See LinkedIn article for professional discussion: www.linkedin.c... See Reddit post for non-professional discussion: / implementing_adjointou... Add me: / davidpenner96
what's the practical point of refining the mesh to a point where the solver converges to 10^-7 error margin? at that point any model inaccuracy (ideal gas, steady flow, etc) will have 10000x greater effect actual on divergence from the experimental data.
To be clear, it's not 10^-7 convergence error/residual, but rather a 10^-7 error (per cell) in the cost function, such as Cl or Cd. Regardless, there's never harm in exploring more efficient mesh refinement strategies - even if they're taken to the extreme for the sake of demonstration, they can still be used to make more efficient meshes at a more practical level.
what's the practical point of refining the mesh to a point where the solver converges to 10^-7 error margin? at that point any model inaccuracy (ideal gas, steady flow, etc) will have 10000x greater effect actual on divergence from the experimental data.
To be clear, it's not 10^-7 convergence error/residual, but rather a 10^-7 error (per cell) in the cost function, such as Cl or Cd. Regardless, there's never harm in exploring more efficient mesh refinement strategies - even if they're taken to the extreme for the sake of demonstration, they can still be used to make more efficient meshes at a more practical level.