NEW YORK — The Metropolitan Transportation Authority has begun a pilot program with a division of Google that uses smartphones to help detect potential subway track defects.
The program announced Thursday, Feb. 27, builds on TrackInspect, a prototype developed with the Rapid Innovation Team at Google Public Sector. Tested on the subway A line of New York City transit, that effort fitted Google Pixel phones in off-the-shelf plastic cases onto R46 subway cars to capture vibrations and sound patterns through sensors with an attached microphone to find locations in need of preventative maintenance.
The sound and vibration data is sent to cloud-based systems using artificial intelligence and machine learning to develop predictions. New York City Transit track inspectors then inspect the locations to determine if there is an issue, providing feedback to help develop the modeling. The AI involved also allows inspectors to ask questions about maintenance history and other details, receiving answers in conversational language.
“By being able to detect early defects in the rails, it saves not just money but also time – for both crew members and riders,” New York City Transit President Demetrius Crichlow said in a press release. “This innovative program – which is the first of its kind – uses AI technology to not only make the ride smoother for customers but also make track inspector’s jobs safer by equipping them with more advanced tools.”
The initial pilot involved 335 million sensor readings, a million GPS locations, and 1,200 hours of audio. That data was combined with existing NYCT information on track non-conformities, and compliments information gathered by MTA track geometry cars.
“How it works is the prototype sends a soundbite or noise clip showing heavy vibration or noise, and then our inspectors follow up by walking the track and verifying any issue found,” said Robert Sarno, NYCT Department of Subways assistant chief track officer. “We then compare that with whatever we find to teach the device noise and decibel levels and then work from there. That’s how we are able to instruct the prototype on what are normal sounds and vibrations, and what are not, and move along through the process.”
Brent Michell, Google Public Sector vice president go-to-market, said the pilot program identified 92% of defect locations found by track inspectors, “illustrating that enhanced data analysis can hep expedite problem identification and resolution to improve railway reliability.”
The project began as a proof-of-concept prototype developed by Google Public Sector exclusively for the MTA at no cost to the agency. Along with this program, the MTA has released a Request for Expressions of Interest for other companies that may have sensors or analytical capabilities that can be added to the system.
If it proves reliable then that will be a time and money saving.
What about wheel defects or flat wheels? Can it differentiate between a track defect and a bad wheel?