with model accuracies as low as they are, think how much more valuable the simulations will be when models are 80% accurate. That's value people will pay for.

Contact Dan Romanchik at editor@aatr.net .