Application of Computing Analysis to Real-Time Railroad Track Inspection

Prevention of failures in track has continued to be a major railroad need. The problem is worsening, due to technical/cost limitations in inspection technology coupled with continuing increase in track usage/stress. Typically, railroads use an ultrasonics, on-track, computer-based, regularly scheduled, real-time inspection. However, the system development emphasis has focused primarily on the ultrasonics, while incorporating more computer science based data analysis could help gain insight and reduce the testing errors. Based upon applied research at the University of Connecticut, together with local industry, this paper describes the pattern recognition needs, and proposes a specific solution approach.