Characterization of three-dimensional surface-breaking slots based on regression analysis of ultrasonic Rayleigh wave simulations

Abstract

Rayleigh waves travel along the surface of a solid structure, with most of their energy focusing within a depth of one wavelength. Thus, the reflection coefficient from a surface-breaking crack is highly sensitive to the ratio between the crack depth and the wavelength. It is possible to characterize the depth of surface-breaking cracks by measuring the features in the reflected waves. However, a feature value can correspond to multiple depth-wavelength ratios, i.e., the mapping is non-univalent, which brings difficulties for crack sizing using the feature. In this work, we use finite element method (FEM) software to perform 3-D numerical analysis on the interaction between Rayleigh waves and surface-breaking slots with various 3-D geometries. Multiple features are selected based on the nearest neighbour regression analysis on a numerical dataset, ensuring that a univalent mapping relationship from the selected features to the slot depth can be established. This relationship is then experimentally used to predict the depth of real slots with different geometries, showing reasonable accuracy.