以產品測試建立3C凸轂之多品質優化設計研究

OPTIMIZATION DESIGN OF MULTI-CRITERIA 3C BOSS THROUGH PRODUCT TESTING

林源憶(Yuan-Yi Lin)

指導教授 ： 王明庸

大同大學/工程學院/機械工程學系所/碩士(2008年)

摘要

本研究針對塑膠製品中凸轂的尺寸設計參數與品質考量問題，而凸轂的強度常為設計時所追求的目標，本研究以建立塑膠射出產品之凸轂多品質優化設計研究，在多目標考量下尋求最適凸轂參數組合。
本研究針對凸轂所能承受的力量為品質的考量下，透過倒傳遞神經網路，建立品質預測機制；針對特定品質目標(軸向容許力、側向容許力、最大鎖附扭力)，使用基因演算法結合倒傳遞類神經網路尋求特定品質目標之設計參數組合(外徑、內徑、高度、螺絲尺寸)；使用田口法與理想解類似度順序偏好法(TOPSIS)整合多品質目標，在多目標考量下，決定最適參數組合，建構一套可靠之凸轂品質預測機制。以賽局理論進行任兩人(雙品質)之協議賽局，經議價賽局概念策略協議上之重點，在優先策略的組合下即可找出最佳協議組合，以協調衝突獲取多人(多品質)多策略之優化方案，發展並確認多品質3C凸轂設計之創新優化機制。
結合類神經網路與基因演算法技術，建立品質預測機制，在提供數據範圍內皆有9成的準確度，經由賽局理論優化結果在整體優勢排名第二，效果相當不錯。藉由本研究之成果，建立凸轂設計參數之參考依據，並提供於實務作業，避免不適設計，以減少試模的次數與費用，縮短業界塑膠產品射出模具的開發時間，有效減少成本，提昇產業競爭力。

關鍵字

凸轂 ； 倒傳遞類神經網路 ； 基因演算法 ； 田口實驗法 ； TOPSIS ； 賽局理論

並列摘要

This research considers the problem to size design parameter and quality of the Boss in the plastic products, and the strength of Boss often is design pursue of target, this research with set up the plastic projection product of multi-criteria Boss optimization design, under many target considerations, seek the parameter alignment of the optimum Boss.
This research aims at the force that the Boss cans bear to descend for the consideration of quality , through Back-Propagation Neural Network set up the mechanism of quality predicts. Aim at a particular quality target(allowed axial tensile force, side tensile force, maximum attaching torque force), use Genetic Algorithm to combine Back-Propagation Neural Network of the combination of the design parameter that looks for a particular quality target(outside diameter, inside diameter, height, screw size), Use the Taguchi Method and TOPSIS theory to combine many quality target, under many target considerations determine the optimum parameter alignment, Build the reliable Boss quality of predicts mechanism. Take the post as two people (double quality) in Game Theory of the agreement Game, In the concept tactics the focal points of agreement through bargaining position game, it can last agreement associations best at associations of tactics preferential,in order to coordinate conflict obtain to many people(Much quality) of the optimization scheme of many tactics, develop and confirm many quality 3C Boss were designed innovate the mechanism of optimizing.
Combining type neural network and Genetic Algorithm technology and set up the quality predicts mechanism, all there is ninety percent accuracy in offering the range of data, it ranks the second in the whole advantage optimizing the result via being comparable to the Game theory, the result is pretty good. With the achievement of this research, set up the reference basis of the Boss design parameter, prevent the improper design, in order to reduce and try number of times and expenses of the mold, shorten construction period when the plastic products of industry shoot the mold. Reduce the cost of effectively and promote industry's competitiveness.

並列關鍵字

Boss ； Back-Propagation Neural Network ； Genetic A

參考文獻

[1] 廖哲新，”塑膠成型製品中凸轂強度設計之研究”，大同大學機研所碩士論文，2007

[2] Hwang, M. W., Kim, M. H., and Choi, J. Y. ,“Second-order Multilayer Perceptrons and Its Optimization With Genetic Algorithms.”, Proceedings of the 2000 Congress on Evolutionary Computation,Vol. 1,pp. 652-658, 2000.

[3] Enab, Y. M. ,“Genetic Algorithm for Identifying Self-generating Radial basis Neural Networks.”, Fourth International Conference on Artificial Neural Network, pp. 65-70 , 1995.

[4] Petridis, V., Paterakis, E., and Kehagias, A. A. ,“Hybrid Neural-genetic Multimodel Parameter Estimation Algorithm.”, IEEE Transactions on Neural Networks, pp. 862-876, 1998.

[5] G.C. Luh and C.Y. Wu ,“Non-Linear System Identification Using Genetic Algorithms.”, Proc. Instn Mech. Engrs, Part I, Journal of Systems and Control Engineering, pp. 105-117, 1999.