古典反應曲面法是用於解決未知反應函數曲面之最佳化問題，利用一階線性模式與二階數學模式，企圖求得影響最適反應曲面之獨立變數值，在古典反應曲面法之一階線性搜尋中，所採之最陡下降法，存在著鋸齒前進與收斂慢的缺點，可能增加實驗時間與成本；在二階求解方面，配適二階模式後即結束，並未企圖反覆進行二階模式的配適；在雜音方面，雜音（noise）將影響一階模式的配適，易造成一階搜尋方向的偏誤，若無法有效處理，將難以決定最佳搜尋方向，雜音亦對於二階求解最佳解造成影響，導致無法求得最佳解；實驗區域（design size）設計方面，一般以經驗或主觀決定實驗區域的大小，然實驗區域均會影響一階與二階模式的配適效果，亦值得深入探討。
本篇論文針對古典反應曲面法中，其一階模式之搜尋方面，利用指數平滑法觀念提出一個權重平滑法（Weighted Smoothing Method），使搜尋方向具記憶性，並考量雜音因素，以降低雜音對於搜尋方向的影響，進而探討實驗區域對搜尋效果與模式配適的影響，以設計一個實驗區域之改變規則，使其在雜音影響下，運用實驗區域的改變降低模式對雜音的敏感度，並討論其在一階線性搜尋與二階求解過程之使用時機。
本研究提出一個改良式反應曲面法，發展新一階搜尋方法，使其在雜音變異程度不一之影響下，具有良好估計方向的能力，進而依據配適統計資料與雜音影響程度，提供一個實驗區域改變的參考規則與使用時機。本研究改善古典反應曲面法的搜尋效率與求解品質，並進行演算實驗，以文獻上的數據來驗証效率與品質，並與其它改良之相關方法作比較，由演算實驗分析結果得知，在相近之搜尋效率下，改良方法一般可搜尋到較佳的求解品質，但搜尋效率則無顯著改善，若在相近之求解品質下，改良方法方可具較快的搜尋效率。整體而言，無論雜音影響大小，改良方法均較文獻上之平均方向策略與古典方法為優。

none

[1]曾珞萍，“以反應曲面法尋找多目標模擬模式之最佳解－以半導體封裝廠印字區為例”，國立成功大學，製造工程研究所，1999。
[2]Averill, M. L. and Kelton, W. D., Simulation Modeling and Analysis, 3rd edition. Mcgraw-Hill, United States, 2000.
[3]Bazaraa, M. S., Sherali, H. D., and Shetty, C. M., Nonlinear Programming: Theory and Algorithms, 2nd edition. Wiley, New York, 1993.
[4]Bisgaard, S. and Ankenman, B. , Standard errors for the eigenvalues in 2nd-order response surface models,” Technometrics, 1996, 38, 238-246.
[5]Bow, G. E. P. and Wilson, K. B., On the experimental attainment of optimum conditions. Journal of the Royal Statistic Society, 1951, B13, 1-38, discussion: 38-45.
[6]Box, G. E. P., The effects of errors in the factor levels and experimental design. Technometrics, 1963, 5(2), 247-262.
[7]Box, G. E. P. and N. R. Draper, Empirical Model Buidling and Response Surface, Wiley, New York, 1987.
[8]Brightman, H. J., Optimization through experimentation: applying response surface methodology. Decision Sciences, 1978, 9, 481-495.
[9]Brooks, S. H. and Mickey, M. R., Optimum estimation of gradient direction in steepest ascent experiment. Biometrics, 1961, 17, 48-56.
[10]Burden L. R. and Faires, J. D., Numerical Analysis, 6th edition, Brooks/Cole, United States, 1997.
[11]Draper, N. R., Ridge analysis of response surfaces. Technometrics, 1963, 5, 469-479.
[12]Edwards, I. M. and Jutan, A., Optimization and control using response surface methods. Computers & Chemical Engineering, 1997, 21, 441-453.
[13]Fletcher, R. and Reeves, C. M., Function minimization by conjugate gradients. Computer Journal, 1964, 7, 149-154.
[14]Hestens, M. R. and Stiefel, E. , Methods of conjugate gradients for solving linear systems. Journal of Research of the National Bureau of Standards, 1952, 48, 409-436.
[15]Hill, W. J. and Hunter, W. G., A review of response surface methodology: a literature survey,” Technometrics, 1966, 8, 571-590.
[16]Hoerl, R. W., Ridge analysis 25 years later. The American Statistician, 1985, 39, 186-193.
[17]Joshi, S., Sherali, H. D. and Tew, J. D., “An enhanced response surface methodology (RSM) algorithm using gradient deflection and second order search strategies. Computers and Operations Research, 1998, 25, 531-542.
[18]Myers, R. H. and Khuri, A. I., A new rrocedure for steepest ascent. Communications in Statistics-Theory and Methods, 1979, A8, 1359-1376.
[19]Myers, R. H. and Montgomery, D. C., Response Surface Methodology. Wiley,, New York, 1995.
[20]Myers, R. H., Khuri, A. I., and Carter, Q. H., Response surface methodology: 1966-1988. Technometrics, 1989, 31, 137-157.
[21]Pan, J. N. and Hwang, D. Y., Product and process quality improvement using a computerized response surface methodolody. Technical Report, 1994, 3, 1-5.
[22]Powell, M. J. D., Restart procedures for the conjugate gradient method. Mathematical Programming, 1977, 12, 241-254.
[23]Safizadeh, M. H., Minimizing the bias and variance of the gradient estimate in RSM simulation studies. European Journal of Operational Research, 2002, 136, 121-135.
[24]Salgado, C. A., Optimization by response surface methods. Transactions of the American Association of Cost Engineers, 1993, 4, 41-48.
[25]Sherali, H. D. and Ulular, O., A primal-dual conjugate subgradient algorithm for specially structured linear and convex programming problems. Applied Mathematics and Optimization, 1989, 20, 193-221.
[26]Sherali, H. D. and Ulular, O., Conjugate gradient methods using quasi-newton updates with inexact line searches. Journal of Mathematical Analysis and Applications, 1990, 150, 359-377.