Title:

PiRail: Railway Fault Detection

PiRail is a budget-friendly compact railroad track infrastructure monitoring system designed to improve inspection through automated collection and detection. The system integrated a Raspberry Pi data logging platform with GPS and IMU sensors to record and geolocate motion data in real time. The structured JSON data is stored for future analysis and transferred to a webapp for live monitoring. This year, a React-based dashboard was designed and implemented with real-time IMU strip charts, allowing an inspector to watch the speed, pitch, roll, and vertical acceleration over time. Changes were made on the server-side to objectively detect anomalies. We collected data on the Cotton Valley Rail Trail in Wolfeboro using a railcar equipped with the PiRail box. An experiment was run to measure the system’s response to artificial bumps on a well-maintained rail segment. The analysis focused on taking the data we collected and running evaluations on z-axis acceleration, pitch, roll, and y-axis angular velocity to find outliers. Unsupervised machine learning models were used to find patterns between labeled points of interest (POIs) and unlabeled data.

Name

Comment*

Comments are viewable only by submitter

Captcha*