AI in Public Transport

Hitachi Rail has the need for rapid identification of infrastructure failure causes and the provision of effective recovery instructions during railway disruptions. Traditional methods, which depended heavily on the experience and knowledge of dispatchers, were often insufficient for managing promptly rare or complex incidents.

To address this issue, an AI-based system utilising Hitachi’s proprietary Recommendation AI technology was developed. This system analyses extensive historical records of infrastructure failures to identify similar past events. By extracting and presenting information from previous fault reports, the AI assists dispatchers in narrowing down potential causes and suggesting appropriate recovery actions. The learning model used was BERT. Around 3,000 historical data points related to failures and malfunctions were used to train the model.

A field trial was conducted using data from JR East, a rail operator in Japan, from 2020 to 2022 to confirm the effectiveness of the AI-based system. During the proof-of-concept phase, a significant reduction in recovery time was demonstrated. When applied to significant past incidents that typically require around two hours for recovery, it reduced the recovery time by approximately 50%, shortening it to about one hour. Following these successful trials, the system commenced full operation.