BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Amzi Jeffs (remote) (Pacific Northwest National Laboratory) DTSTART:20260324T223000Z DTEND:20260324T233000Z DTSTAMP:20260513T193329Z UID:SFUOR/68 DESCRIPTION:Title: A I-Based Algorithms for Real-Time Anomaly Detection in Ground Penetrating R adar Data\nby Amzi Jeffs (remote) (Pacific Northwest National Laborato ry) as part of PIMS-CORDS SFU Operations Research Seminar\n\nLecture held in ASB 10908.\n\nAbstract\nGeophysical inversion is a well-established met hod for conducting non-invasive surveys of underground\nanomalies\, such a s buried infrastructure. Typically\, inversion is performed using numerica l methods:\ngradient descent and physics-based simulations are used to cre ate a subsurface model that accurately\nexplains the surface surveys. Howe ver\, these methods are extremely computationally expensive\,\nsometimes t aking days or weeks to complete. Additionally\, the inverse problems are h ighly\nunderdetermined\, and must be carefully regularized based on expert input. This creates a challenging\nbottleneck for field operators\, who m ust accept lengthy delays between conducting surveys and starting\nwork.\n \nAI-based methods can help overcome these challenges. We will discuss adv ances in this direction based\non work for the Grid Overhaul with Proactiv e\, High-Speed Undergrounding for Reliability\, Resilience\,\nand Security (GOPHURRS) project. Our approach uses thousands of geophysical field surv eys to train\nconvolutional neural networks that can perform subsurface an omaly detection in near real-time. We will\ndiscuss the strengths and limi tations of this AI-based approach\, best practices for using physical data \nwith AI\, and our approach to optimizing our overall workflow.\n LOCATION:https://researchseminars.org/talk/SFUOR/68/ END:VEVENT END:VCALENDAR