Research On Flatness Pattern Recognition And Flatness Adaptive Control Of Cold Strip Mill

The flatness of steel strip is one of the key indicators for high-level strip products.At present,China's automobile industry is in a period of rapid development,and high-strength automobile has gradually become the mainstream based on the background of lightweight development for automobile industry,but the poor plasticity and large deformation resistance of high-strength steel put forward higher requirements on flatness accuracy.In this paper,the 1620 mm cold strip mill of high-strength steel is taken as the research object.Aiming at the problems of poor adaptability of flatness control and the unsatisfactory effect of flatness control for high strength steel plate,this paper takes flatness pattern recognition and flatness closed-loop adaptive control as the main research line,and makes a systematic and in-depth study on the position control of inner ring hydraulic servo system and the decoupling control of flatness and thickness.The details are as follows:(1)For uncertain factors such as parameter perturbation and load disturbance,which Influence on tracking performance of hydraulic servo position system of cold strip mill,the improved Extended state observer(ESO)and fuzzy adaptive observer are used to estimate the total uncertainty of the system,a nonsingular terminal sliding mode controller based on the improved ESO and a nonsingular fast terminal sliding mode controller based on the fuzzy adaptive observer are proposed.Simulation results show that the designed controllers can effectively improve the position tracking accuracy and robust stability of the system.(2)For the flatness and thickness coupling controlled object of cold strip mill with nonlinear and multivariable characteristics,firstly,Convolutional neural network(CNN)with strong feature recognition is used to design the flatness and thickness decoupler of cold strip mill;secondly,Model predictive control is used to design the flatness and thickness predictive controller;finally,the effectiveness of the designed decoupling.(3)In order to solve the problems of large number of optimization parameters and complex training process in neural network flatness pattern recognition method,a method of flatness pattern recognition based on particle swarm optimization for kernel extreme learning machine(KELM)is proposed.This method is based on KELM to design the flatness pattern recognition model and uses particle swarm optimization algorithm to optimize the set parameter of KELM.Finally,the validity of the proposed method is verified based on the measured data of the on-site shape meter.(4)Aiming at the problem that the parameters of flatness control system vary with the characteristics of steel grade and the external rolling conditions,based on the recursive least square method with forgetting factor,the influence matrix from the actuator to the actual output of strip flatness is identified online,and then the strip flatness is self-tuning controlled.Based on the simulation model of the strip flatness system established by CNN and flatness adaptive control experiment platform,the proposed control is verified the validity of the method.