A statistical pattern recognition based damage identification of a bridge using time series modeling of vehicle induced dynamic responses

Abstract

The deterioration of bridge conditions over time is attributed to various degradation processes, including creep, corrosion, cyclic loading, and others. As bridges age and undergo deterioration, monitoring their structural health has emerged as a significant concern in recent decades. The primary focus in this regard is the detection of damage.This research proposes an output only data driven technique for damage identification which will not require any physical model of the structure nor any known input excitation. This study introduces a damage detection technique that relies on evaluating the dynamic acceleration response of a bridge when subjected to the passage of a vehicle.Finite Element Method (FEM) is used to simulate the structure of the bridge and a dynamic half a car model is employed to depict the vehicle and its suspension system.The Vehicle-Bridge Interaction (VBI) is formulated incorporatingthe dynamic models for bridge structure subsystem and vehicle subsystem, interaction constraints and road roughness modelling. Newmark’s-β method is utilized for solving the coupled VBI problem to extract the vehicle-induced acceleration responses.Through the reduction of structural stiffness of a specific element within the bridge, an artificially induced damage is incorporated. The dynamic acceleration responses are fitted to linear and stationary time series models. An automatic model selection method and model order selection method are demonstrated in this study. The appropriate model class and model orders are selected according to the response from undamaged structure using the proposed technique. The model residuals are used as the damage sensitive features. A damage assessing index is presented utilizing a statistical distance metric named Kullback-LeiblerDivergence (KLD). The proposed damage index is used to identify different simulated damage scenarios. An experimental study is also performed on a laboratory scale bridge specimenwitha vehicle model passing over the bridge. The results of the study demonstrate the method’s effectivenessin identifying damage atvarious position of the bridge spanfor different severity levels.