Newton Iterative Identification Methods For Input Nonlinear Equation Error Type Systems

Nonlinear systems widely exist in industrial field. With the development of industrialization, the mathematical model of systems becomes more complex and the computation of the identification processing is also growing. Therefore, the identification for nonlinear system has important theoretical and practical value. This thesis presents the identification methods for a class of nonlinear systems based on the National Natural Science Foundation of China. Based on the iterative identification principle, the hierarchical identification principle, the key variable separation principle and Newton identification method, this thesis launch a detailed research, and makes some findings. The main results are described as follows.1. Considering the identification of the input nonlinear systems with a white noise, we derive the Newton iterative identification method. In order to solve the nonlinear problem and the excessive computing problem caused by the presence of the large dimension matrix, a Newton iterative identification algorithm is derived by using the key variables separation principle and the hierarchical decomposition principle. By comparing the computational analysis and simulation analysis of the three algorithms, verifies the validity and the a small amount of computation of the proposed two algorithms.2. For the nonlinear system with a moving average process noise, that is, Hammerstein nonlinear equation error moving average model, a hierarchical decomposition based Newton iterative algorithm and a key variables separation principle based Newton iterative algorithm are derived by referring the research of the nonlinear system with a white noise. The simulations of these two algorithms and the Newton iterative algorithm verify that the key variables separation principle and the hierarchical identification principle can be applied to the input nonlinear system identification.3. According to the key variables separation principle and the hierarchical identification principle, we research the identification algorithm for input nonlinear equation error models. Since the input nonlinear equation error model, the equation error autoregressive model and the equation error moving average model are simplified form the input nonlinear equation error autoregressive moving average model, therefore the study of such model becomes more universal.To summarize, this thesis studies Newton iterative method of input nonlinear systems, the simulation results show that the obtained two methods do not a?ect the parameter estimation accuracy while reduce the computation. Finally, this paper gives the conclusion and outlook, and some of the di?culties faced by this research topic and the subject to be in-depth study. The convergence of these nonlinear system identification algorithms pend further proof.