神经网络和有限元方法在两轴柔性滚弯中的应用

针对两轴柔性滚弯回弹难预测的特点,本文采用有限元和神经网络技术建立两轴柔性滚弯的工艺参数和滚弯零件直径之间的映射关系。有限元用于实现对滚弯过程模拟,获取神经网络预测模型的训练样本集;人工神经网络用于建立预测模型,获取两轴柔性滚弯的预测值。用实验值检验模型的预测值,通过比较,发现两者吻合良好,仅存在较小差异。这证明该方法是有效的,可以为实际生产过程中参数的选择提供有力的帮助。

Application of Neural Networks and FEM to Two-axle Rotary Shaping

Considering that it is difficult to predict the springback in the two-axle rotary shaping,the FEM and artificial neural networks(ANN) were used to set up mapping relations between the process parameters and the part′s diameter.The FEM was used to simulate the two-axle rotary shaping so as to obtain rotary shaping training sets for the ANN prediction model.The ANN was used to set up a prediction model for prediction values which were tested by experimental data.A comparison indicates that the prediction value and experimental data agree well,having only small errors.This indicates that the model is effective and helpful for parameter selection in production pratices.