Multi-stage Degrading Process Modeling with Change Point Detecting

Abstract

As the changes in the external environment and internal mechanisms, some components of high-reliability products may show the multi-stage failure processes under long-term operation. That is a mutation occurring in a degradation path at some time which we called change points segmenting the degradation process. It is not caused by external shocks but due to the qualitative change caused by the degradation of natural degradation to a certain extent. However, it hard to get the value and position of the change point. In this paper, we establish the model of multi-stage with change point degrading process and give its parameter estimation. Furthermore, we detect whether the change points exist in the degrading process and determine the time interval where the change point happens by using Schwarz information criterion (SIC) iteratively. We also introduce the experiment likelihood method to detect the existence of change points when the degrading process distribution is unknown. Finally, an experimental analysis is conducted by using SIC method iteratively to verify its validity and flexibility in the change point detection.