前端主導下半導體供應鏈層級存貨監控機制

Forward Echelon-based Inventory Monitoring in Semiconductor Supply Chain

林宏文(Hon-Wen Lin)

指導教授 ： 蔣明晃

共同指導教授 ： 郭瑞祥(Ruey-Shan Guo)

國立臺灣大學/工學院/工業工程學研究所/碩士(2007年)

https://doi.org/10.6342/NTU.2007.00722

摘要

半導體產業為一具有高科技、高度資本密集兩種特性的產業，且其整體生產流程繁雜，尤其半導體前端製造廠中的製程具有再迴流作業、批量生產、派工排程、產品等候時間限制以及環境要求等種種特性，使得製造廠在晶圓生產上扮演著重要角色，也因此造成供應鏈內部成員力量大小不一；再加上半導體供應鏈的高度垂直分工特性，上下游不同的企業所追求的目標並不相同，使得整個供應鏈分割為許多片段，導致整體供應鏈之績效不佳，以及難以對整個供應鏈進行監控。因此，為滿足相關的服務水準之要求，半導體供應鏈必須依據其生產特性、資本密集程度與各成員力量大小不一的特性，以前端成員來主導整合半導體供應鏈前後端製程，以提昇顧客的滿意度與改善整體供應鏈的效率。
本研究主要依據半導體的相關特性，發展以前端為出發點之層級存貨量(Echelon Inventory)概念並結合CONWIP系統控制方式，建立以前端為主之半導體供應鏈監控機制。本研究考量生產系統之特性以及CONWIP系統控制方式之影響，依據服務水準目標與層級在製存貨(Echelon WIP Inventory)之間的關係，建構以前端為主之二層級在製存貨模型以滿足所需的服務水準；接著再利用二元搜尋法來提高內部層級的服務水準與遞迴求解來發展演算法，找尋適當的內部服務水準以降低整體在製存貨上界。並利用求解模型所得到之各層級在製存貨上界作為監控指標，經由模擬驗證本研究所設定的以前端為主之二層級在製存貨監督上界之有效性。經由模擬驗證的結果有下列結論：
(1)藉由本研究提出以遞迴求解以及提高內部層級服務水準之演算法，可有效得到各層級在製存貨上界，其結果亦小於階段觀點之整體在製存貨上界；
(2)相較於過去研究，本研究提出以前端為出發點所發展的層級存貨概念，能使整體供應鏈系統在較低在製存貨量的情況下，得到較佳的服務水準。這主要是因為利用本身與整體資訊來進行規劃，使系統中在製存貨配置之不同所致。

關鍵字

半導體供應鏈 ； 層級在製存貨 ； CONWIP

並列摘要

The semiconductor industry with its high capital investment and hi-tech nature faces fierce competition and quick environment changes. Furthermore, semiconductor manufacturing system is a very complicated multistage production process, transferring silicon in a form of polished disk into integrated circuits. Especially in wafer fabrication, it is time-consuming and complicated one, consisting of six major types of process: Diffusion, Lithography, CVD, Thin Film, Etching and Ion Implantation. With the characters of re-entrant production operations, high capital investment and hi-tech nature and vertical disintegration, each member of supply chain has different might and pursues for different goals. Therefore, this results in the poor performance of supply chain and the difficulty in monitoring it. In order to enhance the service levels and improve the performance of the whole supply chain, it is essential for the front-end member to lead all members to share their information and synchronize their operations, i.e., there is a need to integrate the front-end and back-end in the semiconductor supply chain to meet customers’ on-time delivery requirement and service level.
This research develops concept of forward echelon-based inventory and uses CONWIP system to develop a semiconductor supply chain monitoring scheme. The first step in the proposed scheme is to construct a forward echelon-based two-echelon model which takes the features of the production system into consideration with the goal of minimizing the echelon WIP inventory control limits of the supply chain under the service level constraints. Next, the model is solved iteratively through raising the service level of the internal echelon. The algorithm searches the proper internal service level with the concept of binary search. With the resulting echelon WIP inventory control limits from the model, the proposed monitoring scheme has been validated through simulation study.
Based on the simulation validation results, the conclusions are drawn as follows:
(1) By raising the service levels of internal echelon, the proposed scheme can effectively derive WIP inventory limits of two echelons, which is lower than the stage-based WIP inventory limits;
(2) Compared to the traditional stage-based inventory monitoring scheme, the proposed forward echelon-based monitoring scheme can obtain a higher service level under the lower inventory levels. This is the result of allocation of the WIP limits and the utilizing of overall information.

並列關鍵字

Semiconductor Supply Chain ； Echelon WIP Inventory ； CONWIP

參考文獻

[1] 莊貴惠, 「半導體供應鏈多層級存貨監控機制」, 國立台灣大學工業工程學研究所碩士班論文, 2005

[2] 楊博鈞, 「運用CONWIP概念於半導體供應鏈多層級存貨監督機制」, 國立台灣大學工業工程學研究所碩士班論文, 2006

[3] 鄭孟哲, 「半導體供應鏈路徑配置之多目標最佳化模式」, 國立台灣大學工業工程學研究所碩士班論文, 2005

[4] Abdoli, M. and F.F. Choobineh, “Empirical Bayes forecasting methods for job flow times,” IIE Transactions (Institute of Industrial Engineers), 37, 7, (2005), 635-649

[5] Asmundsson, J., Rardin, R. L., and Uzsoy, R., “Tractable nonlinear production planning models for semiconductor wafer fabrication facilities,” IEEE Transactions on Semiconductor Manufacturing, 19, 1, (2006), 95-111