加速壽命試驗是一種快速測量物件壽命的方法。本文考慮單一應力下兩階段的型一逐步設限加速壽命試驗, 在群組資料下假設物件壽命分配為指數分配, 且物件壽命和應力水準有線性與對數線性兩種模型之下, 分別在V-準則、D-準則和A-準則下決定最佳化試驗時間。在貝氏最佳化試驗中, 我們建議決定先驗分布的方法, 並比較其與傳統方法之差異。研究結果顯示, 若我們對參數有充份的訊息, 則貝氏最佳化設計較傳統方法好。 Accelerated life test (ALT) is a quickly method for estimating the life of item in reliability. Optimal design on the experimental time is an important issue on the ALT. In this thesis, we consider a two-step-stress accelerated life test under progressively Type-I censoring scheme, and assume that the lifetime of each item follows an exponential distribution with grouped data. Determinations of the optimal experimental time based on V-optimality, D-optiamlity and A-optimality criteria with the mean life time of the experiment item being a linear and loglinear function of the stress variable are disscussed, respectively. We also study the corresponding Bayesian designs, and provide a way to select the prior distribution compared with the traditional methods. Empirical studies indicate that if we have more information from the parameters, we can do beter than the traditional design.
