How to achieve homogeneous vertical profile for turbulence

saugatshr4 · August 16, 2021, 13:52

Hello,

I am working on turbulence modeling of the Atmospheric Boundary Layer. Here I have been able to maintain a homogeneous wind velocity profile. However, I have been unable to maintain a homogeneous profile for tubrulence.

I have given a manual input for the Turbulent Kinetic energy (TKE). It seems that the TKE seems to dissipate very early in the domain and doesn't travel the domain. I have attached a picture of the TKE in the along wind direction of the domain.

FYI, I'm trying to implement a k-epsilon model.

Can anybody please help me. I have been stuck in this problem for a long time.

I will be very grateful for your help.

Thank You.

With best regards,
Saugat

piu58 · August 17, 2021, 01:13

If you are interested in what happens in the boundary layer, the k-omega model is much better. You may refine the mesh in the boundary layer as far as you want. k-epsilon requires a coarse mesh here.

Nevertheless,, turbulence is only modelled with the k-epsilon and k-omega models. The all in all behaviour should be reproduced (forces and main velocity). If you want to simulate the turbulence in the boundary layer I recommend the LES model. You have to go for 3D for this, at least to 2.5D

Tobermory · August 17, 2021, 05:53

Good answer Uwe, but I fear the problem is deeper than that.

Saugat - the first thing to realise is that the constant stress layer concept for the ABL is an asymptotic approximation to the real world. It's a simple model that allowed early ABL modellers to approximate conditions in the surface layer of the ABL. CFD modellers in the 90s (kicked off by Richards & Hoxey) seem to have latched on to this and assume that it is the "correct" behaviour in the ABL, and have worked hard to try and achieve constant vertical profiles of k. In reality the ABL is just another boundary layer, and has a profile that of course should vary with height.

A shorter answer: the reason why you cannot preserve your vertical profile of k is that the vertical profile does not give a balance between production, dissipation, convection and diffusion of k. I.e. it is not the correct solution for the boundary layer conditions that you have.

You might be able to get better bahaviour if you check carefully that your wall boundary condition is in equilibrium with the inlet profiles of U, k, eps & nut, and if you check your boundary condition on the top of the domain. But keep in mind - the "constant k" profile is not a correct profile, especially if the domain height is more than 50m or so (k will decay with height, towards the edge of the ABL).

saugatshr4 · August 17, 2021, 08:57

Thank you Tobermory and piu58 for the response.

piu58:
For now, I do not have the computational power to run an LES simulation. My intention is to use generic CFD models used in Atmospheric Boundary Layer simulations. I will definitely try the k-omega model.

Tobermory:
I am fairly new to CFD, so this is more like a learning experience. So please forgive me if my question comes across to be a bit stupid.
From what I understand from your response is that my model is is not correct at the moment due to the inability to conserve the turbulent kinetic energy.
You suggested checking the equilibrium of U, k, and eps and nut. Can you recommend methods, how this can be implemented.
I got a recommendation to use a cyclic boundary layer to conserve k. Do you think this would be the right approach?
I really want my model to be correct and give me correct results. Your help in clearing these doubts would be invaluable.

Thank You.

With best regards,
Saugat

piu58 · August 17, 2021, 12:25

Cyclic boundaries are often problematic. I would avoid it.

I see two ways:

1) pseudo-cyclic: mesh a geometry which contains your case three of four times after each other and use the results form the second to last
2) (if you get a non oscillating result after some time): Use the k values at some distance form the input for new input setting. repeat that if necessary.

saugatshr4 · August 18, 2021, 10:02

Thanks, piu58 for the response.
I can try the second option and will let you know how it goes.
Regarding the first option, I am not quite clear. Can you please give a more detailed explanation of this?

I hope you will have time to consider my request.

With best regards,

Saugat

piu58 · August 18, 2021, 12:23

I made a sketch of the pseudo-cyclic geometry.

saugatshr4 · September 17, 2021, 00:15

Thank You piu58.
I learned a lot from your response.
I solved the problem by using the correct wall properties.
It really felt nice to receive help when I was completely stuck with my problems.
Thank You.
Regards,
Saugat

August 17, 2021, 01:13		#2
piu58 Senior Member Uwe Pilz Join Date: Feb 2017 Location: Leipzig, Germany Posts: 744 Rep Power: 15	If you are interested in what happens in the boundary layer, the k-omega model is much better. You may refine the mesh in the boundary layer as far as you want. k-epsilon requires a coarse mesh here. Nevertheless,, turbulence is only modelled with the k-epsilon and k-omega models. The all in all behaviour should be reproduced (forces and main velocity). If you want to simulate the turbulence in the boundary layer I recommend the LES model. You have to go for 3D for this, at least to 2.5D Fouch likes this. __________________ Uwe Pilz -- Die der Hauptbewegung überlagerte Schwankungsbewegung ist in ihren Einzelheiten so hoffnungslos kompliziert, daß ihre theoretische Berechnung aussichtslos erscheint. (Hermann Schlichting, 1950)

August 17, 2021, 12:25		#5
piu58 Senior Member Uwe Pilz Join Date: Feb 2017 Location: Leipzig, Germany Posts: 744 Rep Power: 15	Cyclic boundaries are often problematic. I would avoid it. I see two ways: 1) pseudo-cyclic: mesh a geometry which contains your case three of four times after each other and use the results form the second to last 2) (if you get a non oscillating result after some time): Use the k values at some distance form the input for new input setting. repeat that if necessary. __________________ Uwe Pilz -- Die der Hauptbewegung überlagerte Schwankungsbewegung ist in ihren Einzelheiten so hoffnungslos kompliziert, daß ihre theoretische Berechnung aussichtslos erscheint. (Hermann Schlichting, 1950)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
how to achieve homogeneous vertical profile for turbulence	saugatshr4	OpenFOAM	9	January 24, 2022 04:44
3D UDF Paraboilc Velocity Profile (Can't Maintain)	Sing	FLUENT	12	August 7, 2017 06:25
problems orientating profile	Ralf Schmidt	FLUENT	1	March 30, 2014 10:00
problem with importing and exporting profile BC	Gui	CFX	2	July 26, 2007 08:50
Associating profile files for the UDS though a UDF	Bharath	FLUENT	0	December 1, 2006 15:58

August 17, 2021, 05:53		#3
Tobermory Senior Member Join Date: Apr 2020 Location: UK Posts: 668 Rep Power: 14	Good answer Uwe, but I fear the problem is deeper than that. Saugat - the first thing to realise is that the constant stress layer concept for the ABL is an asymptotic approximation to the real world. It's a simple model that allowed early ABL modellers to approximate conditions in the surface layer of the ABL. CFD modellers in the 90s (kicked off by Richards & Hoxey) seem to have latched on to this and assume that it is the "correct" behaviour in the ABL, and have worked hard to try and achieve constant vertical profiles of k. In reality the ABL is just another boundary layer, and has a profile that of course should vary with height. A shorter answer: the reason why you cannot preserve your vertical profile of k is that the vertical profile does not give a balance between production, dissipation, convection and diffusion of k. I.e. it is not the correct solution for the boundary layer conditions that you have. You might be able to get better bahaviour if you check carefully that your wall boundary condition is in equilibrium with the inlet profiles of U, k, eps & nut, and if you check your boundary condition on the top of the domain. But keep in mind - the "constant k" profile is not a correct profile, especially if the domain height is more than 50m or so (k will decay with height, towards the edge of the ABL).

August 17, 2021, 08:57		#4
saugatshr4 New Member Saugat Shrestha Join Date: Dec 2019 Location: Thailand Posts: 27 Rep Power: 6	Thank you Tobermory and piu58 for the response. piu58: For now, I do not have the computational power to run an LES simulation. My intention is to use generic CFD models used in Atmospheric Boundary Layer simulations. I will definitely try the k-omega model. Tobermory: I am fairly new to CFD, so this is more like a learning experience. So please forgive me if my question comes across to be a bit stupid. From what I understand from your response is that my model is is not correct at the moment due to the inability to conserve the turbulent kinetic energy. You suggested checking the equilibrium of U, k, and eps and nut. Can you recommend methods, how this can be implemented. I got a recommendation to use a cyclic boundary layer to conserve k. Do you think this would be the right approach? I really want my model to be correct and give me correct results. Your help in clearing these doubts would be invaluable. Thank You. With best regards, Saugat

August 18, 2021, 10:02		#6
saugatshr4 New Member Saugat Shrestha Join Date: Dec 2019 Location: Thailand Posts: 27 Rep Power: 6	Thanks, piu58 for the response. I can try the second option and will let you know how it goes. Regarding the first option, I am not quite clear. Can you please give a more detailed explanation of this? I hope you will have time to consider my request. With best regards, Saugat

September 17, 2021, 00:15		#8
saugatshr4 New Member Saugat Shrestha Join Date: Dec 2019 Location: Thailand Posts: 27 Rep Power: 6	Thank You piu58. I learned a lot from your response. I solved the problem by using the correct wall properties. It really felt nice to receive help when I was completely stuck with my problems. Thank You. Regards, Saugat