[kabz]

So i've implemented residuals (rms residuals) and also convergence between two iterations (rms convergence). I get oscillating residuals but all below 1e-5 whereas my convergence is monotonically decreasing. The velocity profiles predicted look okay. So which should i use as my stopping criteria, the residuals or the convergence?

[FMDenaro]

Originally Posted by kabz

So i've implemented residuals (rms residuals) and also convergence between two iterations (rms convergence). I get oscillating residuals but all below 1e-5 whereas my convergence is monotonically decreasing. The velocity profiles predicted look okay. So which should i use as my stopping criteria, the residuals or the convergence?

difference bewteen two iterations is not meaningful for the real convergence

[kabz]

Originally Posted by FMDenaro

difference bewteen two iterations is not meaningful for the real convergence

Can I use both? If i was to rely on residuals alone, my algorithm would stop within 2 or 3 iterations (since they are always below 1e-5)

[FMDenaro]

Originally Posted by kabz

Can I use both? If i was to rely on residuals alone, my algorithm would stop within 2 or 3 iterations (since they are always below 1e-5)

the residual must diminuish below such value