We use laboratory experiments, computational methods, and Earth observations to advance our knowledge in rock fractures and fracturing, rock-fluid interactions, subsurface stress, material integrity, permeability and hazard avoidance.
Our team covers broad research topics including geothermal energy, fossil energy, carbon dioxide sequestration and storage, nuclear waste, basic energy sciences and national security. We comprise experts in rock fracture and fluid flow, seismology, data science and machine learning, computational imaging, and geophysical inversion and modeling.
What We Do
- Rock fracture and fluid flow
- Data science and machine learning
- Inversion and modeling
- Observational seismology
- Rock fractures and fracturing
- Computational imaging and Parameter estimations
- Well integrity and cementitious material characterization
- Carbon sequestration
- Geothermal exploration and exploitation
- Rock-fluid interactions and fracturing
- Material integrity
- Subsurface stress analysis
- Computational imaging (subsurface, medical, etc.)
- Scientific machine learning
Recent advances in algorithms and computing provide an opportunity for efficient, data-driven solutions to previously infeasible problems. The excellent performance of learning-based methods arises from their ability to exploit large amounts of high-quality training data without the need for hand-designed features. However, many scientific disciplines are not a data-rich domains (e.g., subsurface geophysics). To alleviate the data scarcity issue and improve model generalization, there has been growing interest in combining physics knowledge with machine learning (ML). This project aims to incorporate physics knowledge with ML (primarily deep neural networks) to solve challenging problems from energy, Earth, environment, medicine, etc.
Waves are ubiquitous in our world; there are seismic waves in the Earth, ultrasonic waves in materials, acoustic waves in water, electromagnetic waves in space, and many others. Well-established physical models describe how these waves propagate and how that propagation is affected by the intervening media. Wave properties (amplitude, phase, arrival time, etc.) thus carry information about the space through which they travel. The goal of this project is to infer physical properties, such as density and bulk modulus, from measurements of the wave signal. Various computational methods have been developed based machine learning and physics.
Spatio-temporal measurements and observations are ubiquitous in many scientific disciplines. Spatio-temporal data analysis becomes an emerging research area due to the development and application of novel computational techniques allowing for large-scale data. This project is to develop advanced spatio-temporal data analysis techniques with applications to problems from Earth science, biology and space science to demonstrate the efficacy of the techniques.
Modeling of seismic wave propagation is critical to seismology. It plays an important role in many tasks, ranging from survey design methods to imaging and inversion. This project is to develop accurate and efficient seismic wave modeling simulation approaches and effective tools; methods based on machine learning and finite-difference strategies to understand the character of recorded seismic data and to explain wave phenomena that arise in complex, heterogeneous earth models governed by, for example, acoustic, isotropic, anisotropic, viscoelastic, or poroelastic rheologies.
Leakage detection is the cornerstone of any carbon capture and storage (CCS) project. If leaked, the pollutant carbon dioxide (CO2) can migrate from the storage formation to the surface or into other geological formations. In addition to defeating the purpose of CCS, the leaked CO2 poses a set of environmental challenges, such as increasing the acidity of the fresh groundwater and releasing heavy metals. Ensuring the CO2 does not leak from these reservoirs in this time period is the key to any successful CCS project. This project aims to develop cost effective CO2 leakage monitoring approaches, which would yield high sensitivity and detectability. Our effort ranges from those based on traditional geophysical imaging methods to the most recent techniques built on machine learning and data science.