Research On Weighted Multiple Model Adaptive Control For Parameter Jump Systems

With the rapid development of science and technology,the industrial production process and the space exploration present diversity and complexity,which bring more and more serious challenges for the control system of the equipments.In the actual control process,when the sudden failures of the system or changes in working condition occur,the conventional adaptive control system will produce large transient errors.Fortunately,weighted multiple model adaptive control strategy is an effective solution for these cases.The model set covers and approximates the real plant,and the controller set is constructed corresponding to the local models.Based on the local controllers,the global control signals are generated by a weighting algorithm,which can effectively deal with the system response jitter problem caused by switching multiple model control and effectively control the complex and variable system in real time.In this dissertation,a weighted multiple model adaptive control is proposed for parameter jump systems,which focuses on theoretical analysis,simulation verification and method application,as follows:(1)A new weighted multiple model adaptive control framework is constructed for parameter jump systems.A model set including one adaptive model and multiple fixed models is constructed to cover and approximate the real plant,the controller set is designed correspondingly.The weighting algorithm based on the model output error performance index and its convergence proof are given.Simulation experiments verify the effectiveness of the control method.(2)Based on the weighted multiple model adaptive control framework,the virtual equivalent system theory is introduced,and the overall closed loop stability and convergence analysis independenting of the specific local control methods are given.The system stability and convergence are of the general significance.(3)A weighted multiple neural network controller is designed for complex nonlinear systems with jumping parameters.Based on the proportional-derivative-like and positively related neural network learning algorithm,a weighted multiple neural network control strategy including several fixed and a reassignable neural network controllers,is constructed.Meanwhile,the proof of the global system stability is given.Simulation experiments verify the control performance.(4)The study achievements are applied into the control system of space robots.The motion control problems for flexible structure of space robots are analyzed,and the theorietical applicability of the proposed method is explored.Through the combination of multiple model control strategy,dynamic surface control,boundary control and neural network control,the weighted multiple model adaptive controllers are designed respectively to realize the motion control for flexible joints and flexible manipulators with jumping parameters.Simulations verify the effectiveness of the control system.