MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641110361 A) filed by Vellore Institute Of Technology on September 15, 2026, for Edge-Based Dual-Stage Machine Learning System For Railway Track Monitoring And Fault Classification.
Inventors include Mohanraj G; N Balaji Sai; and G. Dineshkumar.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: The present invention provides a real-time embedded railway track monitoring system configured to detect and assess track-related variations using two mechanically coupled sensing channels. An embedded controller continuously establishes an adaptive baseline representing normal signal behaviour and detects deviations exceeding an event trigger. Detected events are represented using event-level features comprising peak deviation, event duration, rise rate, and signal variance. A two-stage conditional machine-learning architecture processes the extracted features, wherein a Decision Tree first distinguishes noise from genuine track-related events, and only validated events are forwarded to a 12-tree Random Forest for severity classification. The processing and machine-learning inference are performed locally on the embedded controller to reduce computational and communication overhead. Each validated event is associated with geographical coordinates obtained through GPS and transmitted through LoRaWAN for remote monitoring and maintenance support. The system thereby provides adaptive, event-driven, and computationally efficient railway track monitoring.
Disclaimer: Curated by HT Syndication.
