Optimal allocation of test times for reliability growth testing with interval-valued model parameters

Reliability growth testing is widely used to identify and remove failure modes in the development of complex systems. Different models have been proposed to track the progress of reliability growth during test, and previous research has addressed the improvement of after-testing system reliability by allocating limited testing resources. The majority of reliability growth testing models are based on the AMSAA/Crow model with known parameters, but there is a lack of work focusing on the situation when the AMSAA/Crow parameters are subject to uncertainty. In this article, we investigate a reliability growth testing allocation problem to series–parallel systems that considers parameter uncertainty in the AMSAA/Crow models. The model parameters are assumed to be known as uncertain-but-bounded values. Interval arithmetic and an interval order relation reflecting decision maker’s preference are used to analyze the uncertain parameters. In order to determine the optimal allocation of testing time for each component or subsystem aiming to maximize the after-testing system reliability, a modified genetic algorithm is developed. A penalty function is used to handle the testing resource limitations, and the techniques of dual mutation and random keys are used in the algorithm to improve searching efficiency. Computational results indicate that the interval analysis is a useful tool to deal with parameter uncertainty of reliability growth testing allocation problems.