Non-parametric hazard rate estimation of hard failures with known mileage accumulation rates in vehicle population

Abstract

Major automobile companies spend billions of dollars annually towards warranty costs. Engineers carry out field reliability studies to obtain appropriate feedback for design, manufacturing, and/or service fix that help to reduce the high warranty costs. Warranty claims data, however, are known to be messy and need to be carefully used during statistical modelling and analysis to extract useful feedback. Hazard rate estimation is an important step for reliability and design engineers in obtaining a meaningful feedback from warranty claims data. The methodology discussed in this paper uses mileage accumulation rates in the vehicle population to calculate the number of vehicles at risk of a warranty claim at the start of a month and to provide a non-parametric estimate of the hazard rate. Although there are many books and articles that discuss reliability and hazard rate analysis in such a situation, this article provides a simple and stepwise methodology with an application example for component/sub-system level hazard rate estimation when dealing with automobile warranty data.