Abstract: Highlights•Neural networks embedded in PDEs can violate physical bounds unless constrained.•Motivated by established schemes, a new neural network approach is introduced.•It is demonstrated for subgrid-scale modeling for turbulent combustion.
Loading