Detection of the Incipient Degradation for Aluminum Electrolytic Capacitors in Track Circuits

摘要

Aluminum electrolytic capacitors used in the tuning and matching units of track circuits are prone to degrade. The decrease of capacitance results in the drop of main track output voltage (MTOV). However, it is difficult to detect the incipient degradation by monitoring the MTOV because the impact of ambient factors such as variations of temperature and humidity is significant. This paper proposes a data-driven method to reduce the detection delay for the degradation of capacitors. As the fault data is difficult to be collected, this method depends only on a few days of historical data for capacitors undegraded. A model is trained by support vector regression based on the on-site data of MTOV in normal condition. Then, this model is used to generate a residual signal for on-line monitoring. In order to verify the effectiveness of the proposed method, it is assumed that the degradation rate is constant and the remaining useful life is 1000 hours when degradation starts. Then, voltage data for degraded capacitor can be synthesized by utilizing a simulated curve of MTOV vs. capacitance. Simulation results demonstrate that impacts of daily variations of temperature and humidity on the MTOV can be reduced considerably from the residual signal. By monitoring the residual with a small threshold, the detection delay for the degradation can be reduced by several thousand minutes compared to method by setting a threshold of 20% drop on the MTOV. Data preprocessing to eliminate the sudden voltage drop by train occupancy and threshold selection are also reported.