具連續作業等候時間限制之平行多機生產系統控制

Abstract

本研究探討具連續作業等候時間限制下，平行多機生產系統的動態允入控制策略。連續作業等候時間限制為：序列式生產系統中任兩連續製程，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。 近代產業常存在作業等候時間限制的問題，例如在半導體產業中，若在製品違反作業等候時間限制，則可能會造成晶圓表面氧化而須重製或報廢。隨著先進製程發展，具作業等候時間限制的製程數急遽增加，另外站點機台數增加也讓派工策略更為困難，因此有必要針對此類型問題，發展妥善的生產管制策略。 本研究基於馬可夫決策過程，以正規化方法將連續時間軸轉換為事件導向的離散時間軸，發展多工作站多機台啟發式控制法(MMHC)。MMHC利用子系統拆解法將多工作站系統拆解為數個兩工作站子系統，針對兩工作站子系統發展動態規劃模型，目標為最小化總生產成本，再將子系統求解結果組合為動態控制策略。最後以系統模擬與其他文獻提出的方法進行比較，結果顯示MMHC控制策略能有效降低總生產成本。

Production control in parallel multi-tool systems under sequential process queue time constraints is studied in this research. When WIP finished an upstream process step, the WIP should be processed at the next process step within a given amount of time. The time constraint is called the process queue time constraint. Any violation of the process queue time constraint seriously impacts yield quality and incurs significant scrap costs. In the semiconductor industry, the violation of process queue time constraint may cause the wafer surface oxidation to be reproduced or scrapped. Under the sequential process queue time constraints and multi-tool production environment, the dispatching strategy is hard to be defined. Besides, machine reliability and demand uncertainties also make the production control difficult, production managers need to balance the need for achieving production targets and yield quality. The Multi-step with Multi-tool Heuristic Control (MMHC) is developed to minimize long-run average production cost. The MMHC strategy is formulated with stochastic dynamic programming. The model objective is to minimize the sum of inventory holding costs and scrap costs. The simulation results reveal that our method significantly reduces total production costs.