等候時長限制下串聯生產系統之動態派工與保養方法

Abstract

本研究針對受等候時長限制的串聯生產系統提出動態派工與預防保養方法結合允入控制。本研究考量系統中常見的隨機事件，以最小化等候和報廢成本為目標，利用馬可夫決策過程配合動態規劃，提出動態派工與預防保養模型，同時本研究提出混整數規劃模型，透過產能配置將大維度問題分解為多個小問題，以克服動態規劃模型求解時間過長的問題。由於生產系統的隨機性是違反等候時長限制的主要原因，本研究提出允入控制方法，透過精準的生產控制避免違反等候時長限制衍伸的高額成本。本研究預期能根據所提出之方法進行動態的派工與預防保養決策優化，並通過與允入控制方法的結合，有效避免產品違反等候時長限制，降低整體生產成本。

This study proposes a dynamic dispatch and preventive maintenance method combined with admission control for serial production system under queue time constraint. This study considers common random events in the system, aims at minimizing waiting and scrapping costs, using dynamic programming with Markov decision process, and proposes a dynamic dispatch and preventive maintenance model. To overcome the problem of long solution time for the dynamic programming model, we decompose high-dimensional problems into multiple smaller problems by a mixed integer programming model that allocate production capacity. Since the randomness of the production system is the main reason for violating the queue time constraint, this study proposes an admission control method to avoid the high costs of violating the queue time constraint through precise production control. We expect to optimize dispatching and preventive maintenance decisions dynamically based on the proposed method, and by combining it with the admission control method, we expect our method can effectively prevent products from violating the queue time constraint and reduce overall production costs.