运用动态BP网络和数学模型预测热轧变形抗力

Flow Stress Prediction With Dynamic BP Networks and Mathematical Models in Hot Rolling

为了提高变形抗力预测精度 ,针对传统热轧变形抗力模型的固有缺陷 ,提出一种将动态 BP神经网络与数学模型相结合的新方法。大量实验数据表明 ,采用这种方法 ,预测结果的相对误差在± 6%以内 。

In view of intrinsic imperfection of traditional models of flow stress of steels with hot rolling, a new method combining dynamic BP networks with mathematical models is proposed to improve precision of flow stress prediction. With the certification of a large amount of experimental data, it is indicated that the relative error is within and the predicted values get by this method is more accurate than that by traditional models.