AI-Driven Track Asset Lifecycle Prediction for Indian Railways Using Laser QR, Weather and Service Loads

Abstract

Railway track assets such as rail sections, fastenings, sleepers and welds are still inspected largely on fixed calendar schedules, so degradation that depends on traffic and climate is detected late or not at all. This paper presents Smart Track AI, a web-based platform that links each track asset to a laser-marked QR tag, collects service-load and weather telemetry at the point of inspection, and returns an estimate of Remaining Useful Life (RUL) together with a three-level health status. The platform couples a browser-side camera scanner with a Flask inference service that serves a pre-trained Random Forest regressor. Because real Indian Railways failure logs were not available to the authors, the prediction pipeline was evaluated on 3000 synthetic asset records whose degradation follows load-, corrosion- and thermal-fatigue-inspired relationships. On the held-out set the Random Forest reached an MAE of 5.20 months and an RZ of 0.945, and mapping the predicted RUL to Optimal, Warning and Critical bands gave 92.3% status accuracy; gradient boosting was marginally more accurate. The results demonstrate that the end-to-end workflow is feasible; validation on real railway data remains future work.