考慮需求不確定和等候時間限制之機台組態決策

Abstract

本研究探討晶圓製造的擴散區之機台組態規劃問題。擴散區之生產情境如下：(1)需求不確定，含需求總量和產品組合的不確定；(2)生產系統是三階段的流程式生產線，在第一及第二階段，各機台的加工單位為單批加工，第三階段為批次加工，相同產品的工件才可並批；(3)各工件在各階段有等候時間限制，若工件在等候區中的等候時間超過上限時，則須重回第一階段重新加工。本研究提出系統化的方法，考慮缺貨與機台折舊成本，以期望總成本最小化為目摽，求解擴散區各階段最適的機台數量。 首先建立一等候網路模型(MSTCB模型)，在給定的機台組合、需求總量情境與產品組合情境下，考慮各階段的重工率限制(等候時間限制)，以最大產出率為目標，估算產出量與機台利用率。並設計電腦模擬程式驗證MSTCB模型的精確度。 隨之構建機台組態規劃模型(TP-MDMX模型)，由MSTCB模型輸入各機台組合下，各種需求情境下的產出量及機台利用率，由此構建各需求情境的需求量、產出量及缺貨量的關係式，缺貨懲罰成本的風險程度限制關係式及機台利用率限制關係式。以期望總成本最小化為目摽，求解各階段最適的機台數量。

This study addresses the tool portfolio planning of the diffusion zone in in wafer fabrication. The production environment of the diffusion zone is as following：(1) There are uncertainties in aggregate demand and product mix.(2) it is a three-stage flowshop problem. Each stage has several parallel machines. Machines in the third stage are batching machines, which can handle up to b jobs of the same product type simultaneously.(3) There are queue time limits before the second stage and the third stage. Jobs which waited for more than the queue time limits should be reworked. At first, we constructed a queuing network model, called MSTCB model, which considers the waiting-time constraint and set the objective as maximizing the output rate to estimate the yield and the utilization of machines under certain tool portfolio, aggregate demand scenarioand product-mix scenario. Then, we address a simulation program to test and verify the accuracy of the MSTCB model. After that, we proposed a tool portfolio planning model, called TP-MDMX model, which consider the relationship between demand, yield and shortage, the risks of shortage cost and the utilization of machines, to determine number of parallel machines in each stage with the objective of maximizing the expected total cost.