Research On Optimization Control Of Strip Thickness And Diagnosis Method Of Rolling Process Health State

Strip thickness is one of the most important indexes for evaluating the quality of hot tandem mill.With the requirement improvement of the quality requirements of rolling products,such automobiles,household appliances,food packaging and construction and so on industry,the requirements of the rolling process on the strip thickness control performance are also raised.Due of the high complexity of the rolling process,it is still difficult to realize the high accurate control of the strip thickness.For the purpose of further improving exit strip thickness control accuracy,this paper carries out research from the two aspects of strip thickness optimization control and rolling process health state diagnosis.The main contents of the paper are as follows:1.Aiming at the problem that strip thickness can not be directly measured and the accuracy of the existing strip thickness prediction model is not high for the thickness setting model system in hot tandem mill,this paper proposes a soft measurement prediction method for strip thickness based on D-S information reconstruction(DSIRPM).Firstly,sensitive rolling parameters are determined according to the ibaAnalyzer analysis software and the graph correlation analysis method;then,sensitive rolling parameters are used to fit the least square polynomial curve to obtain the initial thickness prediction value;finally,grey relational analysis is used to calculate the contribution rate of rolling sensitive parameters to strip thickness,and it is used as the basic probability distribution(BPA)function of D-S evidence theory to fuse the initial strip thickness prediction value to obtain the final prediction results of strip thickness.Compared with single least squares polynomial curve fitting model,GM(1,1)model and the weighted average fusion method,the results show that DSIRPM has better prediction accuracy and stability.2.Aiming at the problem that it is difficult to get satisfactory control effect by traditional control means due to the characteristics of time delay and parameter time-varying of monitor automatic gauge control(AGC)system in hot tandem mill.Starting from two aspects of establishing multi-step prediction model and optimizing controller parameters,this paper presents a kind of optimal control strategy for exit strip thickness based on time series prediction.On the one hand,fractal extrapolation interpolation algorithm is introduced as the prediction model considering time delay,and chaos optimization algorithm is used to determine the vertical scale factor in fractal extrapolation interpolation algorithm,which is developed the improved fractal extrapolation interpolation prediction(IFEIP)method to compensate the system uncertainty caused by time delay;IFEIP method were compared with the existing ARIMA model,BPNN and the developed RARIMA model,ARIMABPNN hybrid model,according to experimental analysis,the results show that IFEIP method is a prediction model for exit strip thickness with excellent performance.On the other hand,in order to further solve the problem that the uncertainty of time delay and time-varying parameter causes the conventional PID controller difficult to guarantee the robust control of exit strip thickness,this paper proposes a delay estimation method based on the integrated wavelet cross-correlation degree to determine the delay step,and IFEIP method is combined with the controller to establish the control model of exit strip thickness based on the time series prediction;the PID controller parameters are optimized by ladder type DMC to solve the parameter time-varying problem,which makes the PID controller parameters adaptive adjustment according to the different condition and ensure exit strip thickness control stability and robustness;the simulation results show that the proposed method can effectively alleviate the control overshoot and oscillation phenomenon of exit strip thickness caused by time delay and has better dynamic and static characteristics under the condition of external disturbance and model parameter mismatch compared with the traditional AGC,Smith-AGC control and IFEIP-AGC control method.3.In order to ensure the stability of long term control performance of strip thickness in hot tandem mill,starting from the health status of rolling process,a method of health state diagnosis for rolling process is proposed based on fuzzy set and deep belief network.Firstly,aiming at"ill" running state between the fault and normal,the concept of health degree is proposed to describe it,and the fuzzy set is used to define the health degree that can describe the health level of rolling process quantitatively to divide the health grade;then,according to the rolling data with "big data" and unbalanced characteristics,deep belief networks with penalty factor is used to calculate the fuzzy membership degree of the monitoring data corresponding to the rolling process health state set in real time,and the health degree of rolling process is obtained by the fuzzy mapping relationship between membership degree and health degree,which realizes the health assessment of rolling process;finally,the health level of rolling process is determined according to the health state division table and the operation and maintenance decision based on FIDBN method is given.The experimental results show that FIDBN method can monitor rolling process health degree in real time,which increases the diagnosis rate of few fault states while ensuring higher overall diagnosis rate of rolling process health state and provides an important basis for maintaining the health quality of rolling process.