Railway point devices act as actuators that provide different routes to trains by driving switchblades from the current position to the opposite one. Point failure can significantly affect railway operations, with potentially disastrous consequences. Therefore, early detection of anomalies is critical for monitoring and managing the condition of rail infrastructure. We present a data mining solution that utilizes audio data to efficiently detect and diagnose faults in railway condition monitoring systems. The system enables extracting mel-frequency cepstrum coefficients (MFCCs) from audio data with reduced feature dimensions using attribute subset selection, and employs support vector machines (SVMs) for early detection and classification of anomalies. Experimental results show that the system enables cost-effective detection and diagnosis of faults using a cheap microphone, with accuracy exceeding 94.1% whether used alone or in combination with other known methods.

A practical condition monitoring method is proposed for the fault diagnosis of railway point machines (RPMs) by considering the difficulty of obtaining in-field failure data. Failures in RPMs have a significant effect on railway train operations, and it is very crucial to detect abnormal conditions in RPMs. However, it is generally difficult to obtain in-field failure data for a classifier training step. A diagnosis method using dynamic time warping is proposed to manage the variation in durations of RPM movement without a training step. On the basis of the experimental results with RPMs operated in Korea, it is believed that the proposed method without a training step can detect abnormal electric-current shapes more accurately than previous training-based methods.

Detecting replacement conditions of railway point machines is important to simultaneously satisfy the budget-limit and train-safety requirements. In this study, we consider classification of the subtle differences in the aging effect—using electric current shape analysis—for the purpose of replacement condition detection of railway point machines. After analyzing the shapes of after-replacement data and then labeling the shapes of each before-replacement data, we can derive the criteria that can handle the subtle differences between “does-not-need-to-be-replaced” and “needs-to-be-replaced” shapes. On the basis of the experimental results with in-field replacement data, we confirmed that the proposed method could detect the replacement conditions with acceptable accuracy, as well as provide visual interpretability of the criteria used for the time-series classification.

No abstract

No abstract