I was wondering if these residuals are ok because they are spiking a little?
thank you,
frank lucci
Hey Frank,
With respect to the spikes, this is normal for the numeric solvers employed.
What I think that can be improved here is the solver tolerances. Notice that at around iteration number 180 the k field gets flat, and the same happens for Uy at around 300. What is happening here is that the solver assumes it is converged and stops solving those fields. But this affects the other related fields (p, Ux, Uz), thus the other curves get flat without having reached the tolerance value. You can prevent this by reducing the tolerances down in the numerics section.
1 Like
Dear @FrankTheTank, I would not call that a spike. It’s basically showing you that for the numerical configuration and domain you have, you can’t improve the solution any further. If you notice, the algorithm is trying to reduce the pressure residuals by making small changes in U(x,y,z) these changes make small fluctuations in the residuals of U(x,y,z), which are the “spikes” you see. As @ggiraldof says, sometimes you can “smooth” the solution by modifying the tolerance configuration for a further reduction in the residuals. The “spikes” are large changes in the residuals from one iteration to another (or one time step to another).
2 Likes