Identification of Nonlinear Behavior of Bridge Structures Using Time Series Analysis of Vibration Signals

Abstract

ABSTRACT One of the most widely used linear techniques for solving damage identification problems and prediction of the system dynamic behavior is modal testing and analysis. However, most of real structures exhibit nonlinear behavior during their lifetime caused by a number of various reasons. The presence of nonlinearity in a structural system changes its dynamic characteristics; hence, the use of linear techniques is improper for prediction of the system behavior. In this paper, a vibration-based nonlinearity identification technique was proposed to identify early changes in a dynamic system prior to significant damages. This technique combines the time series Autoregressive Moving Average with Exogenous Inputs (ARMAX) modelling, probability theory, and Fuzzy C-Means (FCM) clustering algorithm to categorize the linear and nonlinear dynamic behaviors of a dynamic system. The technique aims to categorize the linear and nonlinear behaviors of a structure, when it is subjected to various levels of excitation source. To show the validity of the proposed method, a series of shake table tests was performed on a six-span steel truss bridge model. The bridge model was excited using different amplitudes of ground motion to control the nonlinearity degree of rubber-based supports. It can be concluded from the analysis results that the new vibration-based nonlinearity identification technique was able to identify nonlinear behavior of the bridge model once it was subjected to different levels of earthquake excitations. This algorithm was able to categorize the linear and nonlinear behaviors of the bridge model and define a specific threshold for different excitation sources.