基于神经网络的轧制力模型参数辨识

为了提高热连轧轧制力预设定值的精度，提出卫种新的轧制力模型参数辨识方法。利用人工神经网络对以往的大量生产数据进行了训练、预测、将预测结果结合轧制力模型，对思制力模型中的温度相关关系数ｍ１、变形速度相关系数ｍ３进行只。现场生产实践表明，采用辨识后的模型进行轧制力预设定，带钢头部厚度精度有明显提高。对于象本钢热连轧厂这样的老企业，这种新方法更具有在线应用的可行性。

IDENTIFICATION OF PARAMETERS OF NERVE NETWORK BASED ROLLING MODEL

In order to raise the accuracy of the preset rolling force of the continuous hot rolling mill a new method identifying the parameters from the rolling force model has been put for ward,in which an artificial nerve network is utilized to train and predict a great volume of the past production data and then use the predicted outcome to identify the temperature related coefficiency m 1 and the deformation speed related coefficiency m 3 in combination with the rolling model. production practice demonstrates that the gauge accuracy at the strip head can be drastically raised as long as the identified model is used to preset the rolling force. A greater feasibility of on-line application of the new method exists for the old enterprises just like the Continuous Hot Rolling Mill of Benxi Iron & Steel Corp.