基因演算法應用於動態可靠度之研究

Implementation of Genetic Algorithm into Dynamic Reliability System

翁昭榮(Chao-Jung Weng)

指導教授 ： 陳雲岫

元智大學/工程學院/工業工程與管理學系/碩士(2003年)

摘要

本文主要探討伴隨運作時間增加，元件可靠度逐漸衰退的週期性(periodic maintenance)預防維護作業，研究探討機械系統的週期性的預防維護與預防更換作業為主。有別於過去學者在Renewal模式下的維護探討，本文以動態可靠度描述機械元件隨運作時間遞減的壽命曲線，並在壽命衰退模式(Age reduction model)的基礎下呈現預防維護作業對於可靠度與失效率的影響。本文同時考量可靠度、妥善率與成本三個績效指標，藉由基因演算法建立最佳的預防維護作業組合。經由多目標評估所建立的預防維護作業，在滿足系統可靠度需求的同時，也同時考量了妥善率與成本的情況。
本文藉由等高線圖呈現週期維護階段中，最適預防維護計畫的三個績效指標值。可靠度需求高的系統，妥善率也相對越高；可靠度較低的系統，失效風險與維修保養成本導致整個後勤成本的增加，尤其當風險失效成本遠大於零件預防更換成本時。最後，研究顯示了元件預防維護作業的選擇受到維護因子影響，維護效果較高的元件通常以預防維護為主(1P.)；維護效果較差的元件，通常以預防更換為主，尤其當系統可靠度需求較高的情況。

關鍵字

預防維護 ； 可靠度 ； 妥善率 ； 壽命衰退模式 ； 基因演算法

並列摘要

In this thesis, a periodic preventive maintenance (PM) of a system with deteriorated components is modeled by genetic algorithm. Simple preventive maintenance and preventive replacement are considered simultaneously for a periodic preventive maintenance schedule of mechanical systems. The life deterioration of components is modeled with dynamic reliability equation, and the effect of PM activities with reliability and failure rate of components under an age reduction model is studied.
Reliability, availability and cost are considered simultaneously as the multiple performance evaluation indices in scheduling the optimal maintenance scheme at each PM stage by the genetic algorithm. From this research, the proposed preventive maintenance operation from multiple evaluation aspects is more diversification and eclectic.
We present the three performance evaluation indices of the optimal maintenance scheme on every periodic preventive maintenance stage by Contour map. The more reliability of the system is, the more availability it is. As far as a lower reliability system, the Logistics Support cost would increase because of risk cost and corrective maintenance cost increased, especially when the risk cost is more expensive than component preventive maintenance cost. Eventually, the PM. activities chosen on each stage is affected with the improvement factor of components. The main PM. activity of the higher improvement factor of components is 1P; and on the other hand, the worse improvement factor of components, more the opportunity is 2P. implemented, especially to the system required higher reliability.

並列關鍵字

Preventive Maintenance ； Reliability ； Availability ； Age Reduction Model ； Genetic Algorithm

參考文獻

4 Barlow R. and L. Hunter, “Optimum Preventive Maintenance Policies,” Operations Research Society of Americas, Vol. 8, pp. 90-100, 1960.

5 Canfield, R. V., “Cost optimization of Periodic Preventive Maintenance,” IEEE Transactions on Reliability, Vol. 35, pp. 78-81, 1986.

6 Chan, P.K.W. and T. Downs, “Two Criteria for Preventive Maintenance,” IEEE Transactions on Reliability, Vol. 27, pp. 272-273, 1978.

7 Cox, D. R., Renewal Theory, Methuen & Co., London, 1962.

8 Goldberg, D. E., Genetic Algorithms in Search, Optimization and Machine Learning, Addison Wesley Professional, 1989.