Coupled ANN-Based Prediction of a Foam Expansion Ratio Using an FEM Computational Scheme to Predict the Material EVA Expansion Shape

結合類神經預測發泡膨脹倍率並應用於有限元素方法預測EVA材料之發泡膨脹形狀

鄭宜仁(Yi-Ren Jeng) ； 劉德騏(De-Shin Liu) ； 姚宏宗(Hong-Tzong Yau)

《中國機械工程學刊》 41卷4期 (2020/08) Pp. 409-416

摘要

Accurately predicting the shrink mold shape of direct injection-expanded foam molding is a difficult and important task. This molding method is widely used by sports shoe sole manufacturers to create shock-resistant material. The relationship between the normal thickness of a shoe sole and the expansion ratio of ethylene vinyl acetate (EVA) foam material is crucial to obtaining the correct shrink mold size. This study conducts experiments involving a series of small rectangular specimens with varying z thickness, modified by Artificial Neural Network method (ANN), to qualify the micro expansion ratio along the x, y, and z directions through the normal thickness. Using this relationship as the heat expansion criterion, the expansion of foam could be simulated by the finite element method. Finally, two shoe sole types were used to verify the algorithm. The discrepancy between simulation results with the original designed CAD model in the y-direction was less than 4 mm, meeting the requirements of the shoe sole factory. Thus, the proposed method can be used to predict the shrink mold size rapidly and improve the traditional trial-and-error method.

關鍵字

Finite element ； thickness ； expansion ratio ； Neural Network

並列摘要

對於直接射出發泡製程，準確的去預測縮小模的形狀是一件困難且重要的工作。本文所探討的發泡成型方法，已廣泛運用在運動鞋業之中底部件的製造，可生產具抗震效果的材料。而球鞋法線厚度與EVA材料膨脹倍率之關係，是提供給原設計三維幾何設計模型，進行縮小模計算，並獲得正確縮小模穴的關鍵。本研究藉由隨z軸厚度變化的方形試片，並透過類神經訓練及修正，而得到正確的法線厚度與微觀膨脹倍率關係，從而分析出有限元素之熱膨脹分析法中，各個元素在x, y,及z方向的微觀膨脹倍率。最後我們使用有限元素法，針對兩款球鞋中底來進行模擬以及驗證，結果顯示模擬與原設計的CAD模型比對，y軸的誤差範圍4mm以內，符合鞋廠的需求，故使用本研究之方法可以快速的預測縮小模的形狀，改善傳統嘗試錯誤法的缺點。