Advanced computational methods in acoustic tomography.

Abstract

An alternative approach is based on the field theory of wave propagation. A finite difference solution of the wave equation for an anisotropic medium is used. The material properties can be reconstructed using a modification of the traditional ART (algebraic reconstruction technique). The rotation angle of a fiber-reinforced laminate is reconstructed in this study. The algorithm is successful in detecting two, five and ten-degree misalignments in an orthotropic composite laminate.

A matrix domain decomposition of the ray tracer and the tomography algorithm is presented. This technique is applied to a heterogeneous isotropic material and the velocity is reconstructed. Each processor is assigned a grid point in the discretized model, allowing all source-receiver pair calculations to be performed simultaneously. The CPU time remains relatively constant as the array size increases.

In this work, two methods of acoustic tomography are introduced. Advanced computational methods are implemented to handle the complexity of the tomographic algorithm in isotropic and anisotropic materials. Parallel processing is a technique used to improve the performance and thus the versatility of the method. The speed-up using parallel processing techniques is presented.