基于RBF的轴承钢变形抗力的预测

Prediction of Bearing Steel Deformation Resistance Based on RBF Neural Network

以凸轮式高速形变试验机得到的试验数据为基础,利用Matlab人工神经网络工具箱,建立了轴承钢的变形抗力与其化学成分、变形温度、变形速率及变形程度对应关系的RBF神经网络预测模型。分析了变形温度和变形速率对轧制压力网络模型精度的影响。得出随着变形温度的增加,网络的预测误差逐渐增大;随着变形速率的增大,网络的预测误差逐渐减小的结论。通过与BP网络和Elman网络模型相比较,结果表明,RBF网络模型具有更高的精度和较强的泛化能力。

On the basis of the data obtained from Cam Plastometer, a RBF neural network prediction model was established for the relationship between bearing steel stress and chemistry elements, temperature, strain rate and deformation strain of carbon steel by Matlab neural network toolbox. The precision of network model which was affected by the temperature and the strain rate were analyzed. The precision of network model falls with the in- creasing of the temperature, and the bigger the strain rate is, the higher the precision of the network model. Through compared with the BP network and Elman network, the results indicate that RBF has better accuracy and adantabilitv.