Research On Identification And Control Algorithm Of Nonlinear Process Based On Hammerstein Model Using B-spline Function

In the actual industrial process,there often exists strong characteristics.Therefore,the problem of model identification and optimal control for strongly nonlinear systems has become one of the hotspots in the field of control theory.Hammerstein model is a representative nonlinear model.The model is feasible for a class of nonlinear processes in which static nonlinearity and dynamic linearity can be separated,and it can deal with nonlinear problems in most chemical processes and has a wide range of applications.Considering identification and control problems of systems with strong nonlinearity,identification and control methods of Hammerstein model based on the B-spline function is adopted in this paper.For the model identification,the constrained Gauss-Newton identification algorithm is adopted.The algorithm not only constructs a auxiliary function as the optimization objective to reduce the computational complexity,but also avoids the uncertainty in the separation of the mixed parameters by applying a constraint.For the model control,two control algorithms are adopted.The adaptive PID control based on Hammerstein model incorporates parameters of the PID controller into the objective function of the nonlinear predictive control,and the optimal control law is obtained through the model predictive value and its derivative derived by the De Boor algorithm,and the predictive updating of PID controller parameters is realized.The static nonlinear part is successfully removed by the control of Hammerstein model using the inverse of De Boor algorithm through inverting the B-spline function,and it's only need to design a pole assignment linear controller for the dynamic linear part.The validity of the identification and control algorithms are verified by numerical simulation examples.Aiming at the problem of nonlinear process control with large-scale and multi-operating conditions,the predictive control algorithm based on multi-Hammerstein models is given.Combining the advantages of adaptive PID control and the control using the inverse of De Boor algorithm,the actual control variable is obtained by the inversion fusion of the pre-identified multiple Hammerstein models.For the model identification,it needs to identifymultiple nonlinear static modules,while it only needs to identify a linear dynamic module,thus reducing the part of calculation.And for the design of linear model predictive controller,the current model predictions can be given by the fusion of multiple Hammerstein models,which can reduce the degree of model mismatch and improve performance of the controller.Several simulation examples in pH neutralization process demonstrate the effectiveness and reliability of the algorithm.In order to illustrate the actual control effect of the predictive control based on multiHammerstein model,the experimental study of the predictive control based on multiHammerstein models compared with the adaptive PID control and the predictive control using the inverse of De Boor algorithm is carried out in the actual water tank level control device.The experimental results show that the predictive control based on multi-Hammerstein models has a good control effect on the nonlinear processes with a large range and uncertainty.