RBF神经网络在金属应力状态系数模型中的应用研究  

Application Research on the RBF Neural Network in the Model of Influence Coefficient in Stress State of Metal

作　　者：孟令启[1] 雷明杰[1] 王建勋[1] 吴浩亮[1] 

机构地区：[1]郑州大学机械工程学院,河南郑州450001

出　　处：《矿山机械》2008年第24期25-28,共4页Mining & Processing Equipment

基　　金：国家自然科学基金资助项目(50175031)

摘　　要：以结构钢板的大量实测数据为基础,利用 MatLab 人工神经网络工具箱,建立了结构钢板变形区的应力状态系数与锻前、锻后厚度的 RBF 神经网络预测模型。通过分析应力状态系数的影响因素,结合传统的数学模型,确立了网络的输入层参数,并对函数newrb中宽度系数 spread 的实验调整,确定了最佳的网络结构形式,提高了模型的预测精度,并与 BP 和 Elman 神经网络模型相比较,结果表明,RBF 神经网络能更好地适用于金属应力状态系数模型。On the basis of the data obtained from large scale experiments in structural slab, the RBF neural network prediction model is established for the relationship among the influence cofficient in stress state, the thickness before forging and thickness after jorging by Matlab neural network toolbox. By analyzing influence.factors of coefficient in stress state and taking into account the traditional mathematical model, the parameters of input layer is established. By selecting suitable width coefficient spread in Junction of newrb0, the best form of the network is obtained. So the prediction accuracy is improved. Comparing with BP and Elman network, the result indicates that RBF neural network can be better applied to the model of influence coefficient in stress state of metal.

关 键 词：应力状态影响系数 RBF神经网络 预测 

分 类 号：TG113.2[金属学及工艺—物理冶金]