基于RBF神经网络的中厚板轧机的温降模型

 基于Matlab神经网络工具箱，采用改进的径向基函数(RBF)网络优化计算4200中厚板轧机的轧制温度。通过径向基层散布常数的人工调整以及神经元的自适应调整，提高了收敛速度，确定了最佳的网络结构形式。网络预测结果与在4200中厚板轧机上实测的轧件温度进行了对比，预测精度很高，表明了该网络模型的优越性。该模型可为以温度参数为主要控制对象实行自动化生产的4200轧机提供可靠的参数，也可为人工神经网络在其它自动控制方面的应用提供参考。

Temperature Decrease Model of Medium-Plate Mill Based on RBF Neural Network

On the basis of the Matlab neural network toolbox, the improved RBF algorithm is adopted to optimize the calculation of 4200 heavy and medium plate mill′s rolling temperature. Via the radial substratum dispersion constant modulation as well as the neurotic number self adapting modulation, the rate of convergence increases and the best form of the network is affirmed. Compared with measured temperature of 4200 heavy and medium plate mill rolling, precision of prediction is good, which confirms superiority of the network. This algorithm could count the reliable temperature parameter for 4200 rolling mill automation production control according to the temperature parameter, moreover the ANN network in steel rolling auto control application, which would offer reference for other auto control application.