The Study of a Successive Taguchi Method Applied in Globally Optimizing Problems

View Statistics Email Alert RSS Feed

Information

Details

Project title

Code/計畫編號

NSC98-2221-E019-065-MY2

Translated Name/計畫中文名

循序式田口實驗法應用於問題全域最佳化研究

Project Coordinator/計畫主持人

Jeun-Len Wu

Funding Organization/主管機關

National Science and Technology Council

Department/Unit

Department of Systems Engineering and Naval Architecture

Website

https://www.grb.gov.tw/search/planDetail?id=2009973

Year

2010

Start date/計畫起

01-08-2010

Expected Completion/計畫迄

01-07-2011

Bugetid/研究經費

351千元

ResearchField/研究領域

機械工程
工業工程

"有鑑於田口方法之廣泛應用於業界與學術界，令人不免質疑田口方法所得到之區域最佳解與全域最佳解差距有多大? 追蹤相關研究文獻，筆者發現目前並無相關文獻對此問題探討，是否受限於全因子實驗方法不易進行？因此本系郭信川副教授與筆者乃於近日發展一套新又有效率之因子實驗運算方法，稱為循序式田口因子設計實驗法(Successive Taguchi Method in Design of Experiment, STMDE)，本套方法兼顧因子實驗設計之效率與準確度，亦即能以很少之實驗組配試驗而得到全域最佳組配。本方法以” A New Approach with Orthogonal Array for Global Optimization in Design of Experiments , ”一文投搞至Journal of Global Optimization (SCI, impact:0.813, ranking 63/165 )，已被接受，將於近期刊登。爲使本套開創演算法應用上更爲全面性及可靠性，本研究計畫擬以二年期間進行如下之規劃：第一年計畫: 執行因子實驗設計法，直交表可謂影響實驗結果最主要關鍵。直交表格式與因子數目及各因子之水準數目息息相關，目前普遍被引用之田口直交表有三種類型分別為兩水準(,,,,)，三水準,,，二與三水準混合類型,,因這些分類侷限於二與三水準之設計若遇到水準數目大於三者或因子個數並無適合直交表可應4L8L16L32L64L LLLLLL 用時，此時研究者或設計者將面臨挑戰---推導一為問題可用與可靠之直交。STMDE之另一特點就是其不受限上述問題之限制。因此第一年研究計劃為將STMDE擴展爲多選擇數據(# of levels bigger than 3)参數因子問題求解，執行過程除選擇benchmark問題驗證外，並與Lee & Park[11] 研究比較。Lee&Park 迭代執行直交表方式，發展一套最佳化演算法。我們發現其每次直交表之區域解，若能引用筆者所開創之演算法，相信可大幅降低其數值迭代次數，尤其在高維問題時效率性將更為突顯。第二年計畫: 擬配合模糊集合理論與模糊決策模式，將本套方法發展處理多目標問題之求解。目前相關研究大多配合田口方法進行之，惟田口方法於每一目標所得非全域最佳解，導致計算時間長久而且問題最終解也非全域最佳解。相信以STMDE應用於多目標問題研發，不但大幅降低運算時間同時可保證求得是多目標問題之全域最佳組配解。同樣的我們將藉一工程問題驗證本方法之優越點。由於本套方法能以非常少的成本達到全域之最佳解，毫無疑問，本計劃完成工作結果，將可做爲田口方法使用者非常重要及有意義之參考價值。" "Taguchi Method has been widely used in industry and academic work to improve design process. However the solution obtained by Taguchi method is technically not a global but a local optimum. The difference between both solutions for a designing process of interest, to author’s knowledge, has never been studied because the global one has to be determined through a full factorial approach, which has been prohibitively conducted due to a tremendously huge amounts of optional settings. Lately, the author has developed an approach on a design of experiments able to obtain the global optimum with much less work than the full factorial approach. Hence if the present approach is employed by users conducting a design of experiments, we believe our approach will boost their performance dramatically. Since the developed approach provides a global optimum efficiently and excels Taguchi approach, we’d like to propose a two-year study of the developed approach while applied in different works and further highlight its merits in optimum design. At the first year, we will extend the proposed approach in solving problems of multiple-level (bigger than 3) factors. The work will demonstrate its efficiency and effectiveness in comparison with that by Lee & Park who used the local best solution determined by each orthogonal array to iteratively approach the problem’s solution. Certainly their solution was not a global optimal one to the problem. In the second year we would expand the proposed approach in optimizing multi-response problems in collaboration with Fuzzy set theory and Fuzzy decisive modal. An engineering problem will be used to demonstrate the proposed approach’s performance believed to outperform Taguchi method."

DSpace CRIS

The Study of a Successive Taguchi Method Applied in Globally Optimizing Problems

Details

Description