基于RBF神经网络的金属应力状态系数模型

The Model of Influential Coefficient in Stressed State of Medium and Heavy Plate Rolling Mill Based on RBF Neural Network

以4200 mm轧机轧制71块钢板的实测数据为基础,利用Matlab人工神经网络工具箱,建立了轧制变形区的应力状态系数的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的试验调整,确定了最佳的网络结构形式,提高了模型的预测精度以及网络的泛化能力.测试结果表明,RBF网络模型具有很好的推广能力.与传统的BP神经网络模型相比较,结果表明,RBF网络具有更高的精度和更好的泛化能力.

Based on the experimental data obtained from 71 steel plates rolled in 4200 rolling mill, this paper established a RBF neural network prediction model of influential coefficient in stressed state by Matlab neural network toolbox. By analyzing influential factors of coefficient of stressed state and taking into account the traditional mathematical model, this paper affirmed the parameters of input layer, and by selecting suitable spread in function-newrb（ ） ,this paper affirmed the best form of the network, and as a result, it improved the prediction accuracy and the adaptability of the network. The results of testing the model indicated that the model based on RBF neural network has a good generalizing capability. Compared with traditional BP network, the result indicated that RBF has better accuracy and adaptability of the network.