Research And Experiment Of System Identification Based On Real-Gained Data

As one of the main branches of the automatic control science, system identification has been applied in many fields. In the past, system identification was mostly applied in linear system modeling, and the flawless theory for linear system had come into being for many years. However, with the development of society and science, nonlinear system is more and more important. The conflict between control and model is getting more and more evident. This fact results in the development of nonlinear system identification theory.This paper studies a data experiment and identification problem of an actual system, in which the steering gear and the satellite-satellite pointing/tracking system act as the study object, based on system identification technique. The main factors that influence identification results and problems that should be paid attention to are analyzed. Base on the analysis, Auto-Regressive Moving Average with Exogenous Input Model (ARMAX) for steering gear and a three-layer predictive control neural network model are established.Aiming at the steering gear, a data experiment based on dSPACE is designed. With pseudorandom binary sequence as inspiriting signal, feedback data is collected and time-domain and frequency-domain characteristic of the steering gear is analyzed. According to the results, system delay is determined. Applying predictive error identification method, by comparing the different order model, model structure and parameters of the steering gear is determined.Aiming at satellite-satellite pointing/tracking system, an experiment based on xPC real-time simulation platform is designed. White noise acting as inspirit signal, the experiment data is collected. Utilizing these data and error back propagating identification method, different neuron and input-output delay are selected. By comparing approximation ability and generalization ability, the neural networks model in position mode and velocity mode is identified.