應用田口法與適應性類神經模糊推論系統於塑膠射出成型之多目標最佳化研究

Application of Taguchi Method and Adaptive Network-Based Fuzzy Inference System to Multi-Objective Optimization of Plastic Injection Molding

周格仲(Ge-Zhong Zhou)

指導教授 ： 葉豐輝

淡江大學/工學院/機械與機電工程學系碩士班/碩士(2017年)

https://doi.org/10.6846/TKU.2017.00848

摘要

本文係使用田口法與適應性類神經模糊推論系統於探討塑膠射出成型最佳化問題，並應用多目標決策使最佳化結果能夠選擇多樣性，以達到多目標最佳化之目的。研究中使用繪圖軟體進行實體繪製，接而使用Moldex3D執行模流分析，選擇材料溫度、模具溫度、填充時間、保壓壓力、保壓時間與冷卻時間做為控制因子，針對縫合線長度、翹曲變形、體積收縮率與週期時間等目標做探討。最佳化首先導入田口法利用直交表配合信號雜訊比與變異數分析，獲得一組較佳射出參數與因子的影響程度，由分析結果顯示，保壓時間為這六個因子中影響最為顯著的因子。其次將所得到之結果，再導入適應性類神經模糊推論系統，以得到最佳化加工參數組合。最後運用多目標決策方法獲取各目標間得以平衡的加工參數組合，以達多目標最佳化研究之目的。
　　由模擬分析結果顯示，翹曲變形最佳化可同時達成體積收縮率最佳化，但週期時間須較長，如翹曲變形最佳化為0.089mm，體積收縮率僅為0.874%，但週期時間為51.909秒。由於多目標無法達成所有目標值同時最佳化，故使用多目標決策法，利用每個單一目標下所得之最佳化結果，在翹曲變形與體積收縮率可接受的條件下，降低週期時間，得到一組多目標最佳化加工參數，如翹曲變形為0.251mm，體積收縮率為1.860%，而週期時間降為46.842秒。故本研究可證實田口法與適應性類神經模糊推論系統和多目標決策能有效應用於塑膠射出成型之多目標最佳化。

關鍵字

田口法 ； 適應性類神經模糊推論系統 ； 塑膠射出成型 ； 最佳化

並列摘要

The thesis employs Taguchi Method and Adaptive Network-Based Fuzzy Inference System (ANFIS) to explore the optimization of plastic injection molding, and uses multi-objective decision-making method to diversify the optimization results and thereby achieve the purpose of multi-objective optimization. In the study, the drawing software is used to perform the tangible objects drawing, and then Moldex3D is used to perform the mold flow analysis. The material temperature, mold temperature, filling time, packing pressure, holding time and cooling time are selected as control factors to investigate the warpage, volume shrinkage and cycle time. As to the optimization, the orthogonal array of Taguchi method is applied to examine signal-to-noise ratio and analysis of variance. The influence degree of factors and the better parameters of injection molding are obtained. According to the analysis results, holding time had the most significant influence out of the six factors. Besides, ANFIS is incorporated into the acquired results to obtain an optimized parametric combination. Finally, the multi-objective decision-making method is used to acquire a parametric combination that serves to balance all objectives and to achieve the purpose of multi-objective optimization research.
The results of the simulation analysis show that the optimization of the warpage can be achieved at the same time to optimize the volume shrinkage. But the cycle time grows longer. For example, the warpage is optimized to be 0.089mm, the volume shrinkage is only 0.874%, but the cycle time is 51.909 seconds. With many objectives to be optimized, it is unlikely to optimize all objective values simultaneously. Therefore, using the optimization results acquired from each objective, the multi-objective decision-making method is applied to minimize the cycle time under the condition that the warpage and the volume shrinkage are acceptable, and a set of multi-objective optimization parameters are obtained. For example, the warpage is 0.251mm, the volume shrinkage is 1.860%, and the cycle time is reduced to 46.842 seconds. As proven by this study, Taguchi Method, ANFIS and multi-objective decision-making method can be effectively applied to multi-objective optimization of plastic injection molding.

並列關鍵字

Taguchi Method ； Adaptive Network-Based Fuzzy Inference System ； Plastic injection molding ； Optimization

參考文獻

2. 陳坤村，“應用田口法與遺傳演算法於點字觸摸顯示方塑膠零組件射出成型最佳化參數研究”，碩士論文，淡江大學機械與機電工程學系，2009。

6. 陳采玲，“成型條件之交互作用對光碟機托盤翹曲的影響”，碩士論文，國立臺北科技大學製造科技研究所，2013。

臺灣科技大學機械工程系，2006。

20. 科盛科技股份有限公司，“Moldex3D 模流分析技術與應用”，2007。

22. 科盛科技股份有限公司，“真實三維模流分析技術與應用”， 科盛