Abstract
一般而言，一製程之退化原因不外有二：(1) 老化過程 ( agingprocess )[內部因素];(2) 外力衝擊 ( ext- ernal shock )[外部因素]。為維持製程效率及其穩定性，我們可以利用適當的管制圖以防止製程更趨退化。在此，我們考慮預防保養或適時的替換策略。本文將分別討論連續監看(continuously monitoring ) 及按時監看( scheduledlymonitoring )兩種模式之最佳預防保養替換策略。根據退化程度的不同，我們將系統分成數種不同的狀態( states )，針對每一個狀態應有一個最佳之因應行為( action )。在本文中僅考慮了三種行為： 1. DoNothing:讓系統處於原來的狀態，不對它作任何行為。 2. Repair:修理此系統，使它能回復到某個狀態。 3. Replace: 即替換一個新系統，使其回復到全新的狀態。本文之目的即應用更新--報酬定理(Renewal-Rewards Theorem)以求得最佳狀態--行為之對應組合方式。方法如下：每次的替換當做一個更新( renewal)，兩相鄰更新的間隔時間為一個循環(cycle)，在每一循環中可能會有許多的修理。對每個策略找出其在循環間的運作成本( rew- ards) ，並比較各個策略之長期平均成本，則可以找出最佳之預防保養或適時的替換策略。第一章為緒論，描述研究動機及目的、研究架構與相關研究之文獻回顧。第二章介紹老化過程及連續監看與按時監看之預防保養或適時替換策略模式化。第三章介紹外力衝擊及按時監看之預防保養或適時替換策略模式化。第四章則用一個簡單的例子以驗證模式並作一些結論。第五章列出一些後續研究以供參考。The deterioration of a machine or a process is in general dueto:(1) aging or (2) external shock. We can maintain thestability and performance of a production process in terms of asuitable control chart so that actions can be properlytaken to prevent it from futher deterioration. In this paper,two cases, continuously monitori- ng and scheduledlymonitoring respectively, are considered. In each case, wewill study how to determine an optimal maintenance/replacementpoli- cy. Firstly, system states are defined according todifferent levels of deterioration. Our problem is then todetermine an optimal action for each state. However, in thisthesis, only three kind of following actions are considered:(1) Do Nothing: leave the system as it is, i.e. we do nothingabout it. (2) Repair: after repair, the system returns tostate r. (3) Replace: replace the system by a new one, i.e.the system returns to state 0. The method is to find whatproper actions vs different states are taken in terms of theRenewa- l-Reward Theorem. Each replacement is regarded as arenewal and the time between two successive rep- lacements isconsidered to be an operation-cycle. During an operation-cycle, there may either no repair at all or at least onerepair in between. We can compute the expected cost forvarious maintenance/replacement policies during anoperation-cycle and find out an optimal one after comparingthem.