Traction diesel engine anomaly detection using vibration analysis in octave bands

Abstract

Traction machines are essential parts for a train to run. Therefore, a condition monitoring system (CMS) is being developed, that detects machine failure in the early stages to prevent traffic disruption. The CMS observes the vibrations of a machine and detects abnormal vibrations with a machine learning algorithm. In the CMS, octave-band analysis is performed to extract feature vectors from vibration data. Running tests were conducted to verify the performance of the CMS. Test results showed that simulated abnormal vibrations were clearly distinguishable from normal ones with the CMS.