From the hospital to the car plant: What is GM doing with CT scanners?

More and more, we're seeing imaging technologies and machine learning showing up in automotive applications. It's usually to diagnose some kind of problem like quality control, although not always—the camera-based system by UVeye that we wrote about a few years ago made news recently after Hertz started using it to charge renters for things like scuffs on hubcaps. I have fewer concerns about customer abuse with General Motors' use of CT scanning, which simply seems like a clever adaptation of medical technology into another industry.

Ignore, if you can, GM's business decisions. Maybe you're upset because it killed your favorite brand, changed the shape of the Corvette headlights, or abandoned Apple CarPlay. There are many valid reasons, but none change the fact that the company's engineers are quite creative. (That's probably why it stings so much when the company starts hacking things up.)

GM first turned to X-rays as a way of doing two-dimensional quality control on castings during the development process, according to Ed Duby, manufacturing engineering executive director at GM. "Much like the application to people, when you think about X-ray and CT scan, it's really trying to diagnose something without having to go into surgery. We kind of want to do the same thing with our castings," Duby told me.

"What we're trying to do is use this tool to look at where the defects are, to change our processes so that when we build it, we don't have to deal with the defects and then the ramp plan that comes with it," Duby said. Destructive testing used to be the name of the game—slicing a casting into bits to see if there were any problems in it.

But that was a laborious and time-consuming process. Using CT scanners to 3D image parts during development has improved first-time quality by 90 percent, Duby told me, and has cut development time by a third.

A 10-speed transmission casing rotates inside the CT scanner, evaluating the part three-dimensionally — A casting for GM's 10-speed transmission, rotating in the CT scanner. Credit: General Motors

The datasets have value beyond quality control checks. "The less we have to do physically, the more that we can do with math modeling, the more that we can do with simulation, the more data that you collect," said Mike Trevorrow, senior vice president of global manufacturing at GM. "We're a company who's been doing this for 100 years. We wish we would have collected all the data of all the things we've tried, which would then enable newer technologies like AI and other things to give us even better predictive capabilities."