Abstract: The talk will present the research findings of an ongoing project aimed at characterizing and controlling the propagation of mechanical discontinuity in crustal earth materials. In the first case study, unsupervised manifold-learning method is used to process triggers isolated from the continuous signals measured by the accelerometer array in the enhanced geothermal system (EGS) collab testbed. Second case study achieves accurate mapping of the spatiotemporal evolution of mechanical discontinuity (crack) in a simple 2D material irrespective of its dimension and composition. This requires efficient feature extraction, supervised learning and causal discovery. In the third case study, deep deterministic policy gradient algorithm is used to train four neural networks within a reinforcement learning framework to control the propagation of mechanical discontinuity in a simple 2D material. A well-designed reward function is essential for the robust control of the crack propagation.
Bio: Dr. Siddharth Misra is associate professor in the Harold Vance Department of Petroleum Engineering with a joint appointment in the Department of Geology and Geophysics at the Texas A&M University. He has published two books and developed 4 patented and 5 patent-pending technologies on machine learning and electromagnetic sensing implementations for subsurface characterization. In 2018, he was recognized as Department of Energy Early Career Awardee. In 2020, for his significant contributions to exploration geophysics and subsurface engineering, he received SEG J. Clarence Karcher Award, SPWLA Young Technical Professional Award, EAGE Arie Van Weelden Award, and SPE Gulf Coast Formation Evaluation Award.
The Energy for the Future Seminar Series is being presented as part of the College of Earth and Mineral Sciences 125th Anniversary celebration.