GM says AI cut some vehicle simulations from 15 hours to 1 minute

In short: General Motors says it is using AI to speed up computer testing for new vehicles, cutting some simulation runs from about 15 hours to about one minute.

What's going on

General Motors is leaning more on “virtual” engineering, which means testing ideas on computers before building real parts. Sterling Anderson, GM’s chief product officer, described this as a shift from old, slow trial and error to faster work done in software.

A key change is how quickly GM can run certain engineering checks. Anderson said some FEA runs now take one minute instead of 15 hours. FEA (finite element analysis) is a way to predict how a part might bend or break under force, like stress testing a bridge design before building it.

GM is also combining more pieces of the car into the same virtual test. Jason Fischer, who leads virtual integration engineering at GM, said the company can model hardware and software together, including sensors and control computers, and run many variations quickly. He gave an example of running a digital version of a handling test, like Consumer Reports’ emergency swerve test, and changing conditions like road grip without needing a physical track.

The same approach is also used beyond car bodies and handling. GM says it can use these tools to improve crash design earlier, speed up work on heating and cooling systems, and build “digital twins” of factory assembly lines, meaning a virtual copy of a real production line used to find problems before equipment is installed.

What to watch

If these faster simulations match real world results, GM and other automakers may build fewer physical prototypes and test more ideas in software. That could shorten development timelines, but it also raises a practical question, how companies will prove their virtual tests are accurate enough for safety-critical decisions.

Source: Arstechnica