Professor Peter Kang

This research group focuses on the physical processes of flow, transport, mixing, and reaction in subsurface systems. They study the physical processes across scales ranging from tiny pore-scale (~mm) to large field scale (~km). Each group member is working on different scales, and their research needs access to MSI's high-performance computing resources. The projects are listed below:

• Flow and Reactive Transport at Single Fracture Scale (~10cm): The researchers study a variety of geochemical and hydrological processes in the single fracture scale. The physical length of the single fracture systems ranges from 1mm to 10cm. The researchers simulate flow, transport, chemical mixing, and reaction in this small-scale system by solving the Navier-Stokes equation for flow and the advection-diffusion-reaction equation for reactive transport. They conduct the required direct numerical simulations using multiple computational fluid dynamics software, including OpenFOAM, COMSOL, and MATLAB. The modeling domain is in three-dimensional space, implying additional requirements for computing power.
• Flow and Reactive Transport at Fracture Intersection (~1m): This research studies how the processes of flow-transport-mixing-reaction at different single fractures interact with each other. For this purpose, the researchers systematically change the condition of geometry, flow, and diffusion regimes. For example, they investigate the effects of the inclination between the two fractures, fracture roughness, inertial regimes, diffusivity, and reaction rates. As the number of controlling factors is large, the total number of their combinations is huge. Thus, it is critical to have access to high-performance computing resources at MSI for this research project.

• Flow and Reactive Transport at Fracture Network Scale (~100m): The ultimate goal of this research is to apply improved understanding of flow and transport in fractured media to field scale. This requires consideration of fracture network systems whose length scale can increase beyond 1km. The researchers will use cutting-edge software such as dfnWorks and PFLOTRAN to achieve the research goal. The software is designed to run on massively parallel computing architectures to relieve the computational limitations of large-scale domain systems. The researchers will conduct a global sensitivity analysis by systematically changing the values of controlling factors, such as conductivity, fracture intensity, and dispersivity, to find out the most significant factor that determines the flow-transport-mixing-reaction processes. This ambitious research project is achievable only with state-of-the-art computing powers like the supercomputing resources at MSI. Meaningful preliminary results have already been achieved via running a few simulations using MSI clusters.

• Machine Learning-Based Modeling of Geothermal Systems: Experimental and computational approaches to flow and transport processes have been actively studied in hydrogeology. Advances in computing power and data acquisition techniques have called for integrated approaches to studying flow and transport processes. This research project aims to develop a model-data integration framework which will establish an efficient surrogate model. For this, the researchers need the capacity to train the model by integrating it with a large amount of data. Because it is inherently difficult to collect such a large amount of data from subsurface systems, they will produce the data from a large number of numerical simulations. The most significant obstacle to the research project is that each run of the simulation is computationally demanding, and thus the total computing requirements are huge. Again, the computing resources at MSI are critical to the success of this project.