Data analysis and resource allocation in Bayesian selective accelerated reliability growth

Abstract

The rapid pace of technology advancement has resulted in increasingly complex systems with more potential failure modes. However, it is quite common that multiple key components of such a system may be developed, tested and improved independently during product development. Without taking a holistic approach to system reliability improvement, a significant amount of time and resources may be wasted on over-design of some components, which can be otherwise used for strengthening other under-designed components. The technical challenge is more prominent when accelerated testing is utilized in a reliability growth program in hopes of shortening the system development cycle. To overcome limitations of the traditional reliability growth method using the Crow-AMSAA model, a Bayesian selective accelerated reliability growth method is proposed in this article to accelerate potential failure modes and aggregate component testing results and prior knowledge for predicting system reliability growth and corrective actions. As one of the key steps, the method dynamically allocates limited resources for testing and correcting failures on all system levels. Numerical examples illustrate that the proposed integrated statistical and optimization method is effective in estimating and improving the overall reliability of a system.