CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   experiment and simulation comparison (http://www.cfd-online.com/Forums/cfx/72894-experiment-simulation-comparison.html)

mactech001 February 20, 2010 22:23

experiment and simulation comparison
 
Dear all,

I've been designing a cooling jacket for an electrical machine, and i'm now comparing experimental and simulation results from a fluid analysis point of view.

At the same inlet flowrate and the same outlet static pressure with respect to 1atm, the pressure difference between inlet and outlet from simulation is 30% higher than the experimental measurement [(simulation delta_p - experiment delta_p)/experiment delta_p]. No roughness was applied to simulation.

could the fluid properties such as viscosity, density account for this difference?

Any experience/comments shared will be very helpful. Thank you for your kind attention.

feizaghaee February 21, 2010 04:36

have you checked mesh sensitivity?

mactech001 February 21, 2010 05:27

RE: mesh sensitivity
 
Quote:

Originally Posted by feizaghaee (Post 246757)
have you checked mesh sensitivity?

Hi feizaghaee, thank you for your reply.

my cooling jacket has coolant channel of a helical configuration. should i concentrate more mesh on the corners on the fluid part please?

feizaghaee February 21, 2010 05:36

it depends on your geometery and other effective factors. you should refine the mesh until it dosen't chang anymore. i can't tell you it's enough or not!

mactech001 February 23, 2010 22:12

i've increased the mesh sizing of the fluid from 1.5e-3 to 0.8e-3
No significant difference in calculated pressure difference. Calculated pressure difference is still about 30% higher than measured pressure difference.
I've used WB mesh, not CFX-mesh..... will this bring any difference if i switched to CFX-mesh with inflation?

zandi February 25, 2010 08:19

Could you tell how you use WB mesh in cfx without changing it to cfx mesh?
I thought it's impossible!!
test all the thing you think.
but how about the turbulent modeling method?
I'm not profession in your field but one of other reasons that can affect result to be different from real data is turbulent method.
in literature survey we find methods suitable for our case or we find it ourself with testing them.

mactech001 March 1, 2010 03:48

Hi zandi, thank you for your reply and suggestions.

i've tried using Shear Stress Transport Model and the BSL Reynolds Stress model in the fluid Domain> Fluid Models page > Turbulence Option. Results using both options have made the calculated pressure difference closer to the experimental pressure difference.

A few more enquiries:
1) what is the difference between changing the turbulence model in the Domain and in the Boundary please?
2) how do i choose which turbulence model to use in the early stages of design when i don't have a prototype for experiments?

Hope to hear of any comments/suggestions again. Thank you!

zandi March 1, 2010 17:44

Hi
1)The turbulent model that we use in the domain is the main turbulent model, if Iím right. In the boundary due to conditions we choose T M.
For example if we use k-epsilon for domain because of abilities and our reasons then we could not use it in the boundary, why?! Because of not having k and epsilon values.
In the cfx guide read the topic for inlet (subsonic), part of turbulence:
ANSYS CFX-Solver Modeling Guide | Boundary Condition Modeling | Inlet |

2) you should have a literature survey. See which models have been applied before in your field problems. See the ability of each model and disabilities (in CFD texts, or CFX guide) and match them with the characteristics that your problem needs.
The SST Model have high abilities and accuracy and is a good choose, and itís the cfx powerful solution model instead of k-e and Ö, as the experts and cfx guide say. but itís better to read the CFX help in turbulent models first if you donít have any overview of T M.

mactech001 March 4, 2010 03:57

Is it true that, changing the turbulence model, will not have the slightest effect on the heat transfer from solid to fluid please?

ghorrocks March 4, 2010 06:43

Different turbulence models mean different heat transfer models. So if turbulent heat transfer is significant in your model then different turbulence models will result in different heat transfer.

mactech001 March 9, 2010 22:34

Hi all,

i'm facing another problem: i think one turbulence model for a certain flowrate doesn't work for another flowrate..... is it true???

My comparisons on inlet/outlet pressure difference between experiment and simulation results using BSL Reynolds Stress model have been very good at flowrate of 12L/min, but not for 8.4L/min and 16.6L/min.

any comments/suggestions will be appreciated.

ghorrocks March 10, 2010 07:23

It could be lots of things, including merely luck.

zandi March 10, 2010 13:02

Quote:

Originally Posted by mactech001 (Post 249247)
Hi all,

i'm facing another problem: i think one turbulence model for a certain flowrate doesn't work for another flowrate..... is it true???

I dont think so

Quote:

My comparisons on inlet/outlet pressure difference between experiment and simulation results using BSL Reynolds Stress model have been very good at flowrate of 12L/min, but not for 8.4L/min and 16.6L/min.
is your model convergent? in all flowrates?

mactech001 March 11, 2010 03:31

Yes, the solution all converges to RMS residual 1e-5


All times are GMT -4. The time now is 17:28.