Model in Frequency-Domain Identification of a Fast Steering Mirror System Based on Levenberg-Marquardt Algorithm

Abstract

A method for frequency-domain identification of a fast steering mirror system is provided. The method firstly uses swept sine as dynamic excitation signal to obtain the frequency characteristics of controlled object and frequency response data by dynamic signal analyzer Agilent 35670A. Then the paper uses data pre-processed to analyze magnitude-phase characteristics based on FFT (Fast Fourier Fransform). The error of FFT calculation is relative major. To obtain the frequency characteristics approaching to real system, windowing and Sample Hold is introduced to modify magnitude-phase characteristics. For non-parametric frequency characteristics curve of the system, in combination of Levenberg-Marquardt nonlinear least squares curve fitting and transfer function model based on dynamic mechanism modelling, this paper implements curve fitting for low frequency band with frequency characteristics. With this method, the frequency characteristics of complex system can be converted to accurate transfer function. Levenberg-Marquardt algorithm providing the theoretical support for parameterizing BODE diagram is an optimization strategy. The optimization strategy improves the matching degree between actual system and model system in process of modeling and it is provided for the compensation of the influence which is from the temperature shift and measurement error of components. The paper provides a theoretical foundation for building a more complex electro-optical tracking system.