radial fan; comparison measurement and calculation

Benfa · December 17, 2012, 02:08

Hello,

I am calculating a radial fan with the goal to match the measured pressure rise vs. mass flow curve as close as possible. Therefore I bulilt up a whole 3D model using CFX. The results are that the pressure rise in the design point is very close to the experimental (+- 5%). If I vary the pressure rise my calculated values with mass flow lower tha desig point are lower than the measurement(+-30%). If I vary mass flow higher than design point the pressure rise is to high (+-40%). I looked at several other works an it seems that it is a general trend that the points away from the design point does not fit very well. Does anybody have experience whats the reason for that and how to improve the pressure rise calculation.

I already studied the influence of mesh density, y+, turbulence and transient rotor stator calculation but the influence is not big enough.

ghorrocks · December 17, 2012, 04:37

That is normal - on design the flow is attached and CFD works well. Off design point the flow separates and accurate CFD is far more difficult.

Benfa · December 17, 2012, 05:34

What do you mean with far more difficult? Does it include the correct capturing of the separation process or anything else?
Does it mean that I need to resolve the sublayer (y+ <11) tp predict the correct pressure rise an if necesarry do a transient simulation?
I would appreciate if you could clerifiy my thoughts.

Thanks in advance!

ghorrocks · December 17, 2012, 05:40

Far more difficult = harder to converge and harder to get accurate.

The separtion process is integral to the performance of the device so difficulty in getting the separation right makes the whole thing difficult.

Resolution of the sublayer is only one factor. Different turbulence models handle separations with different degrees of finesse. SST is generally better than k-e, sometimes the turbulence transition model can help, but mainly separations require a DES/LES/SAS approach. These approaches are far more difficult to implement than normal 2-eqn models.

December 17, 2012, 02:08	radial fan; comparison measurement and calculation	#1
Benfa Member Benny Join Date: Apr 2012 Posts: 40 Rep Power: 14	Hello, I am calculating a radial fan with the goal to match the measured pressure rise vs. mass flow curve as close as possible. Therefore I bulilt up a whole 3D model using CFX. The results are that the pressure rise in the design point is very close to the experimental (+- 5%). If I vary the pressure rise my calculated values with mass flow lower tha desig point are lower than the measurement(+-30%). If I vary mass flow higher than design point the pressure rise is to high (+-40%). I looked at several other works an it seems that it is a general trend that the points away from the design point does not fit very well. Does anybody have experience whats the reason for that and how to improve the pressure rise calculation. I already studied the influence of mesh density, y+, turbulence and transient rotor stator calculation but the influence is not big enough.

December 17, 2012, 04:37		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,700 Rep Power: 143	That is normal - on design the flow is attached and CFD works well. Off design point the flow separates and accurate CFD is far more difficult.

December 17, 2012, 05:34		#3
Benfa Member Benny Join Date: Apr 2012 Posts: 40 Rep Power: 14	What do you mean with far more difficult? Does it include the correct capturing of the separation process or anything else? Does it mean that I need to resolve the sublayer (y+ <11) tp predict the correct pressure rise an if necesarry do a transient simulation? I would appreciate if you could clerifiy my thoughts. Thanks in advance!

December 17, 2012, 05:40		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,700 Rep Power: 143	Far more difficult = harder to converge and harder to get accurate. The separtion process is integral to the performance of the device so difficulty in getting the separation right makes the whole thing difficult. Resolution of the sublayer is only one factor. Different turbulence models handle separations with different degrees of finesse. SST is generally better than k-e, sometimes the turbulence transition model can help, but mainly separations require a DES/LES/SAS approach. These approaches are far more difficult to implement than normal 2-eqn models.