Research On Multivariable System Identification And Inverted Decoupling Control Method In Industrial Process

System identification is an effective method to obtain industrial process model.It has been one of the most active research fields in industrial control.The internal mechanism of the process is usually quite complex,for example,the distillation column in industry is difficult to get the corresponding mathematical model by using the existing theory.In order to obtain the control model,a simplified method is to test the characteristics directly in the field,and the dynamic characteristics of them can be studied by the identification method and simulation experiment first.The purpose of identification is to control,so we can further discuss advanced control algorithm based on the process model.In view of the characteristics of multi-coupling,time delay,large inertia and uncertainty in multivariable systems,traditional PID control cannot meet the overall performance and process requirements of the system,new advanced control strategy is urgently needed to achieve the optimal control target of the system.In this paper,the process model and control strategy are derivative and the corresponding simulation test and experimental research are to verify the feasibility,simplicity,efficiency of the proposed method,the main content of the thesis includes the following aspects:(1)A model reduction method for open loop multivariable system identification.First the identification effect of ARX model under different noise levels is studied,the problem of model order selection under different noise levels is proposed through AIC and output error method(OE),it can identified by high order ARX model which has good model adaptation in this criterion,by combining the internal mechanism of the actual process,the approximate model of the process can acquire after the high model is reduced.The effectiveness of the proposed method is verified by the simulation examples and the experiments of the pressure and flow of the pipeline(2)A reduced order closed loop identification method based on AIC.First,the time delay term is approximated by first-order Pade and the equivalent closed-loop transfer function of closed loop system is calculated.Second,the high order discrete model which is determined by proposed method based on AIC can be obtained by the identification of the real closed loop system.The terminal identification model which is gained from after reduction of the high order discrete model has the same model structures as the equivalent closed-loop transfer function.Final,solving methods of process parameters could be deduced by coefficient of same power in the two typical time delay process.The effectiveness of the proposed method is verified by the simulation examples and the experiments of the distillation column.(3)A multivariable inverted decoupling control method suitable for high dimensional and high order process.First,the optimal input and output pairing of the system is gained by relative gain matrix.So,the structure design of the direct control channel and feedback control channel of the inverted decoupling controller is proposed,and then the general expression of the high dimensional inverted decoupling controller is obtained.Finally,the feasibility and the robust stability performance of the decoupling controller are analyzed.The effectiveness of the proposed method is verified by the simulation examples and the experiments of the distillation column.(4)A decentralized PID control method based on reverted decoupling.First,an independent design of a decentralized PID controller for TITO system by reverted decoupling diagonal matrix is proposed for process with time delay or RHP zeros.Second,a new analytic methodology of inverted decoupling internal model control base on internal model control structure is proposed,the controller are stable,causal and proper for process,the proposed design method is aimed at achieving the desired principal diagonal matrix function after decoupling,then,a decentralized PID controller is designed by using the desired closed-loop transfer function matrix.The effectiveness of the proposed method is verified by the simulation examples and the experiments of the pressure and flow of the pipeline.