DeepFlame is a deep learning empowered computational fluid dynamics package for single or multiphase, laminar or turbulent, reacting flows at all speeds. It aims to provide an open-source platform to combine the individual strengths of OpenFOAM, Cantera, and PyTorch libraries for deep learning assisted reacting flow simulations. It also has the scope to incorporate next-generation heterogenous supercomputing and AI acceleration infrastructures such as GPU and FPGA.
The deep learning algorithms and models used in the DeepFlame tutorial examples are developed and trained independently by our collaborator team – DeepCombustion. Please refer to their website for detailed information.
Note
This project is under active development.
qs/install qs/download_dnn_models qs/examples qs/input
solvers/df0DFoam solvers/dfLowMachFoam solvers/dfHighSpeedFoam solvers/dfSprayFoam
utility
contributors citation gpl
pr