工單不確定流動之分析與控制

Abstract

半導體供應鏈是一個變異很高的生產系統，製程良率、機台當機、產品需求與訂單變動等，都是影響生產作業的不確定因素，由於半導體製造管理原本就相當複雜，這些動態事件又經常發生，導致生產績效的變異很高、可預測性非常不理想。就供應鏈本身而言，供應鏈節點的所有權是分散而非統於一的，各節點的營運目標也不全然一致，因此供應鏈所要求的製造服務品質有別於過去一貫注重的節點內（intra-node）工廠效率，其挑戰在於節點間（inter-node）有效協同合作、製造服務的監控、以及可靠的交貨績效等。 達成這些新的要求需要新的管理與控制方法，本研究的目的乃是發展供應鏈協同式生產規劃與排程控制方法，提高供應鏈系統的可控制性及交貨績效的可預測性，並研究「佇列等候時間延遲現象」的成因及其對平均流程時間之影響、行為模式。本研究提出供應鏈工單不確定流動之控制模型，以工單的角度進行服務績效控制，透過供應鏈生產排程方法，在不確定因素的影響下，進行排程修訂，藉此提高服務績效的品質及可靠度，提供顧客準確的延遲交期等進階資訊，達到提高交貨績效可預測性的目的。針對供應鏈中不確定因素（如：製程良率、機台當機及非定期維護等）所造成的產能流失，影響工單在系統內的平均流程時間，進而延遲完工日期，降低服務品質及可靠度，本研究亦透過供應鏈生產排程方法，研究不確定因素、工單來到率及產能對平均流程時間的影響，分析「佇列等候時間延遲現象」的成因，並利用排程模組進行動態系統內等候時間的仿真。

The semiconductor supply chains are full of uncertainties and replete with dynamic events of process variation, engineering changes, operation excursions, and demand changes. These events could have profound effects on the operations for the supply chains, resulting in large variation and low predictability in performance. Besides, the supply chains are long in geographical distance, segmented in ownership, varied in business objectives, and demanding in collaboration. The challenges of supply chains lie in effective collaboration between engineering and manufacturing, service monitoring and control, and achieving reliable delivery performance. The new paradigm of manufacturing services requires new methods of operation control. The objective of this thesis is to propose to develop enabling technologies for collaborative planning and scheduling in semiconductor supply chains, and to enhance the controllability of supply chain operations and predictability of delivery performance. We developed an order-centric operation control model of uncertain job flow in supply chains. In order to enhance the service quality and reliability, we provide customers with the advanced information of delay orders under uncertainty by revised scheduling. On the other hand, we analyzed the factors of queuing delay, the effects of factors on mean flow time and its behavior. We applied scheduling methods to analyze the effects of lots arrival rate, capacity, and capacity loss due to uncertainty on mean flow time, and to simulate the queuing delay in dynamic system.