應用半馬可夫過程於可靠度之研究－以電廠水泵為例

Abstract

目前國內的核能電廠已屆使用年限，因此衍生出來與「可靠度」及「風險」有關的議題受到重視。過去有諸多學者利用馬可夫過程進行電廠的可靠度相關研究，但是往往侷限在失效率為常數的案例當中，忽略了實際的物理現象。因此本研究將以電廠中的水泵為例，利用半馬可夫過程建構出符合實際物理現象的數學模式，以求精確地描述水泵的實際表現。除此之外，針對水泵所生出的風險議題，包括「預防維護排程」與「水泵延役」的問題，本文利用風險管理的手法，利用生命週期成本與層級分析法建立完整的數學分析模式，試圖得到管理上的意涵，提供電廠人員做參考。研究結果顯示水泵的可用度會隨著時間逐漸衰退至穩定的狀態。除此之外，經過計算後我們得到水泵最適合的維護週期，並且發現可用度經過預防維護後確實有顯著的改善作用，因此有施行預防維護的必要性。最後針對水泵延役的問題，我們建議更換某些原有的水泵，藉此提升系統的運轉績效。

‘Safety’ and ‘reliability’ have long been important issues related to nuclear power plants. In the past, several researchers have applied Markov process to assess the reliability of nuclear power plants or their components, but often restricted to cases of constant failure rate. Many of them did not take real practices into consideration either. In the present thesis, a Semi-Markov process is proposed to study the reliability of nuclear power plant components. The real data taken from the maintenance record of a few water pumps in a nuclear power plant are used as demonstrative examples. Other safety related issues of pumps such as ‘preventive maintenance scheduling’ and ‘life extension problem’ are also studied by employing risk management tools such as ‘life cycle cost’ and ‘analytical hierarchy process’ in the analyses. It is found that availability of the studied pumps declines dramatically in a transient manner in their earlier usages, and tends to be steady-state in some time. It is also found that the implement of preventive maintenance is necessary. And an optimal periodic maintenance time-cycle is obtained. With regard to the life extension of pumps, it is suggested that some pumps should be replaced rather than maintained continuously after a certain time of usage in order to improve the overall system performance.