粗轧过程轧制力BP神经网络预报

PREDICTION OF ROUGH ROLLING FORCE BY BP NEURAL NETWORKS

利用BP神经网络 ,以某热轧厂粗轧机组数据库中的数据为训练样本 ,采用两种训练方案 ,对粗轧过程轧制力进行预测。BP网络的预报精度 ,既与训练样本的选取有关 ,又与隐层节点的个数以及相对化系数的大小有着密切的联系。以上因素选取得当 ,能够提高网络的预报精度 ,若选取不当 。

Based on the measured data from a hot strip mill, the rolling force in roughing was predicted by means of BP neural networks which adopted two kinds of training methods. It was proved that the prediction accuracy of BP neural networks was not only relating to the choosing of the training patterns but also the number of the hidden layer and the relativation coefficient. The neural networks prediction accuracy could be improved if the above parameters were set advisably, or else the accuracy would decrease.