M.S. (Master of Science)
Department of Electrical Engineering
Every year, the Oil & Gas industry loses $2B dollars due to fugitive natural gas leaks. Identifying the source of a leak is a complex challenge, especially in suburban areas where gas leakage may be mixed with other sources. Besides the multitude of possible locations for a leak in an urban area, surveying an entire city may take considerable time and funds depending on the chosen method. This work proposes a framework based on unmanned areal vehicles (UAV) to survey a region for gas leaks. To accomplish this goal, we rely on the concept of a Upwind Survey Region (USR). The USR is a region that, if it contains a gas leak, it would produce a gas plume that contains the UAV. As the UAV flies over a region, multiple USRs are projected and we propose a statistical methodology to fuse the various USRs to identify regions that are clear of gas leaks and regions that are likely to contain a gas leak. Our methodology relies on simulating and processing acquired data, to estimate the location, orientation, and dimensions of USRs using machine learning models. The framework was validated using simulated data from the QUIC gas dispersion simulation software. Results show that the estimated regions were successfully classified.
Santiago Rodrigues Souza, Witenberg, "Developing a Machine Learning Framework for Upwind Surveyed Regions" (2021). Graduate Research Theses & Dissertations. 7628.
Northern Illinois University
Rights Statement 2
NIU theses are protected by copyright. They may be viewed from Huskie Commons for any purpose, but reproduction or distribution in any format is prohibited without the written permission of the authors.