dear Mr. and Mrs.
I am not very sure about which factor affect the convergence in cfd,especially in flunet calculaion.How improtant is the under-relaxation factor play the role in fluent calculation and how to set it in order to achieve convergence.
Re: convergence condition
There are many factors, I think, to affect numerical convergence. The most frequent factors which they meet may be the 'quality of grid net(mesh quality)' and under-relaxation factor.
(1) When I was trying the analysis of cyclone separator, one cell near the cyclone inlet was fairly much skewed and I could not obtain converged solution. I re-constructed grid-net to reduce skewness of that cell(only one cell), and the calculation was converged.
(2) When my junior engineer was trying to simulate gas combustion with species under-relaxation factors 1.0(default setting), he could see only oscillating residuals. I recommended to reduce the factors to 0.9 for every species, he could obtain the converged solution.
In addition, there may be many factors.
(3) As far as my experience is concerned, finite-rate-reaction-model is more difficult than eddy-dissipation-reaction-model for convergence. Then, first, use eddy-dissipation-model and next switch to finite-rate/eddy-dissipation or finite-rate-model.
(4) For high swirl, first low swirl and next increase swirl strength.
(5) Sometimes, the basic nature of physical situation itself has problem for convergence. For example, free vortex(inner rotational velocity is strong than outer velocity) is hydrodynamically unstable so that it is difficult to be converged.
(6) If the physical situation itself has unsteadiness, it is very difficult to obtain converged steady solution. For example, Benard convection, Taylor vortex, Kalman vortex and so on. Actually speaking, they should approach transient calculation. If someone try steady calculation, it is difficult to be converged.
(7) And many which I may or may not know ......
|All times are GMT -4. The time now is 13:36.|