Performance Degradation Analysis For Products Under Complex Stresses

In the practical applications of industries, people pay much attention on reliability of products, as products’failures in industries will lead to bad results. In the past, the reliability information can be acquired from data of time-to-failure. Due to the higher reliability and longer lifetimes of products, it is hard to obtain the time-to-failure data from products. Therefore, people turned to concern the degradation information of products. By using degradation data of products analyzing the degradation process, and making an assessment of product’s reliability, it provides a new thought for reliability assessment and it is of great significance to industry processes.In this thesis, we focus on the degradation data under complex stresses, and build the degradation model under complex stresses to analysis the degradation process. Specifically, this thesis focuses on the research from the following aspects:1. A proper degradation model is adopted to describe the degradation process according to the feature of the degradation data. Then an accelerated stress model is selected to describe the stresses influence according to the types and characteristics of stresses. On the basis of the above models, we establish a gamma degradation model with multiple stresses to describe the degradation process, and estimate the parameters by the maximum likelihood estimation method (MLE). The proposed method is validated by a group of mill degradation data.2. As continuous stresses can be described easily by mathematics, researchers pay much consideration on continuous stresses, and the discrete stresses in the degradation process have been rarely considered. So we consider the degradation process under complex stresses in this part, and emphasize on the influence of discrete stresses. Our contribution in this part is to estimate parameters based on MLE and moments, and to test the parameters with hypothesis tesing-- the likelihood ratio test method. The testing results confirm that if the discrete stress changes in the degradation process, the influence of continuous stresses on the degradation process will be changed accordingly. The mill degradation data also validates the conclusions.3. To deal with some stoachastic factors or noises in the degradation processes, we build a gamma degradation model with multiple stresses and random effects. Meanwhile, considering the complicated forms of likelihood functions, the EM algorithm is adopted to estimate the model parameters. Considering that the EM algorithm can not be used to estimate the parameters of stress coefficients directly, the EM method is specifically improved, and the effectiveness is validatied by the mill degradation data.