Hi, friends

Is there any difference between the solvers of CFX4.4 and CFX5.6?

I am modelling natural convection within a simple rectangular box with a small heat patch on the middle of its bottom and two openings at high and low level, incompressible buoyant flow with Boussinesq approximation. The steady state simulation using CFX4.4 provides very nice and robust solution for the flow field, except sometimes the residuals of momentum equations seemed to be unsteady. But when I am using CFX5.6 to model the same case, the steady state solution has terrible convergence problem. The Rayleigh number calculated in the output file is around 1.5E11 which is supposed to be fully turbulent based on the CFX5.6 menu.

Tried to import the CFX4.4 mesh into CFX5.6, still had the same convergence difficulty no matter how small the physical time step is.

Transient state procedure seems to give me more reasonable answer, but the timestep has to be very small, 0.1s at starting stage, unbearable time-consuming by this way.

Why the two codes didn't give me the similar answer for modelling the same case by steady state?

Anyone has any feelings about comparing this two code before?

Thanks in advance for your comments

Regards Forrest

Hi Forrest,

The CFX-5 and CFX-4 solvers are very different. CFX-4 uses a cell centered discretization and segregated solver approach, whereas CFX-5 is node based, having many more integration points, and employs a coupled-multigrid solver. CFX-5 is largely based on CFX-TASCflow numerics.

I would discuss this with your support rep. The terminology for boundary conditions, physics, and numercis in CFX-5 is slightly different and you may have set the problems up differently. CFX-5 is far more robust than CFX-4, so I wouldn't expect to see this kind of issue. You should also be careful about your timestep selection in CFX-5, smaller is not always better.

Regards, Robin

Hi, Robin

Thank you for your comments.

I've been talking to one of the CFX engineers when I was on the two days general introduction course a few months ago. He agreed with the long reputation of CFX4 for modelling buoyant flow, especially with simple rectangular geometry. It should be better to consult CFX rep as you suggested and hopefully, I can get through this soon.

Just let you know, the case I am modelling physically has a small timescale. The auto timestep selection gives me a timestep smaller than 1sec. Tried to use a larger timestep but got a bouncy convergence behaviour, RMS never going to be down to 1E-4. The period of oscillation of the residual plot changes using different physical timestep seems to be similar, which gave me a feeling about transient effects. Transient solution will get me nowhere since the uncertainties (at least now) of the momentum resistances within the sub-domains used (porous volume porosity and resistance constants corresponding with CFX4) plus the time-consuming.

thank you for your help again.

Best wishes Forrest