论文部分内容阅读
冷轧平整机的工作辊直接和带钢接触,其表面粗糙度衰减情况对带钢成品的板形和表面质量有重大影响。因此,分析轧辊磨损机制,对轧辊表面粗糙度的衰减进行精确预测十分必要。首先采用灰色关联度分析对影响平整机工作辊表面粗糙度磨损的因素进行分析,确定了工作辊表面粗糙度评估指标体系。进而应用优化在线稀疏最小二乘支持向量回归模型对冷轧平整机的上工作辊表面粗糙度进行在线预测。通过预测误差准则实现系统的前向递推,采用FLOO(fast leave one out)的修剪算法实现其后向删减,并且采用最速下降法实现了2个超参数的在线优化。经过仿真研究表明,系统预测的绝对误差平均值为0.014 9,与其他方法相比具有明显的优越性,并且系统具有在线自适应的能力,能够随着时间而进化。
The work rolls of the cold-rolling mill are in direct contact with the strip and the attenuation of the surface roughness has a significant effect on the plate shape and the surface quality of the finished strip. Therefore, it is necessary to analyze the roll wear mechanism and accurately predict the roll surface roughness attenuation. Firstly, the paper analyzes the factors that affect the wear of the surface roughness of the work roll of the smoothing machine by the gray relational degree analysis, and determines the evaluation index system of the work roll surface roughness. Then the online sparse least squares support vector regression model was optimized to predict the surface roughness of the work roll of the cold rolling mill. The forward recursion of the system is realized by the prediction error criterion, and the clipped algorithm of FLOO (fast leave one out) is used to achieve the backward reduction. The online optimization of two hyperparameters is realized by the steepest descent method. The simulation results show that the average absolute error of the system prediction is 0.014 9, which shows obvious superiority compared with other methods. The system has the ability of online adaptation and can evolve over time.