Keywords: generative ai, occ, time-series sequence, ai-assisted recovery system, signal and communication control failure information database system, 1st gen., 2nd gen..

This paper reports on the development and deployment of an AI?based Failure Recovery Support System (“AI-Assisted Recovery System”) designed to facilitate rapid restoration in the event of failures in railway signalling systems by leveraging knowledge from past failure cases. The system extracts similar historical cases, estimates potential causes, and proposes recovery strategies when a failure occurs, while also automatically generating the incident report that dispatchers traditionally prepare. The AI model was trained on failure data accumulated since 2005, equipment manuals, and recovery procedure documents, enabling it to provide suggestions comparable to those of experienced personnel. The system has been in operation within the Tokyo Operation Control Center (OCC) since 2023, and demonstration tests confirmed that it can reduce recovery time by approximately 50%. Furthermore, in 2025, generative AI technology was introduced to enhance the system, enabling automatic recognition and recording of operational conditions based on radio communications, as well as providing estimated recovery times and other advanced support functions.

download the full-text of this paper (PDF, 12 pages, 979 Kb)

go to the previous paper
go to the next paper
return to the table of contents
return to the volume description