基于深度学习的热轧带钢头部厚度的命中预测

 针对热轧带钢头部厚度精度较低的问题,提出了一种基于深度学习的热轧带钢头部厚度的命中预测方法。在精轧过程中,带钢头部张力较小,且通常温度较低;同时轧机工艺参数复杂,精准设定存在困难,轧制带钢头部经常会出现厚度不合格的现象。利用深度神经网络的非线性拟合能力,设计带钢头部厚度预测模型,给轧机的参数设定提供参考、提高头部厚度命中率、减少钢材浪费。深度神经网络(DNN)包含输入层、隐藏层、输出层,使用TensorFlow开源机器学习框架设计预测模型并用程序实现。调整神经网络各参数,通过研究它们对模型性能的影响,优化预测模型。最后使用多种厚度的带钢测试数据训练并检验头部厚度预测模型,结果显示,分类预测命中准确率在80%以上。

Accurate prediction of head thickness of hot-rolled strip based on deep learning

Aiming at the problem of low precision of hot rolled strip head thickness, a hit prediction method of hot rolled strip head thickness based on deep learning was proposed. In the process of finish rolling, the tension of the head end of the steel is small, and the temperature is usually lower. At the same time, the process parameters of the rolling mill are complex, and it is difficult to set accurately. The thickness of the rolled strip head is often unqualified. This study intends to use the nonlinear fitting ability of the deep neural network to design the prediction model of strip head thickness, to provide a reference for the parameter setting of rolling mill, improve the hit rate of head thickness and reduce the waste of steel. The deep neural network (DNN) consists of the input layer, hidden layer, and output layer structure. The prediction model is designed by TensorFlow open-source machine learning framework and implemented by program. By adjusting the parameters of the neural network and studying their effects on the model performance, the prediction model is optimized. Finally, the model of head thickness prediction was trained and tested with the test data of various thicknesses of strip steel, and the result showed that the accuracy of classification prediction was more than 80%.