In the case of convergence
 

Hi dear friends,

I have some question about converging:

I am simulating a turbine, I want to converge this CFD model to convergence Criterion of 1e-6, But the solution is not converged to this criterion. Maybe I don't know that how to converge and when the solution is convergence. I started with the default values for all under relaxation factors(URF). In iteration 900, value of torque is 0.087 and the solution is converged in convergence Criterion of 1e-4 and then I have changed convergence Criterion to 10e-5. In iteration 17750, value of torque is 0.0222 and In after iteration 17750,I change momentum and pressure to 0.1.Finally, the solution is converrged in iteration 18000, and again I change convergence Criterion to 10e-6, But the residuals are almost constant and they didn't decreased until the solution reaches to convergence Criterion of 1e-6. Also value of torque is 0.0222 in iteration 19500.But still It does not reach to convergence Criterion of 1e-6.

Now according above explanations, Is the solution convergence?
I want to know that when the iterations increase and the residuals are almost constant, Can the solution change? (in here , the solution is torque), or because the residuals are almost constant, the solution is not changes. Correct?
I want to know that what does your opinion?

I am grateful that guide me about this problem and bold sentences.

Thanks,

Best,

Aja

Yes, I would say the solution has converged. The residuals are a good indication if the solution is converging or diverging, but you should not attach to them too much. depending on the mesh, initial guess, etc. in some cases the solution may not (significantly) change if the residuals are below 10^-3, in other cases the solution may still significantly change below 10^-6. So, best practice is to look at physical monitors - like torque in your case - and check whether these monitors are sufficiently constant (say, fluctuate less than 1%)

In your case, the torque did not change at the level of digits that you mentioned over a range of 2000 iterations, so the solution seems to be stable and constant to me.