RESUMO
A passive data-driven method to localize a defect in a structure using the ambient noise is derived. The approach requires only acoustic measurements in a spatially random noise field and no knowledge of the structure. Measurements are taken before and after the perturbation has occurred and Green's functions are retrieved by cross-correlating acoustic measurements. The difference between measured data reveals the perturbation. A frequency domain method based on matched field processing is then performed to localize the perturbation. The Bartlett, minimum variance and white noise gain constraint processors are implemented and their performances are illustrated on a numerical experiment.
RESUMO
Recent studies show the ability of diffuse ultrasound to characterize surface breaking cracks in concrete. However, derived parameters are sensitive to the presence of partially opened zones along the crack whose pattern may differ from one sample to another. The aim of this letter is to study the variation of diffuse ultrasound parameters while the sample is driven by a low frequency bending load which alternatively opens and closes the crack, allowing to access supplementary information about its morphology. The results show the sensitivity of the method for various crack depths and highlight its potential for concrete nondestructive evaluation.