Design and evaluation of accelerated life testing plans with dual objectives

Accelerated life testing (ALT) is widely used in industry to predict product lifetime and lifetime distribution. Optimal test plans based on statistical optimality criteria, such as D-optimality or U_c-optimality, have recently been explored in Monroe et al. (2010, 2011). However, in these studies, the product use condition was set at a constant; in reality, use stress may vary due to a changing product use environment. In this paper, we consider an I-optimality that minimizes the variance of lifetime prediction over the entire region of possible use conditions. In addition, we investigate the tradeoffs between model parameter estimation and model-based prediction for ALT test plans. Dual-objective optimal test plans are provided to experimenters so that they can make a decision to balance the plan's estimation and prediction properties. We employ some graphical tools, including fraction of use space (FUS) plot, efficiency plot, and Pareto frontier plot, to evaluate diffierent test plans and to compare our proposed method with existing ones in literature.