RBF神经网络在后桥壳体焊接中的应用研究

在分析焊接残余应力影响因素的基础上,将RBF人工神经网络和有限元分析进行了有机集成,并用以解决后桥壳体焊接工艺优化的问题。首先用ANSYS对矩形对接板的焊接进行模拟,从分析结果中获取训练样本,对RBF网络进行学习,获得能够进行焊接结果判断的人工神经网络系统,最后用ANSYS对后桥壳体焊接过程进行数值模拟,模拟结果证明了系统的可行性。

Research on Application of RBF Neural Network in Rear Axle Housing Welding

On the basis of analysis of the different parameters on welding residual stress, RBF neural network was integrated with Finite Element Analysis, in order to solve the problem on Optimization of the welding technology parameters. At first, the welding at Rectangular docking board is numerically simulated with ANSYS, and then RBF neural network was trained by data samples that came from Finite Element Analysis. The ANN system that can estimate the result of welding could be founded. At last, the welding at rear axle housing is numerically simulated with ANSYS; FEM analysis validates the mapping results of the ANN system.