融合工业大数据的热轧厚板轧制力模型研究

目的针对传统解法建立的轧制力模型精度不足的问题,建立一个轧制力整合模型。方法对工业大数据进行归一化处理,系统优化了神经网络模型的结构形式,建立了一个神经网络模型。在此基础之上,利用误差间距补偿的方法实现神经网络模型与已有理论模型的有机融合,从而最终获得了轧制力的整合模型。结果通过与已有的轧制力模型进行对比,表明所提出整合模型预测结果与实测值吻合更好,其中轧制力误差为-4.09%,轧制力矩误差为-4.01%。结论该模型整合方法能够实现理论模型与神经网络模型的优势互补,从而给出物理概念与预测精度均可靠的计算结果。

Investigation on the Model of Rolling Force by Integrating Industrial Big Data

The paper aims to establish an integrated model of rolling force to solve the problem of low precision of rolling force model established in the traditional analytical method. In the paper, all the industrial big data were normalized; and the structure of the neural network model was optimized and a neural network model was established. On this basis, the neural network model was organically integrated with the existed theoretical model through the compensation of error space, and thus obtaining integrated model of rolling force. The comparison with the existed rolling form model showed that better agreement was found between the predicted results of the integrated model and the measured values. In which, the rolling force error was -4.09%, and the rolling torque error was -4.01%. The method of integrating the two models can achieve mutual complementation of theoretical model and neural network model, to provide the calculated results with definite concept and high prediction simultaneously.