產品的可靠度常用一因子的加速衰退試驗進行測試，並推論產品壽命。但可能影響產品壽命的不只一個因子， 因此本文考量了具ED過程之兩因子恆定應力加速衰退試驗(CSADT)的隨機過程模型，並使用數值方法推估產品失效壽命分佈。常見的三種隨機過程Wiener, Gamma, Inverse-Gaussian 過程都是Tweedie ED 過程的特例。 我們考慮兩因子試驗可能有交互作用影響，故交互作用項列入模型的考量。 我們使用兩組兩因子加速衰退試驗的真實資料，進行比較。考慮兩因子是否有交互作用、不同的加速模型，以及Wiener, Gamma, Inverse-Gaussian和ED 過程組合不同的隨機過程模型。主要結論為Tweedie ED process模型在資料上配適上優於其他模型。;Product′s reliability is often obtained by an one-factor accelerated degradation testing and then calculated by the inferred lifetime distribution of products. However, there may be more than one factor that can affect the life of the product. This thesis considers stochastic process model of two-factor constant-stress accelerated degradation testing based on ED process. The numerical method is used to estimate the product′s failure lifetime distribution. The three common stochastic processes: Wiener, Gamma and Inverse-Gaussian, are special cases of Tweedie ED process. We consider that the two-factor test may have an interaction effect and put the interaction term into the models. We used two sets of real two-factor accelerated degradation data to compare with the models by considering the interaction term, different accelerated forms and four stochastic processes. The main conclusion is that the Tweedie ED process model is better than others in model fitting.
