Optimizing Flow Control with Deep Reinforcement Learning: Plasma Actuator Placement around a Square Cylinder
The DRL code used in this study is built upon the
open-source DRL code developed by Rabault et al. The DRL code follows the
structure of a PPO agent implemented using the Tensorforce library. For the
simulation environment, we have employed OpenFOAM-6.
Installation
python3
and pip3: sudo apt install python3-pip
tensorflow: pip3 install tensorflow==1.13.1
tensorforce: pip3 install tensorforce==0.4.2
How to run
Compile
the solver pimpleplasmaFoam in OpenFOAM-6: wmake
Train the
DRL model in folder mode_180_3: python3 training.py
Test the model
in folder model_180_3: python3 testing.py