CFX transition model issue with VATT

PLD · April 17, 2014, 11:40

Hi everyone,

I have a problem with 2D calculations of a straight bladed vertical axis tidal turbine using the Gamma - theta transition model of CFX.The mesh is divided in 3 parts : a rotating ring containing the 3 blades and 2 stators inside and outside the ring.
I have run calculations at several tip speed ratios (TSR) with the fully turbulent SST turbulence model (called SST henceforth) and the SST turbulence model coupled to the gamma - theta transition model (called SST-TM henceforth). I expected differences between the results of the two models but I get big unrealistic differences at high TSR as you can see on the attached picture (averaged Power Coefficients (CP) have been divided by the SST value at TSR = 3 since they have not been published yet).

SST calculations seem fairly good since they agree with data obtained by Paraschivoiu at the same solidity. Howerver, I don't understand why the CP keeps increasing after TSR = 3 in the SST-TM calculations. The value at TSR = 6 is even above the Betz limit...

I have already checked the convergence of the solution (grid, time step, size of the calculation domain). I use a maximum y+ value of 1 on the blades. I have run calculations on a static hydrofoil with both SST and SST-TM models and they were in very good agreement with experimental data. The turbulence intensity is very low just before the turbine (0.1%) but I don't see how it could have such a big effect on the CP.

I have read in the paper of Menter (AIAA Journal, 2009) that one of the limitations of the SST-TM model is that "the transition correlations are formulated non-Galilean invariant". I don't know if it has been corrected in the version I use (CFX v14.0) and this point is not mentioned in the modeling guide or anywhere else in the CFX documentation. If my understanding is good, this should only affect the turbulence intensity calculated as an input of the transition model. I don't understand how it could result in a factor 2 for the CP calculated at TSR = 6.

Do you have any idea regarding this problem ?
Thanks for your help!

PLD · April 17, 2014, 11:58

Sorry. Here is the attached picture:

CP_adim_SST_et_SST_TM.png

ghorrocks · April 17, 2014, 22:05

Is the model tightly converged?

You would have to look at the detail of the flow over the blades. There is probably something weird going on - maybe the flow separates in one model and does not in the other, maybe the laminar separation bubble of the transition model is doing weird things.

And are you modelling it steady state or transient, or something like frozen rotor?

PLD · April 19, 2014, 18:09

Thanks for your answer Glenn!

I am modeling it transient. Convergence is good: residuals (rms) are under 10^(-5) and the calculation is run during 6 to 10 turbine revolutions so that the power coefficient reaches a constant value at the end of the calculation.

A plot of the pressure coefficient shows that the transition model leads to a slightly higher lift and a plot of the friction coefficient shows that the viscous drag is divided by 2 with the transition model. The lift increase and the drag decrease both tend to an increase of the driving force, so the increase of the power coefficient is rather logical. However such a big increase is weird I think...

ghorrocks · April 20, 2014, 06:30

What Re is this running at?

Most wind turbines run fairly low Re numbers, resulting in significant laminar flow regions before transition to turbulence. This means a transition model is to be expected to result in significantly lower drag than a full turbulence model. So your report that the transition model is lower drag is expected, and the drag reduction could be significant (like 50%). This probably explains the difference you are seeing.

PLD · April 22, 2014, 05:21

In my case, at TSR = 6 the Reynolds (based on the chord and the relative velocity) varies between 1.5 x 10^6 and 2.1 x 10^6 during one revolution.
In the calculation, transition generally takes place at x/c = 0.3 on the suction side and very close to the trailing edge on the pressure side, which explains the decrease of the viscous drag as you say. However, transition location should probably be located upstream of these points, I think, and this behavior may be due to the low tubulence intensity (0.1%). I am currently running the same case with a higher turbulence intensity to assess its effect on the power coefficient.

April 17, 2014, 11:40	CFX transition model issue with VATT	#1
PLD New Member Pierre-Luc Join Date: Oct 2011 Location: UK Posts: 16 Rep Power: 14	Hi everyone, I have a problem with 2D calculations of a straight bladed vertical axis tidal turbine using the Gamma - theta transition model of CFX.The mesh is divided in 3 parts : a rotating ring containing the 3 blades and 2 stators inside and outside the ring. I have run calculations at several tip speed ratios (TSR) with the fully turbulent SST turbulence model (called SST henceforth) and the SST turbulence model coupled to the gamma - theta transition model (called SST-TM henceforth). I expected differences between the results of the two models but I get big unrealistic differences at high TSR as you can see on the attached picture (averaged Power Coefficients (CP) have been divided by the SST value at TSR = 3 since they have not been published yet). SST calculations seem fairly good since they agree with data obtained by Paraschivoiu at the same solidity. Howerver, I don't understand why the CP keeps increasing after TSR = 3 in the SST-TM calculations. The value at TSR = 6 is even above the Betz limit... I have already checked the convergence of the solution (grid, time step, size of the calculation domain). I use a maximum y+ value of 1 on the blades. I have run calculations on a static hydrofoil with both SST and SST-TM models and they were in very good agreement with experimental data. The turbulence intensity is very low just before the turbine (0.1%) but I don't see how it could have such a big effect on the CP. I have read in the paper of Menter (AIAA Journal, 2009) that one of the limitations of the SST-TM model is that "the transition correlations are formulated non-Galilean invariant". I don't know if it has been corrected in the version I use (CFX v14.0) and this point is not mentioned in the modeling guide or anywhere else in the CFX documentation. If my understanding is good, this should only affect the turbulence intensity calculated as an input of the transition model. I don't understand how it could result in a factor 2 for the CP calculated at TSR = 6. Do you have any idea regarding this problem ? Thanks for your help!

April 20, 2014, 06:30		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	What Re is this running at? Most wind turbines run fairly low Re numbers, resulting in significant laminar flow regions before transition to turbulence. This means a transition model is to be expected to result in significantly lower drag than a full turbulence model. So your report that the transition model is lower drag is expected, and the drag reduction could be significant (like 50%). This probably explains the difference you are seeing. zkdkeen likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
CFX multiphase flow model	aximefu	CFX	24	February 17, 2018 05:35
Superlinear speedup in OpenFOAM 13	msrinath80	OpenFOAM Running, Solving & CFD	18	March 3, 2015 05:36
How to use non-equilibrium thermal model for porous medium in Ansys CFX 13.0?	Chander	CFX	3	November 28, 2011 14:26
modelling transition with v2f turbulence model	pablo	FLUENT	0	August 7, 2006 15:44
Reynolds Stress Model in Fluent Vs CFX	Tim	FLUENT	0	December 6, 2005 22:03

April 17, 2014, 11:58		#2
PLD New Member Pierre-Luc Join Date: Oct 2011 Location: UK Posts: 16 Rep Power: 14	Sorry. Here is the attached picture: CP_adim_SST_et_SST_TM.png

April 17, 2014, 22:05		#3
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	Is the model tightly converged? You would have to look at the detail of the flow over the blades. There is probably something weird going on - maybe the flow separates in one model and does not in the other, maybe the laminar separation bubble of the transition model is doing weird things. And are you modelling it steady state or transient, or something like frozen rotor?

April 19, 2014, 18:09		#4
PLD New Member Pierre-Luc Join Date: Oct 2011 Location: UK Posts: 16 Rep Power: 14	Thanks for your answer Glenn! I am modeling it transient. Convergence is good: residuals (rms) are under 10^(-5) and the calculation is run during 6 to 10 turbine revolutions so that the power coefficient reaches a constant value at the end of the calculation. A plot of the pressure coefficient shows that the transition model leads to a slightly higher lift and a plot of the friction coefficient shows that the viscous drag is divided by 2 with the transition model. The lift increase and the drag decrease both tend to an increase of the driving force, so the increase of the power coefficient is rather logical. However such a big increase is weird I think...

April 22, 2014, 05:21		#6
PLD New Member Pierre-Luc Join Date: Oct 2011 Location: UK Posts: 16 Rep Power: 14	In my case, at TSR = 6 the Reynolds (based on the chord and the relative velocity) varies between 1.5 x 10^6 and 2.1 x 10^6 during one revolution. In the calculation, transition generally takes place at x/c = 0.3 on the suction side and very close to the trailing edge on the pressure side, which explains the decrease of the viscous drag as you say. However, transition location should probably be located upstream of these points, I think, and this behavior may be due to the low tubulence intensity (0.1%). I am currently running the same case with a higher turbulence intensity to assess its effect on the power coefficient.