Jim Tung, MathWorks Fellow.

Q How important is it to use physical test data to improve simulation models?

A It's very important. When tuning system model parameters, for example, you want to make sure the model's behavior represents actual system behavior. You can also use physical test data as inputs to a simulation to see how a design will react to real-world conditions.

I think there is sometimes a mistaken assumption that models used in model-based design are disconnected from real data. In reality, the models are closely tied to experimental and test data throughout model-based design.

Q What hardware and software is required to successfully use physical test data to improve simulation models?

A Certainly, the test system must first of all provide accurate measurements. Next, the data modeling and analysis software must easily access both live and archived data and have capabilities for system identification and data reduction. The simulation tools should model both the system and the environment, which helps the design of testing scenarios. Those same simulation tools must be able to automatically generate production-quality implementations. That dramatically increases the benefit of reusing the validated models.

Q What's the biggest mistake companies make when trying to feed test data back into the design process?

A Probably the biggest mistake is not doing it. When a company doesn't validate and maintain the models after implementation, they may have a good design but they lose the benefits of reuse.

Q What companies are having the most success doing this?

ATakata is one company that is successful in using test data to improve its simulation models. The company developed and verified an air-bag system controller model that uses a new magnetic sensor technology for crash detection. Takata heavily relied on data from vehicle tests to understand what occurs in a vehicle collision and then used that data when developing and verifying the control algorithm.

Q How do you see the automotive and aerospace industries using test data to feed design? How are they similar or different?

A Automotive and aerospace both use high-fidelity models in design, but they sometimes differ in how they use test data. Aerospace companies normally develop a simulation and then run a low-risk ground-based test to verify the model. They use that data to refine the design model and then perform captive flight tests and finally free flight tests. They increase the model fidelity as they gain more insight about factors such as atmospheric conditions and turbulence.

For automotive companies, particularly those developing powertrains, it is critical that they meet specific physical design specifications such as engine emissions. Feeding back test data is integral to the design process and meeting those specs.