Artificial intelligence is helping the MBTA to predict problems before they happen. They’re doing that by using oil samples taken from diesel Commuter Rail locomotives. And it’s revealing some very big clues.
“Tells a lot about what’s going on inside,” said Ryan Coholan, the chief operating officer at the MBTA. “It’s sort of a molecular black box.”
That could mean fewer trains breaking down and could also reduce delays.
“That’s really the end game,” said Coholan.
NBC10 Boston was given access to the lab inside the Somerville maintenance facility. That lab is where the oil is analyzed.
The MBTA collects samples from each locomotive and brings it to a machine. It looks for abnormal levels of elements like copper, iron and zinc.
“And now it’s looking for change in those values,” said Coholan, while pointing to the computer. “That’s where the AI component really comes in.”
He said the computer can find a problem instantly, something that could take two to three hours to do manually.
“AI really looks at patterns so we can use machine learning to review those patterns constantly across the entire fleet,” said Coholan.
Tom Davenport, a Babson College professor and expert on all things AI, studied the MBTA’s oil analytics program.
“It’s much better to know ahead of time that something is going to break down so you can fix it without stranding passengers,” he said.
Davenport co-authored a book, using the T’s program as an example.
He said this is a way to utilize AI in complement with keeping human jobs. And he believes there is a big return on investment.
“It would be quite difficult for a human to do it on a large scale,” said Davenport. “This is a great example of using an innovative technique to make things better for passengers.”
The MBTA plans to use this same AI technology on its subway trains and buses, and they’re looking at other way to use machine learning to better predict failures for things like track switches.
“It protects our investment,” said Coholan.
Jeff Saperstone
Source link