臺灣博碩士論文加值系統

本研究主要目的為發展一套具有創新性之產能互換解決方案，可分析出精準的產能互換率來取代目前的經驗值及嘗試錯誤法。目前晶圓代工產業的作法分為兩個階段，第一階段先將產品種類高達六千多種的可能產品分成群組並選出少數代表產品來配置產能及建立可承諾量。第二階段在可能產品與代表產品間根據經驗設定一個產能互換率。晶圓代工廠於接單時將使用此產能互換率來配置訂單所需產能並即時提供交貨日期。若產能互換率低估則使得設備利用率降低，若高估則會使得設備產能超載且無法準時交貨。因此，本研究利用設備利用率的變動程度來分析出最佳產能互換率。首先，本研究利用Arena模擬軟體建構具有代表性的十二吋晶圓製造廠並進行模擬與驗證。接著，本研究設計一系列的產能互換率並將其使用於模擬模型中，再藉由平均絕對誤差(Mean Absolute Error, MAE)計算兩產品產能互換後其設備利用率的差異。將所設計之產能互換率經由模擬實驗找出對設備利用率影響最小的產能互換率，進而分析修正產能互換率之方向。研究結果顯示我們分析的最佳產能互換率能有效降低設備利用率的變動程度達0.2%-0.4%。若以台積電在2012年八吋晶圓產值高達台幣2,700億為例，此產能互換解決方案對晶圓代工公司具有相當大的影響力。

The major purpose of this research is to develop an innovative capacity swap solution focusing on determine an optimal capacity swap rate (CSR) for capacity swap problem which were currently solved based on experience and trial and error. The major value of this research is to determine an optimal CSR to improve the capacity utilization and on time delivery rate. Semiconductor foundry industry is capital intensive. While the capital expense of Taiwan Semiconductor Manufacturing Co. (TSMC) is expected to exceed US$ 10 billion in 2013, capacity utilization rate is critical to the capital effectiveness and profitability of TSMC. However, capacity planning is extremely difficult since TSMC has no products of its own. Currently there are over 6000 customer products from previous orders that TSMC has to be ready to take orders from. A route for advance technology product typically consists of over 1000 steps. It is extremely complicate for TSMC to plan capacity usage and generate Available to Promise (ATP) profile based on these routes. Current practice is to first group these over 6000 “possible routes” into clusters. Routes within the same cluster are more similar in terms of tools required and hence capacity can be swapped among these routes. A number of “representing routes” within each cluster will be selected and are used for capacity planning and generating ATP profile. The second step is to define a “capacity swap rate” (henceforth CSR) for each of the possible routes to represent the capacity consumption ratio between it and the representing route. Upon an order, TSMC will use the CSR to estimate required capacity and provide a delivery date in real-time. If the CSR is too low, then the tool utilization rate will be low. If the rate is too high, then the capacity will be overestimated and the product will not be delivered on time. Therefore, we hope the capacity utilization rate have no significant difference after the changeover of capacity between products.
This research builds a simulation models base on the designed dataset to represent the 12-inch fab. The model shows the real-world operation of the fab. The simulation model can be used to verify the feasibility of the resulting CSR from cluster analysis (CA) and principal component analysis (PCA). The simulation model also can potentially generate more accurate CSR than CA and PCA but with significantly more effort and time. This study uses mean absolute error (MAE) to measure the accuracy of CSR that impacts the difference of capacity utilization rate after the changeover of capacity between products. The results indicate that we can improve the variation of capacity utilization rate to 0.2%-0.4%. While TSMC’s total output reaches 14 million pieces of eight-inch equivalent wafers of market value at NT$ 27 billion in 2012, the proposed solution has significant impact on semiconductor foundry company’s revenue.

摘 要 i
ABSTRACT ii
誌 謝 iv
CONTENTS v
LIST OF TABLES vii
LIST OF FIGURES viii
Chapter 1 INTRODUCTION 1
1.1 Research background and motivations 1
1.2 Research objectives 4
1.3 Research scope and limitation 5
1.4 Research processes 5
Chapter 2 LITERATURE REVIEW 7
2.1 Semiconductor manufacturing in fab 7
2.2 Simulation 8
Chapter 3 RESEARCH METHODOLOGY 10
3.1 Simulation 10
3.2 Mean Absolute Error 13
Chapter 4 SIMULATION EXPERIMENTS 14
4.1 Design the dataset 14
4.2 Building a baseline model 15
4.3 Model verification 18
4.4 Definition of optimal capacity swap rate 20
4.5 Analyze the capacity swap rate 21
4.5.1 The definition of scope for swapping route 21
4.5.2 Verifying CSR and finding OCSR 22
4.5.3 Calculate the difference between capacity utilizations 23
4.5.4 Use paired t-test to compare MAEs 26
4.6 Obtain a SR-MMAE diagram 27
4.7 Result discussion 40
Chapter 5 CONCLUSIONS 41
LIST OF REFERENCES 44

[1]J. Chang and F. Huang, "Taiwan''s wafer production capacity highest in world in 2011 available at: http://focustaiwan.tw/ShowNews/WebNews_Detail.aspx?Type=aALL&ID=201201150044 [Accessed 2012/12/5 ]," 2012.
[2]K. Liu, "TSMC Said to Set Capex at US$10B. for 2013 Amid Apple Contract Report available at: http://cens.com/cens/html/en/news/news_inner_41368.html [Accessed 2013/01/10]," 2012.
[3]C.I. Hsu and H.C. Li, "An integrated plant capacity and production planning model for high-tech manufacturing firms with economies of scale," International Journal of Production Economics, vol. 118, pp. 486-500, 2009.
[4]L. Ellis and Bish, "Scheduling of wafer test processes in semiconductor manufacturing," International journal of production research, vol. 42, pp. pp. 215-242, 2004.
[5]C.F. Chien, C.H. Wu, and Y.S. Chiang, "Coordinated capacity migration and expansion planning for semiconductor manufacturing under demand uncertainties," International Journal of Production Economics, vol. 135, pp. 860-869, 2012.
[6]C.F. Chien, Y.J. Chen, and J.T. Peng, "Manufacturing intelligence for semiconductor demand forecast based on technology diffusion and product life cycle," International Journal of Production Economics, vol. 128, pp. 496-509, 2010.
[7]M.C. Wu and W.J. Chang, "A short-term capacity trading method for semiconductor fabs with partnership," Expert Systems with Applications, vol. 33, pp. 476-483, 2007.
[8]C.H. Hsieh, C. Cho, T. Yang, and T.J. Chang, "Simulation study for a proposed segmented automated material handling system design for 300-mm semiconductor fabs," Simulation Modelling Practice and Theory, vol. 29, pp. 18-31, 2012.
[9]J. F. O’Kane, J. R. Spenceley, and R. Taylor, "Simulation as an essential tool for advanced manufacturing technology problems," Journal of Materials Processing Technology, vol. 107, pp. 412-424, 2000.
[10]Y. Kuo, T. Yang, B. A. Peters, and I. Chang, "Simulation metamodel development using uniform design and neural networks for automated material handling systems in semiconductor wafer fabrication," Simulation Modelling Practice and Theory, vol. 15, pp. 1002-1015, 2007.
[11]M. Lutz, X. Boucher, and O. Roustant, "Information Technologies capacity planning in manufacturing systems: Proposition for a modelling process and application in the semiconductor industry," Computers in Industry, vol. 63, pp. 659-668, 2012.
[12]D. Nazzal, M. Mollaghasemi, and D. Anderson, "A simulation-based evaluation of the cost of cycle time reduction in Agere Systems wafer fabrication facility- a case study," International Journal of Production Economics, vol. 100, pp. 300-313, 2006.
[13]I. Mahdavi, B. Shirazi, and M. Solimanpur, "Development of a simulation-based decision support system for controlling stochastic flexible job shop manufacturing systems," Simulation Modelling Practice and Theory, vol. 18, pp. 768-786, 2010.
[14]J. Wang, Q. Chang, G. Xiao, N. Wang, and S. Li, "Data driven production modeling and simulation of complex automobile general assembly plant," Computers in Industry, vol. 62, pp. 765-775, 2011.
[15]S. Su and C.L. Shih, "Managing a mixed-registration-type appointment system in outpatient clinics," International Journal of Medical Informatics, vol. 70, pp. 31-40, 2003.
[16]K. Concannon, M. Elder, K. Hindle, J. Tremble, and S. Tse, Simulation modeling with SIMUL8. Canada, 2007.
[17]P. Flom, "How to Calculate Mean Absolute Error, available at: http://www.ehow.com/how_6426845_calculate-mean-absolute-error.html#ixzz2Px7dU5cQ [Accessed 2013/03/10 ]," 2010.
[18]J. Yang, D. J. Greenwood, D. L. Rowell, G. A. Wadsworth, and I. G. Burns, "Statistical methods for evaluating a crop nitrogen simulation model, N_ABLE," Agricultural Systems, vol. 64, pp. 37-53, 2000.
[19]C. J. Willmott and K. Matsuura, "Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance," CLIMATE RESEARCH, vol. 30, pp. pp.79-pp.82, 2005.
[20]Y.H. Yang, "Using cluster analysis and principal component analysis for capacity swap in the semiconductor foundry industry," National Taipei University of Technology, 2013.