Damage identification using time series analysis and sparse regularization

Abstract

Time series models have been popularly used in structural damage identification because the model coefficients and residual errors are sensitive to structural damages. As the direct relationship between the time series model coefficients (or residual errors) and damage severity is hard to be established, these methods provide limited information about the location or severity of damage. This study theoretically derives the sensitivity of the autoregressive coefficients of autoregressive moving average model to structural stiffness reduction factors. The autoregressive coefficients of both the damaged structure and the undamaged structure are extracted for damage identification. Structural damage identification is an inverse problem, and an underdetermined set of equations is a common obstacle encountered in solving such problem. Afterwards, sparse regularization is used to solve the underdetermined set of equations. The location and severity of damage is identified using the solution. The effectiveness of the proposed method is verified through a laboratory test of a cantilever beam and the Experimental Phase II IASC-ASCE benchmark structure. In addition, the proposed time series analysis-based method is promising to be used in online structural health monitoring systems.