具等候時間限制之下游多產品機台生產系統控制

Abstract

本研究欲探討具有等候時間限制的多產品共用下游機台生產系統的生產控制問題。產品於上游加工站進行加工後，必須在特定時間內進入下游加工站進行加工，此時間限制稱為作業等候時間限制。若下游在製品違反作業等候時間限制，則會產生重工或報廢成本。且此類系統受機台可靠度、新訂單來到等不確定因素影響，若無良好的控制方法將造成產能利用率下降與生產成本增加。而隨著技術演進，許多機台已發展為可多工生產多種產品的樣貌，本研究考量此種機台的特性做生產系統的控制，並期望能對此類機台產生有效運用產能的控制方法。 本研究透過馬可夫決策過程(Markov Decision Process, MDP)加上動態規劃作為求解方法、以最小化總成本的目標進行求解，並由兩產品共用下游機台模型(ACDCW)推廣至可應用於更多產品的生產系統的多產品共用下游機台演算法(MPCWH)。模擬驗證證明兩個控制方法皆能達到改善成本之目的。

This research develops a dynamic scheduling method for multi-product production systems under process queue time (PQT) constraints, wherein waiting time between consecutive processing steps is constrained by predefined upper limits. If the waiting time of a work-in-process (WIP) violates the corresponding PQT constraint, the WIP may be scrapped or have to be reworked due to quality concern. Machine reliability is another concern in the research. Random machine failure in the system may leads to higher risk of scrap and the increase of cycle time. Thus an effective control method with real time machine reliability considerations is important. First, admission control method of two-product production system is developed using the technique of Markov decision process. The objective is to minimize total expected waiting and scrap costs. In the system having more than two products, a heuristic is developed based on two product model. Through the 3 types of experiment design, simulation analysis were implemented. The result of simulation indicates that both two-product model and heuristic used in multiple product system could improve total cost effectively.