- This event has passed.
Ph.D. Proposal: Huda Aldosari
April 20, 2021 @ 2:00 pm - 3:30 pm EDT
Doctoral Dissertation Proposal
Title: New Architectures using Machine Learning to Detect and Predict Underwater Oil/Gas Pipelines Defects
Ph.D. Candidate: Huda Aldosari
Major Advisor: Dr. Reda Ammar
Associate Advisors: Dr. Sanguthevar Rajasekaran, Dr. Song Han, Dr. Raafat Elfouly ( External Associate Advisor)
Review Committee Members: Dr. Walter Krawec and Dr. Haitham Ghalwash
Date/Time: Tuesday, April 20th, 2021, 2:00 pm
Location:
Meeting link: https://uconn-cmr.webex.com/uconn-cmr/j.php?MTID=m34075a97d6be645c1ef438c227df1cc6
Password: uaPWGRdY243
Meeting number: 120 794 6755
Join by Phone: +1-415-655-0002
Access code: 120 794 6755
Abstract:
The largest cause of fuel consumption around the world is oil and gas. Most oil and gas are shipped by piping from one place to another. Due to aging, violent environmental causes, poor construction and insufficient safety and repair, residual pipelines worldwide have been exposed to corrosion. This also involves thorough servicing, restoration, and renewal procedures or removing parts to ensure optimum efficiency. In the event of infrastructure breakdown, pipeline reliability is of primary consideration to oil and gas producers, regulatory officials, customers, and other stakeholders because of its detrimental impacts on human health, safety, and considerable economic implications. Pipeline accidents can never be eliminated. However, the cumulative number of fatalities can be minimized to an adequate extent by choosing appropriate risk control techniques. Although various researchers have done the right amount of work by proposing different monitoring architectures to make oil and gas pipelines more reliable and defectless, these pipelines usually suffer from some corrosions or defects which require an immediate response from an onshore control station to avert any catastrophe that may happen. Oil and gas pipelines still need a robust and reliable monitoring architecture to help relative authorities inspect these pipelines in real-time with accurate and less information. In this study, an oil and gas dataset comprised of magnetic flux leakage signals has been used to design various robust and reliable monitoring architectures to help relative authorities inspect these pipelines in real-time with accurate and less information. For the analysis and localization of pipeline faults, machine learning-based various strategies have been proposed. To design such monitoring architecture the optimization of machine learning algorithms has been also considered in this work. The presented results can be utilized to design an automated monitoring architecture embedded with machine learning techniques to monitor and predict the oil and gas pipeline defects in real-time.
