| In modern industrial processes, the man made systems, for high quality and large quantity products, usually exhibit nonlinearities, time-varying, unpredictable parameters and other complexities, accordingly solely using linear system methodologies often cannot provide a holistic description of the dynamic characteristics of the systems. Neural network based inductive techniques, subject to their inherent approximation capabilities, have been found to be very supportive for modeling and controlling such class of complex dynamic systems. Delay is also a common phenomenon encountered in the actual industrial production process, which makes the control system design even challenging. The aim of the research conducted in the PhD program is to take in neural network prototype to deal with time delay related problems in the control domains, particularly to study the stability and synchronization, and nonlinear system identification and control issues embedded in delayed cellular neural networks, therefore provide efficient solutions. A brief introduction of the major research achievement is outlined below.Firstly, it studies the stability of cellular neural network with time delay, and obtains global asymptotic stability and exponential stability based on the Lyapunov stability theory. Stability criterion is divided into two kinds of delay independent and delay dependent, in the case of small delays, delay dependent criterion usually more meaningful. Stability criteria given in the form of LMI can be implemented by software. It studies synchronization problem with chaotic delayed cellular neural networks. In the case of disturbances associated chaotic neural networks, the synchronization controller is designed based on the sliding mode control strategy; in the case of the chaotic systems with impulsive, the impulsive synchronization controller is designed for the unified chaotic neural network.Secondly, it studies the identification and control by neural network for multiple input multiple output (MIMO) nonlinear systems. For general nonlinear systems, based on Lyapunov stability theory, the indirect adaptive controller is designed by sliding mode control method. Using neural networks for nonlinear system identification, neural network weight coefficient is adjust by adaptive algorithm, then using the model identified, the controller is designed to control the system state to track the pre-set trajectory. In the further research, the dynamic compensation method is used to design the indirect adaptive neural network controller, according to the model which identified by neural network and add the proper dynamic compensation, and then the controller is designed to guarantee the system state to track those predefined trajectories.Thirdly, it studies triangular structure of the single input single output(SISO) nonlinear systems. Neural network direct adaptive controller is designed based on backstepping strategy, and the error closed-loop system between output and desired trajectory to be uniformly bounded is proved by Lyapunov stability theory. In the further research, indirect adaptive neural network controller is designed based on backstepping strategy. Based on the neural network model of the system, apply proper dynamic compensation, and then the error closed-loop system between system output and desired trajectory can be proved uniformly bounded stable in term of Lyapunov stability.Finally the temperature control of pipeline welding process, a typical nonlinear dynamic system, is selected as a bench test study. First, based on the pipeline welding process, the static mathematical model is established in order to analyze the variation of temperature welding process. It can be observed from the variation that the accurate dynamic model cannot be established for welding process, and because the temperature control is a subject of some uncertainty in electricity, voltage and other factors. It has been gained from previous research that the form of nonlinear systems with characteristics of triangular structure, therefore the indirect adaptive control method with the backstepping strategy is selected. In the experiment, the major work includes building pipeline welding temperature control platform, implementing the communication between Matlab and Kingview through OPC technology, running the operational demonstration of the pipeline welding temperature control system. |
