Discussion Forum

Hi

Well I usually integrate over the inlet/outlets respectively "nitf.rho*nitf.U" (adding the *2*pi*r for 2D axi)
and to get the difference I use "nitf.rho*nitf.U*(-1)^(dom==3)" where "3" is in my last model the inlet boundary "id"

For the energy balance it's more model dependent I sometimes integrate over the external boundaries the nitf.tfluxMag = Total heat flux magnitude or nitf.tefluxMag = Total energy flux magnitude and check that it corresponds more or less to the nitf.Qtot = Total Heat Source (if applicable)

--
Good luck
Ivar

Dear Ivar,

Thank you very much for your answer.

I am attaching a simple model of conjugate heat transfer with the steady state solution. The model is of a copper channel with air flowing inside.

My heat input is of a 1000 W/m2 in one of the exterior surface boundaries, this one having a total surface of (1m*5m=5m2), therefore, my total Power input is 5000 W. The rest of the surfaces are insulated.
The velocity at the inlet is 0.10 m/s. The temperature at the inlet is 293.15 K.

I just want to confirm that the heat removed by the air, Q = mcp (Toutlet-Tinlet) holds in the solution.

First, I am checking the energy balances with surface integration at the boundaries.

Then, the total heat crossing the boundaries of the interior boundaries, should be 5000 W.

When I integrate the normal conductive heat flux on the surface (nitf.ndflux), gives the result of -5000 W, (that’s ok).
However, if I integrate the conductive heat flux magnitude (nitf.dfluxMag) at the same surface I get 38150 W? Why is that number larger than the 5000 Watts input? I also obtain a similar number if I integrate the total energy flux magnitude (nitf.tefluxMag), 38150.74 W.

The mass balance at the inlet seems pretty good. I do a surface integration at the inlet of (nitf.rho*nitf.U (kg/s)) 0.010132687 kg/s. A back of the envelope calculation would be given by: 1.2kg/m^3*0.1m/s*0.1m^2 = 0.012045 kg/s. The surface integration at the outlet of (nitf.rho*nitf.U (kg/s)) gives 0.0098984 kg/s, or a relative difference from the integration at the inlet of 2.4%.
If I integrate the total energy magnitude at the inlet (nitf.tefluxMag), I get 2985.58 W. Doing the same operation at the outlet, I get 3590.96 W. The difference (oulet-inlet) being 605.38 W.

If I integrate the temperature at the inlet, T, I get = 29.3149 m^2*K. After dividing by the surface area (0.1m2), I get 293.149 K (very close to the initial condition set at the beginning of 293.15K).
Integrating the temperature at the outlet I get 37.105 m^2K, then 371.05 Kelvin.

Finally my energy equation looks like

Q= mcp(Tou-Tinlet) = 0.010142687 kg/s*1100 J/(kg-K)*(371.05K-293.149K)= 869 W.

How come this number is not even close to the 5000 Watts of input I have, when I don't even have any energy sinks or thermal losses?

Thank you everybody for your comments,

Hi

Indeed you have some interesting questions there, I would appreciate that COMSOL adds a chapter per physics about model validation, among other how to correctly check in/ out energy, momentum etc

In fact I'm also stuck in the same considerations, but for a long thin water pipe dragging out heat from the tube external surface, and I'm facing the same questions as you.

A few things concerning your model, I see you have no boundary layers in the fluid, I can only recommend to use them, needs still some hand tweaking, probably automatic in next release.
And to make the model converge better I mostly add pressure drop (Poiseuille) and a parabolic velocity profile as initial conditions.

Then for the heat leaving with the fluid, do not forget to integrate

velocity[m/s]*density[kg/m^3]*heat_capacity[J/kg/K]*Temperature[K] over the area[m^2]

to take into account the velocity variation, and then to subtract the same for the inlet, because of the absolute value of the temperature T in K

--
Good luck
Ivar

Dear Ivar,

Thank you very much for your help.

I have contacted customer support to see what they recommend. Do you think this might be a bug in the GUI?

I have not been able to find a sample model of Conjugate heat transfer or CFD describing step by step how to compute the heat/mass fluxes.

I will keep you posted,

Cheers,

Hi

I haven't found many true bugs in last patch of v4.1, and none today of the type wrong physics (so far, but I haven't tested absolutely everything ;)
However, as often, there are a few issues where the programmer and the end user do not think/understand things the same way, this might be one such case.
That is why I'm still searching, and I might also end up sending also a mail to "support" ;)

--
Good luck
Ivar

Hi

Do not forget to take a look at the doc v4.1, in particular chapter 2, "A Note on Heat Flux", page 23

--
Good luck
Ivar

I typically ignore the boundaries that use the default "insulated" boundary condition. One expect these to be shown zero using the integration tool in the GUI, but are not. There is a knowledge-base question on this topic : www.comsol.com/support/knowledgebase/973/
The detail they describe using weak constraints is only necessary if you really need to be very precise about exactly what is being computed. I found that the integration tools available in the GUI are sufficient if you simply ignore the insulated boundaries in your energy balance, and then you will indeed get near balance. As you add more elements, the converged solution approaches zero energy balance as you would expect. In fact, we just went through this learning exercise once again in our own work here. You just simply trust that the insulated boundary is indeed insulated. You should be able to verify this graphically anyway from your solution.

Hi

I have noticed that integrating over all external boundaries for the "normal" components, gives reasonable results, while if you integrate the "total" values you add also the heat gradients running along the "isolated boundaries". Which correspond I believe to your remark about ignoring the "isolated boundaries"

Then one should not forget the "energy" in the moving fluid, pressure work (if applicable) etc that changes the balance, depending if you look at the energy ot the heat flux.

Still I have a another issue in TS, where I get only half the power out of the constant temperature BC compared to the influx heat, but the solution looks reasonnable. But when I make a simpler model the balance is OK

Furthermore I'm fighting to stabilise the pressure in NITF with stationary simple laminar water flow + HT case, even by giving almost the correct pressure drop, while solving, the pressure is fluctuating madly, and funnily on the odd (not on the even) solver steps it even starts (new initial conditions) to invert, which means long time to converge

--
Good luck
Ivar

Dear Ivar,

Thank you very much for your help.

I have contacted customer support to see what they recommend. Do you think this might be a bug in the GUI?

I have not been able to find a sample model of Conjugate heat transfer or CFD describing step by step how to compute the heat/mass fluxes.

I will keep you posted,

Cheers,

Hey,

I would like to know that what is the outcome of the contact with customer service. I am doing similar analysis and could not figure out the same problem also. Can you explain how did you solve the problem?

Thanks a lot.

I also face the problem of matching different quantities with the measurements from real experiments. Please let us know if anyone found a solution to this problem.

Hi Ivar
Did you figure out what is the problem? Because I have the same problem