81 lines
2.8 KiB
Plaintext
81 lines
2.8 KiB
Plaintext
|
Download the report on pyCALC-RANS at
|
||
|
|
|
||
|
|
https://www.tfd.chalmers.se/~lada/postscript_files/py-calc-rans.pdf
|
||
|
|
|
||
|
|
|
||
|
|
For instructions how to run the code, see Section 11.3.
|
||
|
|
|
||
|
|
|
||
|
|
***********
|
||
|
|
|
||
|
|
The NN model for creating the NN model for EARSM is found in the folder NN/
|
||
|
|
|
||
|
|
The NN model is incorporated in pyCALC-RANS in the folder channel-10000-earsm-NN/
|
||
|
|
|
||
|
|
The work is presented in the paper
|
||
|
|
|
||
|
|
L. Davidson, "Using Neural Network for Improving an Explicit Algebraic Stress Model in 2D Flow",
|
||
|
|
J. Tyacke and N. R. Vadlamani (eds.), Proceedings of the Cambridge Unsteady Flow
|
||
|
|
Symposium 2024, pp. 37--53, 2025,
|
||
|
|
|
||
|
|
https://doi.org/10.1007/978-3-031-69035-8_2
|
||
|
|
|
||
|
|
|
||
|
|
***********
|
||
|
|
|
||
|
|
The PINN model for computing prand_k is located in folder PINN-etmm15/
|
||
|
|
|
||
|
|
compute vist_{t,PINN}: compute-vist_k-solving-PDE-using-PINN.py
|
||
|
|
|
||
|
|
Then run compute-c_k-and-c_omega_2-from-balance-of-k-and-omega-eqns.py
|
||
|
|
|
||
|
|
It is used in the CFD code in folder channel-2000-half-channel-PINN/
|
||
|
|
|
||
|
|
It is used in the paper
|
||
|
|
|
||
|
|
L. Davidson,
|
||
|
|
"Using Physical Informed Neural Network (PINN) to Improve a k-omega Turbulence Model",
|
||
|
|
ERCOFTAC Symposium on Engineering Turbulence Modelling and Measurements (ETMM-15), Dubrovnik on 22-24 September 2025.
|
||
|
|
https://www.tfd.chalmers.se/~lada/postscript_files/Using-Physical-Informed-Neural-Network-PINN-to-Improve-a-k-omega-Turbulence-Model.pdf
|
||
|
|
|
||
|
|
https://www.tfd.chalmers.se/~lada/PINN-improve-a-k-omega-turbulence-model.html
|
||
|
|
|
||
|
|
***********
|
||
|
|
|
||
|
|
PINN and NN scripts are found in folder PINN-NN
|
||
|
|
|
||
|
|
** vist-diffusion-pinn-5200-half-channel-save.py
|
||
|
|
|
||
|
|
** vist-diffusion-pinn-5200-half-channel-save.py
|
||
|
|
|
||
|
|
are used for solving the eq for \nu_{t,PINN}. The former is used to create good initial weights and
|
||
|
|
biases which are stored in vist-diffusion-pinn-5200-half-channel-save.ct. The latter
|
||
|
|
is then used for converging the \nu_{t,PINN} eq. by loading vist-diffusion-pinn-5200-half-channel-save.ct.
|
||
|
|
|
||
|
|
** compute-c_k-and-c_omega_2-from-balance-of-k-and-omega-eqns.py
|
||
|
|
|
||
|
|
is used for computing C_{k,PINN} and C_{\omega 2, PINN}
|
||
|
|
|
||
|
|
create NN model for c_{K,NN}: PINN-NN/neural-k-omega-c_k-vist-over-y-and-uv_tot.py
|
||
|
|
|
||
|
|
create NN model for c_{omega 2,NN}: PINN-NN/neural-k-omega-c_omega_2-vist-over-y-and-uv_tot.py
|
||
|
|
|
||
|
|
create NN model for prand_{k,NN}: PINN-NN/neural-k-omega-prand_k-vist-over-y-and-uv_tot.py
|
||
|
|
|
||
|
|
replacd the NN model for c_{K,NN} by pySR (Symbolic Regression):
|
||
|
|
|
||
|
|
Use NN models in CFD code:
|
||
|
|
|
||
|
|
channel-10000-half-channel-NN-PINN-from-channel-NN-vist-over-y-uv_tot-2nd-submission-t_int-3000/
|
||
|
|
|
||
|
|
boundary-layer-k-omega-ni-150-nj-100-yfac-yplus-0.8-ymax-4.5-ML-NN-from-channel-NN-vist-over-y-uv_tot-2nd-submission/
|
||
|
|
|
||
|
|
The PINN model is used in:
|
||
|
|
|
||
|
|
channel-5200-half-channel-PINN-vist-over-y-uv_tot-2nd-submission
|
||
|
|
|
||
|
|
Lars Davidson. Using Physics Informed Neural Network (PINN) and Neural Network (NN) to Improve a $k-\omega$ Turbulence Model
|
||
|
|
|
||
|
|
https://www.tfd.chalmers.se/~lada/Using-Physical-Informed-Neural-Network-PINN-and-NN-improve-a-k-omega-turbulence-model.html
|
||
|
|
|