Research On Flatness Control System For Cold Rolling Process

Flatness is an important quality index of cold rolling strip steel.It directly affects the yield of steel products and subsequent processing.Flatness control technology is the key technology of modern high precision cold rolling mill,but because of the complexity of the rolling process,flatness quality is still the most common and most difficult problem to solve in actual production of cold rolling.Moreover,it has not yet reached the stage of stable and mature.Based on artificial intelligence theory,this thesis does the theory and simulation research of flatness control system in cold rolling process from three aspects of pattern recognition,flatness prediction model and control method,which has achieved some results.Aiming at the shortcomings of the traditional flatness pattern recognition methods,this thesis adopts the Elman neural network to solve the problem of flatness pattern recognition.Because Elman neural network has slow convergence speed,higher dependence on initial weights and easily trapped in local optimal value,this thesis uses the differential evolution(DE)algorithm to optimize the initial weights and threshold of Elman neural network.Furthermore,the differential evolution algorithm is improved by using the variation factor and enables it change adaptively in the process of optimization.This method enhances the ability of searching the global optimal solution.After the optimization by differential evolution algorithm,this thesis uses the momentum BP algorithm to search locally and trains the optimal Elman neural network.In the end,with Legendre orthogonal polynomial as the base model,this thesis establishes DE-Elman network flatness pattern recognition model based on the Euclidean distance.The model adopts the three input parameters and three output parameters.In this model,the internal layers' physical meaning of network is clear.It has high computing speed,high identification accuracy and can also meet the requirements of high precision flatness control in cold rolling process.Flatness control system has the characteristics of nonlinear,time-varying and serious interference.By the analysis of the deficiency of traditional mechanism modeling,this thesis establishes the DE-Elman network flatness forecasting model.Based on the analysis of the influence factors of flatness and field production experience,this model adopts the ten rolling parameters and three flatness adjustment parameters together as input parameters,three flatness characteristic parameters as output parameters.The data is collected from the third colding rolling factory of Benxi steel and after the data filtering and pretreatments,this thesis selects the representative training samples for network training.Simulation results show that the flatness forecasting model which based on DE-Elman network has higher prediction accuracy.In comparison with the traditional BP network,the forecast error is smaller.So it proved the practicability of the model.Finally,because the DE-Elman network flatness forecasting model is constructed,according to the theory of effectiveness function,this thesis gives the calculation method of effectiveness matrix.On the basis of the results of simulation research and production experience,this thesis determines three key factors affecting the flatness and establishes the effectiveness matrix table.After that,this thesis also adopts three-dimension centroid interpolation algorithm to calculate actual working condition point of effectiveness matrix on-line.This algorithm is simple and has fast computing speed.It can also satisfy the requirements of the flatness online control.Simulation results show that flatness closed-loop control method which based on the key influencing factors effectiveness matrix table can give full play to the regulation of the flatness control ability.This method has high control precision,stable control process and faster convergence speed.It provides a new control scheme for the flatness control system and greatly improves the capability of flatness control online.So,it has great significance to improve the quality of cold rolling steel products.