本文結合類神經網路與有限元素法對長方杯引伸作胚料形狀最佳化預測,使成形杯只具有少許的耳化,而此耳化現象主要係因胚料形狀之不恰當。研究以不同杯高與各種長短軸比為製程參數,利用有限元素法分析不同參數下之材料流動特性,然後將分析結果與類神經網路結合,建立胚料形狀最佳化預測模型,最後驗證以此模組所預測之胚料形狀。本研究亦實際製作模具進行引伸,並比較理論預測結果與實驗結果之差異,發現所建立之胚料形狀預測模型在淺杯引伸時有合理的準確性,但在較深杯引伸時,由於角隅部位產生摺疊使得成形杯緣產生劇烈之高低落差,準確性稍差。
In this paper, the finite element method is employed in conjunction with the abductive network to predict the optimum blank contour of a rectangular cup with ear-less in deep drawing. The earring behavior of the drawn cup is caused primarily by the unsuitable shape of blank. Different cup heights combined with various aspect ratios of the rectangular cup are taken into account as the process parameters in this study. A finite element-based code is utilized to investigate the material flow characteristics under different process parameters, and the abductive network is then employed to synthesize the data sets obtained from numerical simulations, thus establishing a predictive model. From this model, an optimal blank contour for producing a rectangular cup with ear-less can be found. Experiments have been carried out with low carbon steel sheet blanks at room temperature for some cases. The theoretical results for the drawn cup heights agree well reasonably with the experimental data.