Optimization of system reliability robustness using accelerated degradation testing

In the system design stage, several component types (or modules) with competitive reliability are usually available for the design of each subsystem; however, their robustness against uncertain operating conditions might be significantly different. Designers attempt to address the design of a reliable and robust system by choosing the "best" components from those candidates. The problem is more difficult when constructing reliable systems with highly reliable components due to the difficulty of evaluating the reliability and robustness of such components under normal operating conditions. In this paper, we investigate a 3-step procedure for system optimization regarding the reliability robustness through Accelerated Degradation Testing (ADT) experiments. In the first step, baseline ADT experiments are conducted to obtain degradation data for each degradable component type of interest using a small number of test units. In this step, an ADT model along with the parameters of each component type is estimated based on the initial data (baseline data). The second step refines the estimation accuracy of those critical model parameters by designing an optimum ADT plan. The optimum testing plan belongs to a D-optimal plan which minimizes the determinant of the variance-covariance matrix of the parameter estimates. Finally, in the third step the reliabilities of alternative component types are predicted under nominal operating conditions and their robustness is evaluated considering the uncertainty of the operating conditions. The design procedure is demonstrated using an example of a brake system design.