自组织特征映射神经网络在薄板冲压中的应用

从影响薄板冲压成形结果因素和有限元网格法出发,研究了基于神经网络预测毛坯尺寸模型的方法.选取模具参数和工艺参数等作为影响冲压成形结果的因素,用正交表和随机法产生径向基函数神经网络的学习样本;利用自组织神经网络对样本进行分类,用有限元网格法反算的毛坯的长度作为神经网络的输出;设计了神经网络流程,定义了神经网络输出与有限元分析数据的相对误差.通过仿真试验证明,提出的预测毛坯尺寸模型的方法是有效的.

Application of self-organizing feature map neural network for sheet forming

By studying the factors influencing the resuh of sheet forming and with the method of FEM mesh mapping, the method for prediction of blank sheet size based on neural network is studied. Mould parameters and process parameters are chosen as the factors that influence sheet forming. Randomized orthogonal table is used to produce the specimens for radical basis function neural network. Self-organizing feature map is used to classify the specimens. The blank sheet length calculated with FEM mesh mapping method is used as the output of neural network. Also the process of the neural network is designed and the relative error between neural network and FEM is defined. The simulation proves the effectiveness of the presented method.