Hybrid physics based-data driven approach for reliability allocation of early stage products

This paper presents a novel approach for reliability demonstration of a product operating in dynamic environment. The device is subjected to mechanical loads, humidity and temperature swings. The form and magnitude of mission loads is uncertain a-priori and best estimates from simulations are used to develop an initial reliability test plan. The focus of this paper is to design reliability tests that accurately represent field conditions and develop a framework to iteratively update the test plan based on field data. The paper uses design failure mode and effects analysis to evaluate risk and prioritize tests. System reliability model is used to flow down requirements and statistical techniques are used to develop empirical life models from field data. Finally, a Bayesian method is used to update models and test plans iteratively as new data becomes available. This framework is used to analyze the failure of a MOSFET based switching circuit used in AC Motors by correlating field failures to reliability test data.