SBB uses AI to inspect 2,000 kilometres of track each month

Swiss public broadcaster SRF reports that two SBB diagnostic trains use artificial intelligence to inspect around 2,000 kilometres of railway track every month, identifying about 1,500 mostly minor defects.

The trains travel at 120 kilometres per hour on a new line where speeds of up to 200 kilometres per hour are permitted. Cameras and measuring equipment record the tracks and track bed as the trains move.

The AI can detect defects on the top of a rail and problems with its fastenings, such as a bent clip. The data is analysed at a computing centre in the depot rather than on the train itself.

SBB began using AI to detect track defects 10 years ago. The system initially produced many errors, but the company now says it has reached the required level of reliability and makes the work more efficient.

AI will not replace human workers, however. It can identify only certain defects in specific parts of the railway. Objects outside the track section, such as overhead power masts, are not detected. Trained staff are therefore still needed to identify defects and carry out repairs.