Nano-scale transistors fill warehouse-scale supercomputers, yet their performance still constrains development of the jets that defend us, the medical therapies our lives depend upon, and the renewable energy sources that will power our generation into the next. We are the Computational Physics Group at Georgia Tech CSE and we develop state-of-the-art computational models and numerical methods to push these applications forward. Formulations leverage domain expertise in physics and biology and data-driven tools like machine learning and data assimilation. Our open-source scientific software utilizes these methods and scales to the world’s largest computers. Check out the links above to learn more, or visit this page if you’re interested in joining our group.
Bubble cavitation can ablate kidney stones, but wreaks havoc on marine propellers. We developed a data-driven sub-grid method for simulating this phenomenon. It utilizes a LSTM recurrent neural network to close the governing equations at low cost. MFC, our open-source multi-phase flow solver, demonstrates this method. MFC is also capable of fully-resolved multi-phase fluid dynamics via the diffuse-interface method.
The spectral boundary integral method leads to high-fidelity prediction and analysis of blood cells transitioning to chaos in a microfluidic device (above). We developed a low-order model for the cell-scale flow, important for guiding microfluidic device design and improving treatment outcomes.
20 September, 2022 Preprint on application performance on ARM+GPU supercomputers submitted and posted on the arXiv! Anand and Henry contributed the MFC portion.
24 August, 2022 NVIDIA awards the group four BlueField2 DPUs!
18 August, 2022 Congrats to Ph.D. student Jack Song on winning a CRNCH fellowship to work on quantum computing for PDEs!
8 August, 2022 Ben Wilfong joins the group as a PhD student! He’s working on multiphase computational fluid dynamics.
22 June, 2022 AMD donates an MI200-series GPU server to the group!