晶圓製造廠區段基礎式週期時間估算法

Abstract

週期時間是生產規劃過程中的重要元素，然而由於晶圓製造廠製程複雜性與機器設備多樣性等因素的交互影響，故難以精確掌握週期時間的變化。本文之研究目的，即在於發展出一個具備高預測準確性與高即時反應能力之週期時間估算法，以提昇晶圓製造廠週期時間預估能力。 本文所發展之週期時間估算法，係以晶圓廠物流分析為基礎，以確認晶圓批等候時間之形成因素。透過實際觀察與分析，吾人發現物流阻塞的主要原因，來自於關鍵工作站的高負荷量，及批次工作站的載入批量限制。由於上述兩項因素所對應之等候時間形成特性，在本質上有所差異，因此，吾人將上述兩種等候時間形成因素，定義為「負荷因子」與「批量因子」，並分別運用M/M/c等候模型及批量因子流程時間估算式(Batching-factor flow time estimation algorithm; BFFT)來進行估算。上述結合負荷因子與批量因子估算觀念的方法，統稱為區段基礎式週期時間估算法(Block-based cycle time estimation methodology; BBCT)。 BFFT演算法之基本運作程序，共分為三個階段。第一階段為「分解與辨識」階段，其目的為依據物流相似性，將複雜的製程「分解」成為獨立的區段與子區段，並「辨識」出各子區段所對應之流程時間模型；第二階段為「基本運算」階段，其目的為發展各子區段流程時間估算之數學式，並進行估算；第三階段為「組合運算」階段，其目的為將各子區段流程時間「組合」成為區段流程時間，乃至整個製程之流程時間。最後，再組合負荷因子等候時間，成為週期時間值。此外，吾人亦將晶圓製造廠中普遍存在的工程實驗工件批，納入週期時間預測模式中。 經由實例驗證結果顯示，BBCT法之週期時間預估準確度，不論是在系統高負荷(90%)、中負荷(70%)或低負荷(50%)之情況下，均顯著優於ConwaySu估計式[87]與M/M/c等候模型。在具實驗工件批之生產環境中，BBCT法亦同樣表現出優於另外兩種方法週期時間之預估能力。此外，BBCT法在工作站利用率推估上，亦具有相當不錯的成效。

Cycle time plays a critical role in production planning. However, the interactive effects between some factors such as process complexity and variety of workstations, leads to the difficulty in estimating cycle time. The purpose of this dissertation is to develop a cycle time estimation algorithm with high estimation accuracy and response capability. The core base of cycle time estimation is to find out the reasons for material flow congestion, caused by the heavy machine loading and batch lots accumulation. Recognizing that the batch size difference among machine types is a major factor contributing to the congestion of production flow, the process flow of each product type is divided into the same number of blocks as that of the batch machines passing through. This dissertation thus develops an block-based cycle time (BBCT) estimation algorithm to estimate the product cycle time, consists of the waiting time due to batching and loading factor in every block. The corresponding waiting time was estimated using the M/M/c queueing model and batch factor flow time (BFFT) estimation algorithm. The basic BBCT estimation algorithm is developed for the system only producing normal class products. Considering engineering lots usually existed in shop floor, a revised BBCT estimation algorithm is proposed also. To verify the performance, a simulation model is constructed to evaluate the accuracy of cycle time estimation with BBCT algorithm. The experiment study showed that the BBCT performance was better than that of M/M/c queueing model and of the revised Conway’s formula by Su (1998) under the system environment of higher loading (90%), medium loading (70%) and lower loading (50%) for all normal class scenario and for engineering lots existing scenario. The related experiments also showed that the BBCT could effectively estimate the workstation utilization rate. BBCT thus is a good tool for production planning and control.