摘要
应用Matlab中内建的神经网络工具箱建立了三层前向反馈BP神经网络模型,并利用拉深试验中采集数据样本集对模型进行前期学习训练,达到指定目标误差后再利用另外一些实际试验样本集来验证所建模型。结果表明：此三层BP模型模拟计算结果与试验结果的相对误差在1%之内,可有效地预测薄板拉深成形过程中的成形性能参数设置是否合理,从而为实现薄板拉深成形过程的智能化预测奠定一个基础。
A neural network model with three layers of BP algorithm was built with the aid of neural network tool box in software Matlab. After trained by a series group of data collections sampled from testing, this model could finally achieve the specified goal error and then was validated by another group of data collections from testing. The relative error between simulation results and test results was limited within 1%. So this neural network model could be effective in judging whether the rationality of set forming characteristic parameters was good or not in advance of actual testing, and also established a good preconditional basis for the following intelligentization in sheet deep drawing.
出处
《机械工程材料》
CAS
CSCD
北大核心
2006年第4期20-22,82,共4页
Materials For Mechanical Engineering
关键词
神经网络
薄板
拉深成形
智能化
neural network
sheet
deep drawing
intelligentization
