Please help! (incompressible navier-stokes)

taka · March 18, 2009, 08:22

Hi, i'm getting unconvincing flow reading from my solar tower model. The tower's air exit velocity is in the order of 10^5 when i'm expecting below 100m/s. I can't find where the source of the problem is, as all other aspects (the visualization and temperature distribution) are making sense.

Any of you has this kind of problem before, please comment on this thread. thx!

sbaffini · March 18, 2009, 08:55

If i understand well what a solar tower is(*), there could be a problem with the incompressible flow assumption. In fact you are not simulating a simple adiabathic flow but a very complex one in which different effects contribute to the divergence of the velocity field as the heating from the external source, the vertical size on which the motion takes place and the expected exit velocity which is at the upper limit of the incompressible approximation. All these things put together could well made the flow be far from the incompressible case.

Obviously i assume that everything else is fine, like the grid, boundary conditions, proper dimensions, etc.

(*) For solar tower i mean an hollow axialsymmetric building, a very tall one, in which the air, heated at the base by the solar heat, flows through the top exit where some kind of turbine (probably not present in your simulation) is placed. Is this right?

taka · March 18, 2009, 09:11

you are absolutely correct about how my solar tower model looks like. And it's a simple 2D model, which i initially hoped could give better results than in 3D, but turns out not

I've checked the boundary conditions, buoyancy formula and input correct air properties value as well as using Lagrange quadratic in the subdomain setting. couldn't find where's wrong. If you don't mind, I wish you could do me a favor by checking the boundaries again :P

sbaffini · March 18, 2009, 13:26

Actually, if you are using the RANS approach, the correct model is the axialsymmetric one (Fluent has this kind of model) and there is no reason to switch to a 3D model.

I don't actually understand what you mean for "checking the boundaries" but, considering that everything seems fine (even if you MUST switch to the axialsymmetric model, the 2D approach it's not unreasonable and usually just gives you wrong results with the same order of magnitude of the right ones), i think that you need to switch to the compressible solver for the reasons that i stated before.

taka · March 18, 2009, 20:29

the boundary conditions are the elements that i specified before solving the model. for example no-slip conditions at the wall and convective flux at the tower's air exit. anyway, i might wanna try out your suggested solution to switch to compressible solver, which i currently use incompressible (i just assumed it's incompressible cuz the 100metre-high tower wont make much difference in the air compressibility property)

Thanks sbaffini

sbaffini · March 19, 2009, 04:38

I'm not sure of what i'd pick at the inflow boundary because fixing the mass inflow will obviously fix the mass outflow which i think is something you want to compute and not fix. Probably, having fixed the inflow (if this is your case) and selected the incompressible flow has resulted in most of the heat converted in kinetic energy, which gave you the wrong result.

I think that with a compressible flow approach a pressure-inlet/pressure-outlet set of boundary conditions is more suitable, but i'm not expert of internal compressible flows so you should probably check somewhere if this is the right choice.

The issue about the compressibility of this kind of flow is not that simple.

Relevant parameters in this case are $M^2 , \frac{M^2}{Fr} , \frac{\beta}{Re Pr}$ , with $Fr = \frac{L g}{U^2}$ and L being a vertical lenght scale on which the velocity varies of U.

Even if each of these parameters is not high enough, their are not definitely low so their interplay could not be that obvious.

harishg · March 19, 2009, 10:05

What numerical method do you employ ? What BC do you apply ? Which solver do you employ ? Does it include RANS modeling? A solution of the order of 10^5 suggests that your solution has exploded. Check the implementation of your boundary conditions. If you employed central differencing switch to upwind and see if it works.

taka · March 19, 2009, 19:43

Hey guys thanks for the many help! harishg, i'm using the lagrange quadratic method and linear solver. the inflow is not fixed too as this is what i wanted to find. Up until now,i did not realize there's a mistake in my navier stokes equation. I was suppose to input a formula written as Fy = beta*gravity*density*volume*(T-293)

where beta = coefficient of heat expansion of air
T = temperature

however since my model is 2D, i wasnt aware of the volume cuz i do not have z-axis!!! i never thought of this before, until i decided to check the equation again...sigh..anyone who can tell me how do i input a volume into a 2-D model?

dominic · March 19, 2009, 23:25

I suppose, you can replace volume by area for two-dimensions... I assume that the volume you have in your expression is the volume of your solar tower..

taka · March 20, 2009, 03:12

Hi there's one more thing that's making me gone crazy at the moment. I suspect that the X and Y coordinates in FEMLAB are not exactly showing the SI dimension (metre) which I believe is the reason why i'm getting blown-up velocity value. For instance, I drew my model at 800metre long, (i followed the '800' on the x-axis), the resulting velocity output is beyond supersonic! then i changed the length to 8metre, then velocity becomes 30m/s, which is way making more sense.
I suspect that the 'number' at both the axes are factored or scaled-up representation, but I could not find anywhere in the help files that could verify this. FEMLAB only claim that the dimension is in SI units by default, but there must be something wrong with the axes dimension.
Help anyone? thanks

sbaffini · March 20, 2009, 03:25

Quote:

Originally Posted by sbaffini

Obviously i assume that everything else is fine, like the grid, boundary conditions, proper dimensions, etc.

Checking the dimensions is the first thing to do when there is such difference between the result and the expected one.

taka · March 20, 2009, 04:33

yes sbaffini, the dimension is the main problem right now, but the problem is, i dont know what unit does the scale means. there are only numbers, so i can't be sure if '10' means '10 meters' and there's no way for me to find out. Things are not gonna change even though i draw the model in scale, but the results depends on the scale factor that FEMLAB has here. So what i need now is to verify the scale factor of the dimensions and hopefully, using that to scale down the blown-up velocity results.

March 18, 2009, 08:22	Please help! (incompressible navier-stokes)	#1
taka New Member LingTu Join Date: Mar 2009 Posts: 6 Rep Power: 17	Hi, i'm getting unconvincing flow reading from my solar tower model. The tower's air exit velocity is in the order of 10^5 when i'm expecting below 100m/s. I can't find where the source of the problem is, as all other aspects (the visualization and temperature distribution) are making sense. Any of you has this kind of problem before, please comment on this thread. thx!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Coding 2d incompressible Navier Stokes	Simon	Main CFD Forum	22	December 1, 2020 04:01
Steady Incompressible Navier Stokes	selim	CFX	0	October 24, 2007 07:52
navier-stokes	queram	Main CFD Forum	7	February 25, 2005 07:20
Incompressible Navier stokes with source term	mayur	FLUENT	0	August 6, 2004 06:07
2D incompressible Navier-Stokes solver for diffuser analysis of centrifugal pump WANTED !!!	LI, J	Main CFD Forum	1	November 19, 2003 23:49

March 18, 2009, 08:55		#2
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,151 Blog Entries: 29 Rep Power: 39	If i understand well what a solar tower is(), there could be a problem with the incompressible flow assumption. In fact you are not simulating a simple adiabathic flow but a very complex one in which different effects contribute to the divergence of the velocity field as the heating from the external source, the vertical size on which the motion takes place and the expected exit velocity which is at the upper limit of the incompressible approximation. All these things put together could well made the flow be far from the incompressible case. Obviously i assume that everything else is fine, like the grid, boundary conditions, proper dimensions, etc. () For solar tower i mean an hollow axialsymmetric building, a very tall one, in which the air, heated at the base by the solar heat, flows through the top exit where some kind of turbine (probably not present in your simulation) is placed. Is this right?

March 18, 2009, 09:11		#3
taka New Member LingTu Join Date: Mar 2009 Posts: 6 Rep Power: 17	you are absolutely correct about how my solar tower model looks like. And it's a simple 2D model, which i initially hoped could give better results than in 3D, but turns out not I've checked the boundary conditions, buoyancy formula and input correct air properties value as well as using Lagrange quadratic in the subdomain setting. couldn't find where's wrong. If you don't mind, I wish you could do me a favor by checking the boundaries again :P

March 18, 2009, 13:26		#4
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,151 Blog Entries: 29 Rep Power: 39	Actually, if you are using the RANS approach, the correct model is the axialsymmetric one (Fluent has this kind of model) and there is no reason to switch to a 3D model. I don't actually understand what you mean for "checking the boundaries" but, considering that everything seems fine (even if you MUST switch to the axialsymmetric model, the 2D approach it's not unreasonable and usually just gives you wrong results with the same order of magnitude of the right ones), i think that you need to switch to the compressible solver for the reasons that i stated before.

March 18, 2009, 20:29		#5
taka New Member LingTu Join Date: Mar 2009 Posts: 6 Rep Power: 17	the boundary conditions are the elements that i specified before solving the model. for example no-slip conditions at the wall and convective flux at the tower's air exit. anyway, i might wanna try out your suggested solution to switch to compressible solver, which i currently use incompressible (i just assumed it's incompressible cuz the 100metre-high tower wont make much difference in the air compressibility property) Thanks sbaffini

March 19, 2009, 04:38		#6
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,151 Blog Entries: 29 Rep Power: 39	I'm not sure of what i'd pick at the inflow boundary because fixing the mass inflow will obviously fix the mass outflow which i think is something you want to compute and not fix. Probably, having fixed the inflow (if this is your case) and selected the incompressible flow has resulted in most of the heat converted in kinetic energy, which gave you the wrong result. I think that with a compressible flow approach a pressure-inlet/pressure-outlet set of boundary conditions is more suitable, but i'm not expert of internal compressible flows so you should probably check somewhere if this is the right choice. The issue about the compressibility of this kind of flow is not that simple. Relevant parameters in this case are $M^2 , \frac{M^2}{Fr} , \frac{\beta}{Re Pr}$ , with $Fr = \frac{L g}{U^2}$ and L being a vertical lenght scale on which the velocity varies of U. Even if each of these parameters is not high enough, their are not definitely low so their interplay could not be that obvious.

March 19, 2009, 10:05		#7
harishg Senior Member N/A Join Date: Mar 2009 Posts: 189 Rep Power: 17	What numerical method do you employ ? What BC do you apply ? Which solver do you employ ? Does it include RANS modeling? A solution of the order of 10^5 suggests that your solution has exploded. Check the implementation of your boundary conditions. If you employed central differencing switch to upwind and see if it works.

March 19, 2009, 19:43		#8
taka New Member LingTu Join Date: Mar 2009 Posts: 6 Rep Power: 17	Hey guys thanks for the many help! harishg, i'm using the lagrange quadratic method and linear solver. the inflow is not fixed too as this is what i wanted to find. Up until now,i did not realize there's a mistake in my navier stokes equation. I was suppose to input a formula written as Fy = betagravitydensityvolume(T-293) where beta = coefficient of heat expansion of air T = temperature however since my model is 2D, i wasnt aware of the volume cuz i do not have z-axis!!! i never thought of this before, until i decided to check the equation again...sigh..anyone who can tell me how do i input a volume into a 2-D model?

March 19, 2009, 23:25		#9
dominic Member Dominic Chandar Join Date: Mar 2009 Location: United Kingdom Posts: 31 Rep Power: 17	I suppose, you can replace volume by area for two-dimensions... I assume that the volume you have in your expression is the volume of your solar tower..

March 20, 2009, 03:12		#10
taka New Member LingTu Join Date: Mar 2009 Posts: 6 Rep Power: 17	Hi there's one more thing that's making me gone crazy at the moment. I suspect that the X and Y coordinates in FEMLAB are not exactly showing the SI dimension (metre) which I believe is the reason why i'm getting blown-up velocity value. For instance, I drew my model at 800metre long, (i followed the '800' on the x-axis), the resulting velocity output is beyond supersonic! then i changed the length to 8metre, then velocity becomes 30m/s, which is way making more sense. I suspect that the 'number' at both the axes are factored or scaled-up representation, but I could not find anywhere in the help files that could verify this. FEMLAB only claim that the dimension is in SI units by default, but there must be something wrong with the axes dimension. Help anyone? thanks

March 20, 2009, 04:33		#12
taka New Member LingTu Join Date: Mar 2009 Posts: 6 Rep Power: 17	yes sbaffini, the dimension is the main problem right now, but the problem is, i dont know what unit does the scale means. there are only numbers, so i can't be sure if '10' means '10 meters' and there's no way for me to find out. Things are not gonna change even though i draw the model in scale, but the results depends on the scale factor that FEMLAB has here. So what i need now is to verify the scale factor of the dimensions and hopefully, using that to scale down the blown-up velocity results.