Research On Intelligent Adaptive Control Of Complex Nonlinear System

The adaptive control of the complex systems with nonlinearities, uncertainties, time delays and chaotic factors is surveyed in this thesis. Several control schemes are presented for corresponding classic engineering systems by combining the adaptive control with the H∞robust control, neural network control, fuzzy control, multiple-model control, and so on, in order to ensure the controlled system is closed-loop stable and satisfies the desired dynamical control performance.At first, a kind of dynamic structure adaptive neural network is presented, which is modified from the fully adaptive neural network and the dynamic structure neural network, so that all of the network's parameters including weights, widths, and centers can on-line regulated, and the number of the hidden units can on-line increase as well. Due to the powerful adaptive property, this kind dynamic structure adaptive neural network avoids the intolerant inherent error caused by unsuitable original experiential arrangement, and approximates the unknown function with optimal parameter allocation and topologic structure to improve the approximation precision.Next, an adaptive control scheme is discussed by introducing the dynamic structure adaptive neural network to the design of the robust adaptive controller for two different complex nonlinear systems. For an uncertain nonlinear interconnected large-scale system, a decentralized H∞robust adaptive controller based on DRBF neural network is presented to make the whole system stable. For a flight control system, an adaptive tracking controller based on DRBF neural network is presented which can have the system stable and tracked the given command fast and accurately by resoling the negative influence produced by the powerful nonlinearities, powerful coupling, and modeling error exited in the system.In the following chapter, the problem is further discussed that most engineering systems usually work under the complex condition which causes the parameters or structure of the system model to abruptly change, and for the case, just one controller will not do well work. Due to outstanding performance of the multiple-model method for controlling the system under complex condition, a kind of adaptive control law based on switching multiple-model method is presented for the flight control system. This multiple-model adaptive controller has the model set formed by one fuzzy model and several fixed linear model, with their corresponding controllers. The optimal present controller is chosen by a proper switching performance index and a switching bound condition. At the same time, the DRBF neural network is introduced into the multiple-model adaptive controller to ensure the system stable, and make it achieve full-envelop flight tracking with satisfied control performance.Finally, a special complex engineering system, the uncertain time-delay nonlinear system, is studied as controlled plant for the adaptive control in-depth survey. Especially, for a class of chaotic systems with uncertain multiple time delays, a time-delay compensator is designed by the Lyapunov-Krasovskii function, and then it is augmented by the DRBF neural network to constructe an adaptive tracking controller, which deals with the difficulties produced by chaotic factors, uncertainties, nonlinearities, and time delays, so that make the system achieve desired static and dynamical performance. The successful application on the uncertain time-delay systems proves the presented adaptive tracking controller is effective, and also proves the neural-network-based adaptive control method surveyed in this work is suitable for complex systems once more, which can realize different control intention by modification and complementarity for the specific controlled system.