GTC 2020: Extensions of TensorFlow-Based Computational Fluid Dynamics
After clicking “Watch Now” you will be prompted to login or join.
Click “Watch Now” to login or join the NVIDIA Developer Program.
Extensions of TensorFlow-Based Computational Fluid Dynamics
Dirk Van Essendelft, The National Energy Technology Laboratory
The National Energy Technology Laboratory (NETL) recently developed a completely TensorFlow-based computational fluid dynamics code for single-phase flow as a potential next-generation solver for Multiphase Flow with Interphase eXchanges (MFIX). NETL demonstrated over a 3x speedup using NVIDIA V100 GPUs on a DGX-1 before any domain decomposition was implemented. NETL is currently working on implementing parallel linear solvers in TensorFlow. In addition, NETL has strategically chosen to invest in a TensorFlow-based Multiphase Particle-In-Cell (MP-PIC) methodology. We'll cover the most recent developments to further the TensorFlow-based solver, and discuss plans for AI/ML acceleration.