2018 年 84 巻 862 号 p. 17-00594
Railway vehicles are equipped with many kinds of machines such as traction diesel engines and their failures sometimes lead to service disruptions and accidents. Vibration monitoring systems are expected to prevent their failures by detecting their abnormalities at an early stage. In order to make an effectual action after abnormal vibration detection, it is necessary to make a root cause diagnosis. To address this issue, a simple diagnosis method is proposed in this paper. In the method, a measured vibration octave spectrum is divided into three frequency bands and abnormality detections are conducted for the spectra to narrow down the root cause of the vibration. A one class classification method, which is an anomaly detection method in machine learning, is used in the abnormal vibration detection to make a general purpose vibration monitoring system. In addition, the ratios of abnormal vibration are calculated for the three frequency bands to show the progress of the fault. The effectiveness of the method is verified using vibration data acquired on simulated abnormality tests of traction diesel engines, i.e., auxiliary drive shaft failure test and engine abrasion test. The test results show that the proposed method is effective in detecting the abnormality and diagnosing the cause and the degree of the failure.