Structural damage detection with two-stage modal information and sparse Bayesian learning

Abstract

Sparse Bayesian learning (SBL) has been proved to be an effective damage detection strategy. In this research, a structural damage detection technique with two-stage modal information is incorporated into the SBL framework. As an approximate representative, the equivalent damage factor requires sufficient sparsity for damage locating in the first stage. To content this, the mode shape energy (MSE) with SBL is performed in the model updating. With limited sensors, more accurate elemental damage will be estimated using mode shapes in the second stage. For further simplified process of damage detection, model reduction method is conducted to decrease both the calculation of eigenvalue and the second derivative corresponding to damage factors. The performance of SBL in damage detection is significantly improved. The accuracy and efficiency of the proposed method is verified by a bridge numerical simulation and a simply supported beam test.