Easy question about residuals
Hi all!
I've sometimes read posts asking why residuals don't go down as far as expected and the asnwer is that this could be due to (scaled) residuals are based in the initial solution and if it is good enough, residuals won't decrease so much. Now imagine the following situation: I run a simulation in 1st order discretization schemes until I reach a continuity residual of 1e-4 after 200-300 iterations. Then, I switch to 2nd order schemes and I continue my simulation until it is converged (all parameters like residuals, Cd, Cl, mass-weighted avg... remain more or less constant), but now continuity residual is higher than 1e-3: is this because this last residual is calculated from the solution I got in the 1st order simulation and, so, residual cannot go down further or simply this high residual show that there's something in the simulation that doesn't work properly? Many thanks! |
Re: Easy question about residuals
Scaled residuals are different from normalized residuals. Please check the manual!! The 1st order upwind is much more robust then the 2nd order and originates a more diffusive solution that is numerically better for residuals (chech the manual). Don't look too much at the absolute value of residuals but at their oscillation and relative decrement/increment.
|
Re: Easy question about residuals
dfadsfa
|
All times are GMT -4. The time now is 10:19. |