System Identification of Nonlinear Dynamic Structures Based on Stochastic Time Series Model Fitting

Summary

A new technique for dynamic system identification by use of a stochastic time series models such as the AR model or the ARMA model, is proposed and investigated. After showing the advantage of the methods over former major identification methods based on FFT, this paper focuses on a double variate AR model method which can be developed to identify nonlinear dynamic systems. Here the dynamic characteristics are evaluated in the form of equivalently linearized system parameters. As a typical example, a simple nonlinear system having a bi-linear stiffness characteristic is adopted. Through numerical simulation for the model, the validity and applicability of the method to such a nonlinear system is examined.