Obstacle Detection on Railway Tracks Using Vibration Sensors and Signal Filtering Using Bayesian Analysis

Abstract

A railway track sensing technology was tested using accelerometers integrated along railway tracks at specific sites. Rock and timber of different weights were dropped on rail head, ballast, and sleeper at various distances and vibration generated in the railway track was measured using accelerometer. Rocks and timber drop on the rail head could be detected from a distance of 500 m. In the next phase, a locomotive was moved along the railway track and the signal generated by obstacle drop was filtered out from extremely high level of acoustic noise generated by locomotive motion using a novel Monte Carlo-based Bayesian analysis. The results indicate that the application of Bayesian analysis with the capability of filtering out signal from heavy acoustic noise in vibration sensing technologies can radically improve the reliability of sensor networks.