本論文首先提出一個新的修正逐步混合型I設限計畫。然後,我們根據 Childs et al. (2003) 和 Childs et al. (2008) 的結果推導出指數型I設限和修正的逐步混合型I及型II設限資料下之最大概似估計值的分配。利用不同設限資料下所得之最大概似估計值的分配,我們分別針對簡單和一般性的損失函數建立抽樣計畫之貝氏風險函數,再應用 Lam (1994) 的離散分割法或模擬退火演算法找出最佳的抽樣計畫。最後,我們呈現一些數據及比較來驗證本論文所提出的方法之有效性及穩定性。 In this dissertation, we propose a new adaptive Type-I progressive hybrid censoring scheme. We follow the work of Childs et al. (2003) and Childs et al. (2008) to derive the exact distributions of the maximum likelihood estimator of the mean lifetime of an exponential distribution under Type-I censoring and both types of adaptive progressive hybrid censoring schemes. Based on the distributions of maximum likelihood estimator, we obtain the explicit expressions for the Bayes risks of sampling plans when a simple or general loss function is used. The discretization method of Lam (1994) and the simulated annealing algorithm are then used to determine the optimal sampling plans under different censoring schemes. Some numerical examples and comparisons are presented to illustrate the effectiveness of the proposed method.