Type-Ⅰ censored reliability qualification test design for weibull products based on expert judgments

Abstract

Reliability Qualification Testing (RQT) is crucial for confirming that a product complies with specified reliability standards. Traditionally, practitioners design a lifetime test plan by referring to the GJB899A-2009 "Reliability Qualification and Acceptance Testing", which provides test plans subjected to the test specifications and risks. However, these plans often demand lengthy testing periods and are not practical in many scenarios where conclusions need to be provided in a short term. In this paper, a framework of test plan design with a derivation method for risk calculation is proposed for Weibull-distributed products. In this framework, to estimate the probability density distribution of product lifetime T with multiple expert judgments, a nonlinear optimization problem is modeled, and imprecise prior information is converted to combinatorial constraints. To search for parameters' approximate optimal solution, PSO-MEM is constructed, which is a framework integrating the particle swarm optimization (PSO) with the maximum entropy principle (MEM). What's more, to cover a relatively comprehensive range of test specifications and types, equations are modified under three typical distributions and two kinds of lifetime test plans. Besides, the proposed method is compared with authoritative standards like GJB 899A-2009 under different test conditions. The results show that test plans searched in this paper yield lower risk value than standard plans, which highlights the effectiveness of the new framework in deriving RQT plans under multiple imprecise information and different test modes.