Temperature issue for walls

Hr_kules · February 15, 2023, 08:44

Hi,

i am simulating a pipe to compute a Nusselt Number and check the results with The Dittus Boelter equation.
My simulation by itself runs pretty good, i got a mesh with a good y+ value and the energy balance in total checks out. However to compute the heat transfer coeffcient i am lacking the desired temperature difference.
Calculation by hand suggests a temperature difference of ~0.5K between the wall and the fluid bulk.
My simulation doesn't match this at all, but still matches the temperature difference between in and outlet and the prescribed wall heat flux of 1000w/m^2 on the entry and the assumed developed region. So Q = mdot cp Delta matches. But h = q/(t_wall - t_fluid) doesn't match.

What could be the reason?

Hr_kules · February 16, 2023, 06:37

So far, ich have tried everything i could think of to resolve the problem.
I am using OpenFoam Version 9 and buoyantSimpleFoam. Could changing the solve resolve the issue? My working fluid is water and right now i am confussed by the statements of different documentations.

https://doc.cfd.direct/openfoam/user...rs#dx13-103005
states that this solver is for compressible flows, one of the tutorials uses water. If i choose the wrong solver, what would be the correct one for a simulation of convection in a pipe flow with a turbulent, incompressible flow?

Hr_kules · February 17, 2023, 09:19

Is really nobody able to help?

Tobermory · February 18, 2023, 07:37

The fact that you have the correct overall deltaT for the water stream (inlet to outlet) means that you are matching the heat transfer from the wall correctly, so that's a good start. The problem then is how to compare your calculated (HTC, T) values against the expected (empirical) values. Clearly, the HTC value is VERY dependent on the T value you use.

My question to you is: how are you calculating you HTC in OpenFOAM, and what T is it using? Is it the local HTC, based on the temperature of the cell adjacent to the cell? If so, then this will clearly be different from the empirical bulk HTC.

Hr_kules · February 18, 2023, 08:03

Hey!

The HTC is computed from the area average Temperature of the wall, where i assume that the flow is developed flow ≈10*Pipe diameter) and the bulk temperature of the Fluid that is the mass flow average temperature of the flow.
Yesterday i investigated the temperature profile on the wall and found it to be of a near parabolic shape, it could be that the flow isn't developed at all and the Initial assumption Was wrong. Maybe my boundary layer is not Dimensioned right since the boundary layer length i used was the length of the whole pipe, and not the length Till flow is developed. This would at least Match openfoams yplus functionobject that prints that my yplus is below 1 and not 1 like is used to compute the first layer height.

Why is this important? The empirical correlation is only valid for developed pipe flow.

Bloerb · February 20, 2023, 17:26

You might want to look at the so called "Graetz Problem" which gives you an equation for the temperature profile in a pipe, either heated with a constant heat flux or constant temperature and laminar flow. For the undeveloped as well as developed region of the temperature profile.

For a pipe you should find equations or measurement data for the bulk temperature along the length for several Reynolds numbers as well. This might make it easier to validate as Nu numbers or htcs or often very depended on how they are expressed and made non dimensional. And this varies a lot in the literature.

Hr_kules · February 21, 2023, 04:35

I was able to make some good changes to the case and i am now able produce results.
The emprical correlation i use is the dittus boelter equation (Nu = 0.023*Re^0.8*Pr^0.4) withmy calculated heat transfer coeffcient i was able to compute a Nusselt number that matches the dittus boelter equation with a difference of about 1%.

The biggest change was the usage of the k-omega SST Model. I saw in some old notes of my CFD lecture that this was a good model to investigate turbulent flow with heat transfer effects.

There are some things in my simulation especially regarding the y+ values and boundary conditions that i need to look further into, however these should be put in a seperate threat.

February 15, 2023, 08:44	Temperature issue for walls	#1
Hr_kules Senior Member Nico Join Date: Jan 2022 Location: Germany Posts: 122 Rep Power: 6	Hi, i am simulating a pipe to compute a Nusselt Number and check the results with The Dittus Boelter equation. My simulation by itself runs pretty good, i got a mesh with a good y+ value and the energy balance in total checks out. However to compute the heat transfer coeffcient i am lacking the desired temperature difference. Calculation by hand suggests a temperature difference of ~0.5K between the wall and the fluid bulk. My simulation doesn't match this at all, but still matches the temperature difference between in and outlet and the prescribed wall heat flux of 1000w/m^2 on the entry and the assumed developed region. So Q = mdot cp Delta matches. But h = q/(t_wall - t_fluid) doesn't match. What could be the reason?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Weird (unphysical) temperature rise with chtMultiRegionFoam	Phil910	OpenFOAM Running, Solving & CFD	3	November 9, 2022 10:25
Averaging over iterations for steady-state simulation	CFD student	Fluent UDF and Scheme Programming	8	September 22, 2022 04:39
Problem with zeroGradient wall BC for temperature - Total temperature loss	cboss	OpenFOAM	12	October 1, 2018 07:36
unexpected constant Temperature on a clip surface	Sungki	OpenFOAM Running, Solving & CFD	0	August 4, 2015 05:50
Temperature Issue in OpenFOAM-2.2.0	prasant	OpenFOAM	0	March 12, 2013 08:17

February 16, 2023, 06:37		#2
Hr_kules Senior Member Nico Join Date: Jan 2022 Location: Germany Posts: 122 Rep Power: 6	So far, ich have tried everything i could think of to resolve the problem. I am using OpenFoam Version 9 and buoyantSimpleFoam. Could changing the solve resolve the issue? My working fluid is water and right now i am confussed by the statements of different documentations. https://doc.cfd.direct/openfoam/user...rs#dx13-103005 states that this solver is for compressible flows, one of the tutorials uses water. If i choose the wrong solver, what would be the correct one for a simulation of convection in a pipe flow with a turbulent, incompressible flow?

February 17, 2023, 09:19		#3
Hr_kules Senior Member Nico Join Date: Jan 2022 Location: Germany Posts: 122 Rep Power: 6	Is really nobody able to help?

February 18, 2023, 07:37		#4
Tobermory Senior Member Join Date: Apr 2020 Location: UK Posts: 747 Rep Power: 14	The fact that you have the correct overall deltaT for the water stream (inlet to outlet) means that you are matching the heat transfer from the wall correctly, so that's a good start. The problem then is how to compare your calculated (HTC, T) values against the expected (empirical) values. Clearly, the HTC value is VERY dependent on the T value you use. My question to you is: how are you calculating you HTC in OpenFOAM, and what T is it using? Is it the local HTC, based on the temperature of the cell adjacent to the cell? If so, then this will clearly be different from the empirical bulk HTC.

February 18, 2023, 08:03		#5
Hr_kules Senior Member Nico Join Date: Jan 2022 Location: Germany Posts: 122 Rep Power: 6	Hey! The HTC is computed from the area average Temperature of the wall, where i assume that the flow is developed flow ≈10*Pipe diameter) and the bulk temperature of the Fluid that is the mass flow average temperature of the flow. Yesterday i investigated the temperature profile on the wall and found it to be of a near parabolic shape, it could be that the flow isn't developed at all and the Initial assumption Was wrong. Maybe my boundary layer is not Dimensioned right since the boundary layer length i used was the length of the whole pipe, and not the length Till flow is developed. This would at least Match openfoams yplus functionobject that prints that my yplus is below 1 and not 1 like is used to compute the first layer height. Why is this important? The empirical correlation is only valid for developed pipe flow.

February 20, 2023, 17:26		#6
Bloerb Senior Member Join Date: Sep 2013 Posts: 353 Rep Power: 21	You might want to look at the so called "Graetz Problem" which gives you an equation for the temperature profile in a pipe, either heated with a constant heat flux or constant temperature and laminar flow. For the undeveloped as well as developed region of the temperature profile. For a pipe you should find equations or measurement data for the bulk temperature along the length for several Reynolds numbers as well. This might make it easier to validate as Nu numbers or htcs or often very depended on how they are expressed and made non dimensional. And this varies a lot in the literature.

February 21, 2023, 04:35		#7
Hr_kules Senior Member Nico Join Date: Jan 2022 Location: Germany Posts: 122 Rep Power: 6	I was able to make some good changes to the case and i am now able produce results. The emprical correlation i use is the dittus boelter equation (Nu = 0.023Re^0.8Pr^0.4) withmy calculated heat transfer coeffcient i was able to compute a Nusselt number that matches the dittus boelter equation with a difference of about 1%. The biggest change was the usage of the k-omega SST Model. I saw in some old notes of my CFD lecture that this was a good model to investigate turbulent flow with heat transfer effects. There are some things in my simulation especially regarding the y+ values and boundary conditions that i need to look further into, however these should be put in a seperate threat.