Ultrasonic detection of contour node represented voids and cracks

Abstract

In current evaluation of ancient stone monuments, 2-D ultrasonic tomography is widely used. The detection result, however, has deviation in the location and the shape since the ultrasonic velocity of voids and cracks area change much more drastic than the healthy parts. Base on the homogeneous material assumption, a contour node representation is therefore proposed fitting the best defect shape and location. With the contour node representation, a quick ray tracing method is proposed. This inverse problem is solved under the Bayesian framework in which we use Markov chain Monte Carlo to search the global optimal result. Through applying more prior knowledge to the model and the quick ray tracing model, the computational cost of the inverse process is highly reduced. Compared to the ultrasonic tomography result, the proposed method has more accurate location and shape in 2-D inspection. Finally, the laboratory experiment also tests the ability of the proposed method.