基于神经网络和自适应预报模型参数的平整轧制力模型

冷轧带钢平整过程存在带材较薄,压下率较小等特点,应力状况较为复杂,采用传统模型对轧制力预报误差较大。采用神经网络模型对应力状态系数进行预报,可提高轧制力的预报精度。针对轧制过程中轧件特性发生缓慢变化的特点,采用符合平整轧制过程特点的变形抗力自适应模型,并与神经网络模型结合,预报平整轧制力。计算结果表明,该模型计算值与实际值吻合精度较高,90%的预报结果相对误差控制在5%以内。

Temper rolling force model based on predicting parameters by neural network and self adaption

The character of temper rolling process is thin strip and little reduction,so the stress state is complex,the error is heavy if predict the rolling force by traditional model.Predict the stress state coefficient by neural network model,it can improve the precision of rolling force.As to the slow variety of the mill characteristics during the rolling process,adopts self adaption deformation resistance model which accord with the character of temper rolling process,and combined with the neural network model to predict the temper rolling force.The result indicate that the rolling force predicted and measured agree well,the relative error of 90% predicting result within 5%.