Reaserch On The Advanced Control Of Continuous Stirred Tank Reactor

Chemical industry,Petro industry and rubber industry are all the main industry of the national economic development.CSTR is the major equipment of chemical reaction in petrochemical industry,so the study on its automatic control is significant.There are many typical characteristics of this method in the complicated manufacturing process,such as nonlinearity and time varying,which bring great difficulty about quality in the manufacturing process.Based on the problems.This paper focused on the parameter of temperature and the concentration of the reactant in this process.The traditional PID control algorithm adjust proportional,integral and differential,it can obtain satisfactory control effect,but it is difficult for the nonlinear continuous stirred tank reactor(CSTR)system to solve the system stability,and the temperature control precision is difficult to guarantee.Model Predictive Control is not need to understand the internal mechanism of the process,which apply to constraint conditions,the big lag,nonlinear process.And the rolling optimization strategy has a good dynamic control effect.Based on the comparison of the process,Soft sensing the secondary variables by computer modeling to obtain the direct measurement of variables.This paper mainly briefly introduces the related concepts about the production flow in CSTR.It also introduces how to select the control index in the production process of the CSTR,and analyzes the dynamic behavior of the reaction kettle temperature,also affects the quality of products.This paper establish the heat exchange system by setting the optimal PID parameters of temperature controller for the complexity of CSTR.It also introduces the soft measurement technology and Sugeno fuzzy neural network model,which establishes the soft sensor model based on Sugeno fuzzy neural network to predict the concentration and temperature in the reaction kettle through material flow velocity.Finally,this paper introduce the technology of model predictive control.The DMC,MAC,GPC three kinds of prediction model is established respectively,and the simulation of MAC model has stronger robustness and antiinterference compared with other two kinds of temperature prediction model.The simulation results show that the optimized simulation convergence speed is fast,the approximation precision is higher,has certain actual application value.