基于KPLS与SVM的热连轧板凸度预测  

Strip crown prediction of hot-rolled strip based on KPLS integrate SVM

作　　者：姬亚锋[1] 宋乐宝 原浩 刘光明[2] JI Ya-feng;SONG Le-bao;YUAN Hao;LIU Guang-ming(School of Mechanical Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,Shanxi,China;School of Materials Science and Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,Shanxi,China)

机构地区：[1]太原科技大学机械工程学院,山西太原030024 [2]太原科技大学材料科学与工程学院,山西太原030024

出　　处：《中国冶金》2021年第1期20-24,30,共6页China Metallurgy

基　　金：国家自然科学基金资助项目(52005358);山西省面上自然基金资助项目(201901D111243)。

摘　　要：热连轧作为典型的流程工业过程,具有多变量、强耦合、过程非线性的特点,轧制机理非常复杂。针对传统方法难以获得准确的数学模型从而导致板形质量预测精度较低的问题,采用基于数据驱动的核偏最小二乘(KPLS)方法以有效处理工艺参数和质量指标之间的非线性关系,以此为基础,建立了基于KPLS结合支持向量机(SVM)的板凸度预测模型,并采用粒子群优化算法(PSO)优化支持向量机参数,进一步提高热连轧板凸度预测精度。预测结果表明,96.86%的板凸度预测值绝对误差小于5.5μm,整体具有较高的预测精度,对实现板形质量精确控制、提高热轧产品质量具有重要意义。As a typical process industry, the hot strip rolling process is multivariate, strong coupling and nonlinear, and the rolling mechanism is very complex. Aiming at the issues that it is difficult to obtain an accurate mathematical model with traditional methods, which leads to unsatisfactory qualification of strip crown, the data-driven kernel partial least squares(KPLS) method is used to effectively handle the non-linear relationship between process parameters and quality indicators. Based on this, based on KPLS method, a predictive model for strip crown is proposed combined with support vector machine(SVM), and the particle swarm optimization(PSO) algorithm is employed to optimize these parameters of the SVM to further enhance the forecast accuracy of the strip crown. The prediction results indicate that the absolute error of the predicted value of 96.86% of the strip crown is less than 5.5 μm, and the overall prediction accuracy is high, which is of great significance for achieving precise control of the shape quality and improving the quality of hot-rolled production.

关 键 词：热连轧 支持向量机 粒子群优化算法 核偏最小二乘算法 板凸度 

分 类 号：TG335.5[金属学及工艺—金属压力加工]