以修正雙界法構建晶圓製造廠投料及派工法則

Abstract

半導體晶圓製造是目前國內具競爭力的產業之一，但因其製程複雜，以致造成其在生產管理上的困難，因此有必要針對晶圓製造廠的特性發展一套良好的投料及派工法則。而當景氣不佳時，能適時地滿足客戶交期便成為企業的一項競爭優勢，因此本研究之首要目標在於充分降低交期延誤，在此目標之下，進一步充分利用限制資源的產能、降低現場在製品數量及平均生產週期時間。 由於投料法則相對於派工法則在生產績效上影響較大，因此為了避免不必要的投料動作，吾人採取更嚴謹的投料法則，亦即在投料法則上，吾人結合了雙界法(TB)及提拉式生產控制系統的投料理念為本研究的投料法則，亦即利用看板系統、限制資源(CCR)在第一個層級的計劃產出量及計劃在製品量兩項參數，來共同控制系統的投料動作。 在派工法則方面，吾人以雙界法(TB)的派工理念為基礎，並依據工作站是否為限制資源(CCR)而有不同的派工模式，此外，派工模式中並考量了訂單權重、各層級之在製品量是否達到均勻化等因素。 經由實例驗證，本模式不僅在降低交期延誤上有良好的表現外，另在產出量、平均生產週期時間、平均等候時間與平均在製品數量等方面，亦有良好的績效表現。

Wafer Fabrication is a highly competitive industry today. Its manufacturing process is so complex to manage. Knowing how to develop a good releasing and dispatching rule according to the production characteristic is one of most valuable know-how of a wafer fabrication factory. Being able to meet the customer's delivery is an important competitive advantage to compete in the industry. Therefore, the main goal of this thesis is to develop a new model of production which will enable IC makers to reduce the delay about the customer's delivery date, increase utilization of capacity with constraint resources, and lower down the wip level in the shop floor and the flow time of production. Because the releasing rule have more influence on the performance of production than that of dispatching rule, the thesis takes a more strict releasing rule to avoid unnecessary material release. The thesis combines the releasing idea of TB method and Pull-based production control system to construct a releasing rule. It means to control the release operation by Kaban-System and two parameters, planned throughput and wip on the first layer of the constraint capacity resource. Concerning dispatching rule, the thesis develops a dispatching rule on the base of the dispatching idea of TB method. A different dispatching model can be generated according as the workstation is a constraint resource or not. The thesis also considers some other factors, such as, the weight of the orders, and the balancing of the wip level on every layer of the constraint resource. The findings of this research show that the new model has excellent performance on production, i.e., reducing the delay about the customer's delivery date, increasing throughput, reducing flow time of production, lowering queuing time and the wip level.