Monitoring The Time To First Failure From Warranty Claims

The warranty data reflects the status of the product, and the warranty data analysis is critical to improve product quality and reliability. Under an assumption that the lifetime of products follows a Weibull distribution, we propose an analysis model of warranty data using two parametric methods for monitoring the time to first failure in warranty claims data, one is a cumulative sum(CUSUM) chart and the other is an exponentially weighted moving average(EWMA) conditional expected value(CEV) chart. Both of these two control charts are developed for censored data to monitor changes in scale parameter with a fixed shape parameter.In this thesis, data experiments are implemented to study the effects of some related parameters of the control charts based on average run length(ARL1). As a metrics, the zero state ARL1 is used to compare the CUSUM control chart and the EWMA CEV chart based on the likelihood ratio. Further more, the EWMA CEV control chart is used in the warranty data analysis model and a modification is implemented to improve this model, considering the time delay by the process of collecting the warranty data. The influence of different m value for warranty data model is discussed and related suggestions are presented.