基于支持向量机灵敏度的轧制过程厚度控制

Thickness control for rolling process by sensitivity based on support vector machine

利用支持向量机(SVM)建立非线性轧制力模型,由模型分别对各个输入变量进行偏微分解决轧制过程模型"代数环"问题,获得轧制过程出口厚度灵敏度系数,并建立基于灵敏度的轧制过程增量线性模型.由所建立基于灵敏度的轧制过程线性模型作为轧制过程内模估计实现内模控制,用于预测轧机出口厚度的变化,消除由传感器检测厚度所产生的纯滞后的不利影响,缩短过渡过程.仿真结果表明,提出的控制方法比PID控制具有更快响应速度和更高控制精度.

A linearization method for the rolling process nonlinear model and the calculation for the sensitivity factors of a linearized model are obtained by the differential of the rolling force model based on support vector machine. It has an advantage without an ＂algebraic loop＂ in the rolling process. A controller is also proposed with the characteristic of the exit thickness predicted by the rolling process linearized model based on sensitivity factor as an internal model in place of the virtual exit thickness measured by the sensor with the disadvantages of a time delay. The comparisons of simulation results show that the proposed controller has a better control performance and higher precision than PID controller.