基于改进BP神经网络铆接接头力学性能预测

Prediction of Mechanical Properties of Clinching Joints Based on Improved BP Neural Network

针对钢铝材料的铆接接头优化设计和结构车身混合材料接头技术问题,使用改进的BP神经网络模型研究钣金材料厚度、硬度、接头底部直径等的接头技术参数与材料自身剪切力与剥离力强度等力学参数的映射关系。ALM算法被用来优化改进的BP神经网络预测模型连接权值,提高了神经网络模型的预测精度和泛化能力。对改进的神经网络的预测结果进行检验的结果表明,训练后的神经网络模型能够准确有效地预测铆接接头力学性能,证实了改进神经网络应用于铆接接头力学性能预测的可行性与可靠性。

The mapping relation between sheet thickness, hardness, joint bottom diameter and the material itself shearing strength and peeling strength and other＇ mechanical parameters were studied using the improved BP neural network aiming to the optimizing design of steel aluminum materials clinching joints and the technical problem of structure mixed materials joints. ALM algorithms are used to optimize the improved BP neural network prediction connection weights of the neural network model to improve the prediction accuracy and generalization ability, which improved prediction accuracy and generalization ability of neural network model. The testing verification to prediction results of improved neural network indicates that the trained neural network model can accurately and efficiently predict the mechanical properties of clinching joints. It confLrrned the feasibility and reliability of improved neural network used in mechanical properties predicting of clinching joints.