Simulations in Problems of Ultrasonic Tomographic Testing of Flat Objects on a Supercomputer

Abstract

This paper presents a computer simulation study on a problem of ultrasonic tomographic imaging of welded joints in flat metal objects. The developed algorithms and supercomputer software for reconstructing sound speed images and identifying defects in welded joints were tested on model problems. A specific feature of welded joint inspection is that in most cases the inspected object is accessible only from a single side. This study investigates a tomographic scheme in which a flat object is sounded from a single side by transducer arrays, and reflections from the flat bottom of the object are taken into account. Various schemes of tomographic imaging are investigated and compared for the cases in which the position of the bottom of the object is known, partially known or unknown. The results of this study showed that taking into account reflections from the bottom is of fundamental importance for high-quality image reconstruction. This method allows us to greatly increase the angular range of sounding and to register the waves transmitted through the object in order to achieve high-precision sound speed image reconstruction. The computations were carried out on “Lomonosov” supercomputer at Lomonosov Moscow State University. The developed software was optimized on a supercomputer for varying number of MPI processes per node. The optimal setup yielded more than a two-fold performance increase.