Abstract
In manufacturing, a change sometimes occurs in the raw materials and/or in the production process after which the lifetime distribution of the products is different from the distribution before the change. Usually the time at which the change occurred is unknown. In this article, we investigate a method to detect the change-point when only the marginal failure counts for the product are available. We present a nonhomogeneous Poisson process model for the repairable products and discuss nonparametric estimation of the parameters, the mean number of failures per product, before and after the change. We apply the EM algorithm to obtain maximum likelihood estimates of the parameters and use the Akaike Information Criterion (AIC) to detect the change-point. An example is given which is based on the field reliability data of an electrical product, and simulation studies are conducted to evaluate the performance of the proposed method.
