A Study on Health Indicator for Rivet Breakage in Train Door System Using Motor Current Signal

Abstract

In this study, health indicator is developed to diagnose the degradation state against the rivet breakage, which is one of the critical failures in the train door system. For this purpose, motor current signals are acquired during the routine maintenance for the train doors of new, 13 and 15 years old, in which the rivet breakages are found in some of those in the 15 years. Features are extracted for decomposed signals after performing the wavelet packet decomposition. Useful features are selected that show monotonicity with respect to the door age and rivet failure, while those with redundancy are removed. Health indicator model is constructed using the selected features by regression to assess the current health against the rivet failure. As a result, it is found that the health indicator increases as the door ages and reaches the maximum value for those with rivet failures, which may be valuable information to the maintenance engineer.