Composite rectification in modeling of rolling force for hot rolled thick plate

Abstract

Abstract In order to eliminate the predicted bias of the traditional Sims model, a new method, called the composite rectification method, is firstly proposed. Firstly, the deformation resistance model based on industrial big data is built based on the generalized additive principle. This new model is adopted to replace the deformation resistance model in the Sims model. Through this factor replacement, the effect of deformation resistance bias due to the traditional regression method was eliminated, resulting in the once-time corrected Sims model with significantly improved accuracy compared to the original Sims model. On this basis, to solve the mathematical form imperfection caused by the introduction of many assumptions during the derivation of the Sims model, a back propagation (BP) neural network model on the bias of the once-time corrected Sims model is built. Ultimately, the double correction of the Sims model is realized through the additive compensation method, and an integrated model of rolling force is ultimately obtained. By comparing, it is shown that the average error of the traditional Sims model is as high as 34.22%. This error can be minimized to 9.40% with once-time correction and further reduced to 3.06% with the double correction. The results show that the prediction accuracy of the rolling force can be improved gradually by the proposed composite rectification method, and the bias brought by model influence factor and mathematical formcan be eliminated. The composite rectification method presented in this article can provide a new way of modeling complex systems with high precision.