Design Of Adaptive Controller With Degree Constraint And Its Application

Adaptive control theory is the important component of the control theory. Parametersestimation and control law design, as two key parts of adaptive control system, are researched.The stability margin, robustness, the capacity of resisting disturbance and noise contaminationare simulated in Matlab. The proposed adaptive control strategy is applied to adsorptionseparation plant of the m-xylene.The project algorithm is introduced into the adaptive control system based on recursiveleast-square algorithm. The projection algorithm can sidepass refreshing the parameters at theprice of slower convergence. It is an engineering trade-off. A new adaptive control strategywith degree constraint is proposed, which depends on the improvement of the Diophantineequation solution. By expanded Euclidean algorithm, the Diophantine equation is solved toachieve the minimum degree poles placement. The stability margin and the robustness arestudied. The results of simulations show that the unmodeled dynamics can make theperformance become worse, even the whole system unstabilized. The capacity of resistingdisturbance and noise contamination of adaptive control system are discussed in the thesis.The disturbance and the white noise are injected into the control system in different places inSimulink. The simulations demonstrate that the parameters are convergence to the true values,which also proves the parameter estimator has good robustness.Combined with the oil refining chemical industry production, the adaptive controlstrategy based on system identification is applied to the adsorption separation plant of them-xylene. The model parameters are identified based on RLS, which is provided for adaptivecontrol strategy. After that, the proposed adaptive control system is compared with thetraditional PID control. The Comparisons show that the former control strategy can increasethe valve response speed, and reduce the output fluctuation, which has the steady, fastresponse.