Reliability prediction and testing plan based on an accelerated degradation rate model

Accelerated Degradation Testing (ADT) is a viable alternative to accelerated life testing with censoring to estimate reliability without waiting for actual failures to occur. However, the estimation accuracy relies greatly on both precise representation of covariates' impacts on degradation behaviour and a carefully designed ADT plan. In this paper, an ADT model, called Accelerated Geometric Brownian Motion Degradation Rate (AGBMDR) model, is proposed by modelling degradation rate in order to explain covariates' effects and inherent degradation rate variation precisely. Based on baseline parameter estimates of the model through a pilot ADT experiment, a local optimum ADT plan is developed to refine estimation accuracy of interests. The objective considered is to minimise the Generalised Variance (GV) of parameter estimates. A numerical example is provided to demonstrate the reliability inference procedure and the optimum ADT design methodology. The result shows that the optimum ADT plan leads to a more efficient experiment than the traditional ADT plan in terms of relative efficiency criterion.