中厚板轧机宽展的神经网络预测

中厚板轧制过程是一个复杂的非线性变形过程，传统的宽展数学模型很难准确表达轧制宽展特性，不能保证计算精度。利用Matlab人工神经网络工具箱，采用改进的BP网络Levenberg Marquardt训练规则优化计算中厚板轧机的宽展，通过对该网络隐含层神经元个数的调整，使收敛速度加快，提高了宽展的预测精度。该方法使中厚板轧机自动控制精度的进一步提高成为可能．

Width Variable Prediction in Medium Plate Mill by Means of Neural Network

The medium plate rolling process is complicated and non-linear one. The traditional width variable model is hard to express the rolling characteristics exactly and pledge precision. This paper presented optimum calculation method of the width variation in medium plate mill, based on the improved BP network Levenberg-Marquardt training rules. This method increases the convergence speed, and improves the prediction accuracy, increases the possibility of high accuracy in auto control.