Real-Time Response-Based Fault Analysis and Prognostics Techniques of Nonisolated DC–DC Converters

Abstract

DC-dc converters have been extensively used in industrial systems and their failure may cause the systems to work abnormally and even accidents. Current research on circuit fault analysis is mainly based on a model or complex signal processing. In order to avoid modeling difficulties and complex selection of signal processing methods, a fault analysis method based on circuit output signal is proposed in this paper, which can judge the circuit health status by directly monitoring the circuit response. This method of monitoring output signal also avoids the problem of inconvenient access to internal nodes of the converter circuit. By continuously monitoring the output voltage, two advantages are simultaneously achieved: 1) the fault components range of an abnormal circuit is located and 2) its remaining useful performance (RUP) can be predicted. The particle filter algorithm commonly used to predict RUP often encounters the problem of particle degeneracy, which leads to the decrease of particle diversity and makes the final RUP prediction accuracy lower. In view of this problem, the developed novel way of RUP prediction by monitoring the circuit output voltage is based on a newly developed conditional particle filter, which tracks the degradation process of the circuit according to collected historical degeneration data. The case studies of two dc-dc converter circuits show that the developed fault analysis method can easily analyze the fault condition of the converter circuits and the improved particle filter algorithm achieves excellent prediction accuracy.