Developing new EV battery technology takes years of research. General Motors, for example, says it’s put a decade of effort into lithium manganese-rich, or LMR, batteries, which aren’t yet in any of their vehicles on the road.

But recently, the automaker says it made a big leap in that work. Using computer modeling, GM was able to virtually test more than 300,000 LMR battery pack configurations, advancing from first concept to formal approval in under a month. 

That represents more than 300 million CPU hours, which would take a typical computer 34,000 years, while running 24 hours a day, to complete, according to GM.

GM is leaning into artificial intelligence and virtual simulations to develop its vehicles. That not only speeds up the time to do testing, per the company, but allows it to test more options, rather than picking the best of just a few configurations. 

Those 300,000 formulations, for example, looked at different chemical combinations, temperature cycling, and other approaches to creating a battery cell, Sterling Anderson, GM’s chief product officer, told Fast Company. 

“If we had not had the luxury of some of these tools, we would have had to sample more sparsely from what was possible,” he says. 

GM is bringing this AI approach beyond battery testing and into all areas of its vehicle development, with the aim of slashing the time it takes to develop a car in half. 

Typically, cars are developed within a four- or five-year time frame. That allows a “limited number” of all kinds of tests, Anderson says, from wind tunnel to crash tests.

Source link