Sparse Bayesian learning for structural damage detection under varying temperature conditions

Structural damage detection inevitably entails uncertainties, such as measurement noise and modelling errors. The existence of uncertainties may cause incorrect damage detection results. In addition, varying environmental conditions, especially temperature, may have a more significant effect on structural responses than structural damage does. Neglecting the temperature effects may make reliable damage detection difficult. In this study, a new vibration based damage detection technique that simultaneously considers the uncertainties and varying temperature conditions is developed in the sparse Bayesian learning framework. The structural vibration properties are treated as the function of both the damage parameter and varying temperature. The temperature effects on the vibration properties are incorporated into the Bayesian model updating on the basis of the quantitative relation between temperature and natural frequencies. The structural damage parameter and associated hyper-parameters are then solved through the iterative expectation-maxi mization technique. An experimental frame is utilized to demonstrate the effectiveness of the proposed damage detection method. The sparse damage is located and quantified correctly by considering the varying temperature conditions. (C) 2020 Elsevier Ltd. All rights reserved.