Ultrasound Tomography Based on the Inverse Coefficient Problem as a Way to Combat Pattern Noise

Abstract

The paper proposes to use the ultrasound tomography method based on the solution of the inverse coefficient problem to reduce the level of pattern noise. Mathematical models used in ultrasound tomography well describe such physical effects as refraction, diffraction, and rescattering. It is logical to expect that reconstruction of the internal structure of metallic samples using ultrasound tomography is more efficient compared to digital antenna focusing (DFA) techniques. Due to the nonlinearity of the inverse problem of ultrasound tomography, an iterative MultiStage method is used to ensure convergence to the global minima of the residual functional. The paper presents the results of numerical experiments to reconstruct the image of the internal structure of a welded joint that may contain side drilled holes and crack models. A domain of welded metal is represented in the form of sections constructed according to the principle of Voronoi diagrams. In each section, the velocity is constant and its value is randomly distributed. In the model adopted in the paper, the pattern noise is formed due to multiple scattering at the boundaries of sections with different sound velocities. It was assumed that the antenna array is located on the outer surface of the test object of known thickness. The obtained results show that the tomographic method allows one to determine the shape and speed of sound in low-contrast reflectors, for which the DFA method is ineffective.