Experimental study of kriging-based crack identification in plate structure

One of the major issues that existing crack identification methods utilizing dynamic responses are facing is the limitation of engineering feasibility. How to suppress the effect of measurement noise and improve the identification accuracy is still challenging. In this work, an effective method is proposed to identify the size of an arbitrary internal crack in plate structure based on a Kriging surrogate model, and a series of laboratory tests are designed to verify the practicability of this strategy. The initial Kriging surrogate model is constructed by samples of crack parameters (tip locations) and corresponding root mean square (RMS) of random responses as the inputs and outputs, respectively. To further improve the surrogate accuracy and reduce computational cost during the inverse problem, an optimal point-adding process for Kriging model updating is then carried out. Experimental results of crack identification in a cantilever plate indicate that the proposed method can be an alternative to conventional crack detection methods even in the presence of measurement noise and modeling errors.