基于时间序列和支持向量机耦合分析的冷轧辊系热变形预报

Thermal deformation prediction of cold rolls based on time series and support vector machine coupling analysis

板带轧机辊系热变形预报对板厚和板形控制具有重要的意义。在分析轧辊热变形机理的基础上,以300实验轧机为研究对象,基于时间序列分析方法,分离提取出轧辊趋势性热变形与周期性热变形。取总轧制力、轧制速度作为周期性热变形的影响因子,提出采用时间序列分析和支持向量机回归相结合的方法对其进行预测;最终将趋势性热变形与周期性热变形预测值叠加得到累计热变形预测值,提高了对热变形的周期性变化的预报能力和预报精度,具有良好的应用前景。

It is a significance issue to predicate the thermal deformation of roll system in strip mill in order to the control of plate thickness and shape.The first,the trend thermal deformation and periodic thermal deformation of 300 experimental rolling mills areseparated and extracted based on the analysis of the mechanism of roll thermal deformation and time series analysis method.The second,the total rolling force and rolling speed aretaken as the influence factors of periodic thermal deformation,and the support vector regression method isused to predict the thermal deformation.Finally,the cumulative thermal deformation prediction value isobtained by superposing the predicted values of trend thermal deformation and periodic thermal deformation.The new prediction ability and accuracy of the periodic variation of thermal deformation are improved,which has a good application prospect.