G-renewal process as a model for statistical warranty claim prediction

A brief overview of the statistical aspects of warranty prediction is given as an introduction. The main discussion then focuses on warranty claim prediction for repairable products. Introduced by Kijima and Sumita (1986), a g-renewal process (GRP) can be considered as a model for major repair assumptions encountered in repairable product reliability analysis. These assumptions include “good-as-new”, “same-as-old”, the intermediate “better-than-old-but-worse-than-new”, and “worse-than-old”. A statistical procedure is developed for estimation of the GRP parameter, which is suggested to have engineering meaning of the effectiveness of the repair actions. A practical example of the GRP application in statistical warranty prediction is given as an illustration of the proposed estimation method. The paper arrives to the following conclusions: The GRP provides high flexibility in modeling real life failure occurrence processes by covering major repair assumptions encountered in practice. A Monte Carlo simulation can be considered as a method for statistical estimation of the GRP. Warranty claim prediction based on GRP provides a higher accuracy compared to the ORP or the NHPP