Railway Interchange 2017 Buzz

Sessions Speakers Floor Plan Buzz

A New Way to Detect Bridge Strikes with Neural Network Pattern Recognition (Room Sagamore 2)

19 Sep 17
9:00 AM - 9:30 AM

Tracks: AREMA Committee Meeting, AREMA Technical Sessions- Structures

The objective of this research is to use artificial neural networks to identify bridge strikes using data collected from accelerometers and tilt-meters. The data has been collected from four Metro-North Railroad (MNR) bridges over a period of 18 months, during which multiple bridge strikes occurred. The data is used to train artificial neural networks that can evaluate data streams, identify patterns in time histories indicative of vehicular strikes, and distinguish vehicle strikes from train traffic. Both neural networks and traditional signal processing techniques are used to classify bridge strikes in collected data. The MNR New Haven Line receives approximately 200 bridge hits from vehicles per year. Each time a bridge is struck, MNR must reduce train speed or stop traffic until the structure is cleared by an MNR Railroad Bridge Inspector. Prior to the installation of these sensors, the railroad was reliant on MNR personnel, the police and the general public to report strikes. Without any notification, a bridge may go uninspected but remain open to traffic. As a result, it could be many hours or days until debris or new damage is noticed and the bridge inspected. The goal of this research is to detect bridge strikes that go unreported and reduce the time it takes for key personnel to be notified in the event of a bridge strike.