Identification of nonlinear structures using discrete-time Volterra series

Abstract

Mathematical modeling of mechanical structures is an important research area in structural dynamics. The goal is to obtain a model that accurately predicts the dynamics of the system. However, the nonlinear effects caused by gaps, backlash, joints, as well as large displacements are not as well understood as the linear counterpart. In this sense, the Volterra series is an interesting tool for the analysis of nonlinear systems, since it is a generalization of the linear model based on the impulse response function. This paper applies the discrete-time Volterra series expanded in orthonormal Kautz functions to identify a model of a nonlinear benchmark system represented by a Duffing oscillator. The input and output data are used to identify the Volterra kernels of the structure. After the identification of the model, the linear and nonlinear components of the response of the system can be analyzed separately. The paper concludes by indicating the main advantages and drawbacks of this technique to model the behavior of nonlinear systems.