CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Transient CHT (http://www.cfd-online.com/Forums/cfx/72777-transient-cht.html)

 Siva_aero February 17, 2010 14:21

Transient CHT

Hello All,

I am working on a Transient CHT analysis problem. In which the flow enters in to a big cavity at a Mach number of 0.2. due to the large area the velocity looses its magnitude and the maximum velocity inside the cavity worked out to be 2 to 5 m/s.
There are 2 fluid domains and 4 solid domains. the simulation (change in BCs every 300 Secs) need to be performed for 1800 Secs (using a time step of 0.5 sec currently). Node count of the TET mesh is 0.22 Million (1 element extruded mesh). the solving time for one outer loop iteration worked out to be 5 Mins.

Based on this the total time required to complete 1800 secs seems to end up in weeks.

Could anybody help me out in a way so that the transient simulaiton time can be reduced considerably.

 ghorrocks February 17, 2010 17:52

You have to do a sensitivity study on convergence, mesh and time step size. Only after that can you say you have optimised the CFD simulation - and that may well tell you your run time has to be longer than it currently is, not shorter.

I guess by your comment "1 element extruded mesh" you are doing a 2D simulation? If you are doing 2D stuff and run time is a problem then CFX is the wrong software for you. CFX does not have a proper 2D solver, it solves 2D problems by solving the 3D equations and the equations in the third dimension just end up being zero. This is highly inefficient. You will get an enormous improvement in run time by going to a CFD code which has a true 2D solver like Fluent or CFD-ACE.

A true 2D solver must be the most requested feature for CFX but they have never provided. It is a clear weakness of the CFX code.

But on the other hand, if you have a Mach 0.2 jet going into a big cavity I doubt the flow is 2D anyway. Have you checked your 2D assumption is valid?

 Siva_aero February 18, 2010 04:50

Hi Glenn, Thanks a lot for ur reply. i did a Mesh sensitivity check and the Mesh was good enough. The max Y+ was observed to be 4 (SST model). the time step size was calculated based on the 1D conduction equation. however to decrease solver time i increased the time step after 100 time steps. will this be OK?

Yes, it is a 2D problem but the sides were defined with periodicity. Since we didnt Benchmark other CFD software it may not be possible to move to other CFD code for 2D simulation.

The flow as such has a high Tangential component. From the results i feel the 3D features of the flow was captured.

Will it be OK if i reduce my no of coefficient loops from existing 5 to 1.

 ghorrocks February 18, 2010 17:53

Quote:
 the time step size was calculated based on the 1D conduction equation. however to decrease solver time i increased the time step after 100 time steps. will this be OK?
You need to determine this with a sensitivity check.

Quote:
 it is a 2D problem but the sides were defined with periodicity.
You would only do this if the flow has a third dimension component. If the flow is 2D you should use symmetry planes rather than periodic boundaries.

Quote:
 From the results i feel the 3D features of the flow was captured.
What do you mean? How did you capture 3D features with a 2D model?

Quote:
 Will it be OK if i reduce my no of coefficient loops from existing 5 to 1.
You need to determine that with a sensitivity check. But I would be surprised if that works. Generally 3-5 coeff loops per timestep is best.

 All times are GMT -4. The time now is 23:24.