Based On The Complex System Of Neural Network Model Identification Technology And Its Application

With the rapid development of science and technology,the control systems became more and more complex, and the demands for accuracy of the control systems is higher and higher. Because the systems with complex nonlinear can not be described by linear models,it is of great significance to study the identification means of the nonlinear system models. Therewith the ability of upstanding nonlinear mapping, self-learning adaptation, associative memory and parallel information disposal mode as well as the ability of random approach for nonlinear function, neural network offers a fast and effectively means for the identification of nonlinear systems. Moreover, it provides a kind of universal solution to the identification problems of complex models. Thereby, it is of great academic value and immense practical prospects to study the application of neural network in the identification of complicated systems.First of all, this paper summarizes the development of system models identification and main application study techniques, also it presents some new improved methods and control tactics for the applications of neural network in the field of system models identification which are based on the structure characteristics and the analysis of the training calculation way of neural network. Secondly, two typical structures of artificial neural networks-BP network and RBF network have been discussed in detail. And it verifies the upstanding ability of neural networks in identifying of nonlinear system models by simulations. Several kinds of neural network structures and the improved algorithms of parameter self-adapting that this paper puts forward are verified by the actual examples. The effectiveness of algorithm is shown clearly. Finally, by the applied examples of neural network in the fault diagnoses, it is verified further that neural network has good classification and learning ability as well as generalization ability, so it can provide a fast and effectively method for the fault analysis and diagnose of complex systems.All experiments and analysis of this paper are completed under the MATLAB environment platform. By the support of this high-powered environment, the strong simulation analysis and research works in this paper can be done. Certainly, from the point of view of actual applying in the engineering, identifying or error diagnosing, there may still be some problems (e.g. real-time performance, etc). We should consider combining the designed methods with the other stronger software platform for the implementation of system.