Simulador de tomógrafo óptico para escoamentos bifásicos acelerado por GPU

Abstract

Optical tomography is a two-phase flow sensing technique that is non-intrusive and low-cost. In this technique, phase distribution images from the tube’s cross section are generated by the processing of light attenuation measurements. Until now, this processing is accomplished by image reconstruction techniques that disregard the system’s physics. Therefore, advanced reconstruction algorithms can improve the image’s resolution, thus increasing the range of application. In this thesis, an original optical tomography system simulator was developed to enable a detailed analysis of the physical behavior of light interacting with two-phase flows. As a consequence of that, new reconstruction algorithms can be created in further developments. The simulator models the direct problem, that is, the tomographic readings that are generated from flow data. The inverse problem, the flow reconstruction based on tomographic readings, will be the topic of future papers. The physical-based light simulation was based in the Path Tracing algorithm, traditionally used in computer graphics to solve global illumination problems. Due to the fact that the simulation is computationally intensive, a GPU implementation using the CUDA platform was developed. A comparison of simulated tomographic readings and real data shows that the simulator system works correctly.