Estimating the reliability model parameters through a simulation of warranty claims: How much data is needed?

Abstract

Manufactured goods are put through a quality control process that ensures an acceptable reliability level before being released to the public. However, manufacturing processes can change over time, which may affect the quality of a product. Warranty claims provide an opportunity to obtain feedback on the manufacturing process. In this study, a Weibull failure model is used to simulate the occurrence of warranty claims. The parameters of the Weibull model used to generate the data are to be estimated from the simulated warranty claims. The length of time required to obtain accurate parameter estimates is examined. The study is based on claims arising from the failure of a single component, but the techniques can also be used for a simple product or a subsystem of a more complex product that can be suitably modelled by a Weibull distribution. Our study shows that at least two years of data are required to obtain acceptable estimates of the Weibull parameters.