# Flow Network model - missing some heat

I have a test problem that uses heat transfer with a pipe flow network. There is no pressure drop calculation, only mass flow and capacitance are considered (test model below).
The figure shows the issue. I apply a convection on the base, a mass flow and inlet temperature to the pipe. I expected the convection, conduction and thermal mass transport to all be the same, but the pipe flow “cool down” is about 30% too low. Any help would be appreciated!

---------------- pipetest.inp
*NODE
1,0,0,0
2,0.01,0,0
3,0.01,0.01,0
4,0,0.01,0
5,0,0,0.01
6,0.01,0,0.01
7,0.01,0.01,0.01
8,0,0.01,0.01
9,0,0,0.02
10,0,0,0.03
11,0,0,0.04
12,0,0,0.05
13,0,0,0.06
14,0,0,0.07
15,0,0,0.08
16,0,0,0.09
17,0,0,0.1
18,0.01,0,0.02
19,0.01,0,0.03
20,0.01,0,0.04
21,0.01,0,0.05
22,0.01,0,0.06
23,0.01,0,0.07
24,0.01,0,0.08
25,0.01,0,0.09
26,0.01,0,0.1
27,0.01,0.01,0.02
28,0.01,0.01,0.03
29,0.01,0.01,0.04
30,0.01,0.01,0.05
31,0.01,0.01,0.06
32,0.01,0.01,0.07
33,0.01,0.01,0.08
34,0.01,0.01,0.09
35,0.01,0.01,0.1
36,0,0.01,0.02
37,0,0.01,0.03
38,0,0.01,0.04
39,0,0.01,0.05
40,0,0.01,0.06
41,0,0.01,0.07
42,0,0.01,0.08
43,0,0.01,0.09
44,0,0.01,0.1
45,0,0.02,0
46,0,0.02,0.05
47,0,0.02,0.1
48,0,0.02,0.025
49,0,0.02,0.075
*ELEMENT,TYPE=DC3D8
1,1,2,3,4,5,6,7,8
2,5,6,7,8,9,18,27,36
3,9,18,27,36,10,19,28,37
4,10,19,28,37,11,20,29,38
5,11,20,29,38,12,21,30,39
6,12,21,30,39,13,22,31,40
7,13,22,31,40,14,23,32,41
8,14,23,32,41,15,24,33,42
9,15,24,33,42,16,25,34,43
10,16,25,34,43,17,26,35,44
*ELEMENT,TYPE=D
11,45,48,46
12,46,49,47
*NSET,NSET=pipe_flow_mdot
48
49
*NSET,NSET=temp_inlet
45
*NSET,NSET=pipe_flow
45
46
47
*ELSET,ELSET=wpipe
11
12
*ELSET,ELSET=solid
1
2
3
4
5
6
7
8
9
10
*MATERIAL,NAME=dummy_thermal
*DENSITY
7700
*CONDUCTIVITY,TYPE=ISO
0.001,0
*SPECIFIC HEAT
3850
*SOLID SECTION,ELSET=solid,MATERIAL=dummy_thermal
*FLUID SECTION,MATERIAL=WATER,ELSET=wpipe
*MATERIAL,NAME=water
*FLUID CONSTANTS
4186,8.9E-4,310

*STEP
1,1,0,0
*BOUNDARY
temp_inlet,11,373.15
*FILM
2,F3,293.15,1
1,F3,293.15,1
3,F3,293.15,1
4,F3,293.15,1
5,F3,293.15,1
6,F3,293.15,1
7,F3,293.15,1
8,F3,293.15,1
9,F3,293.15,1
10,F3,293.15,1
*NODE FILE,GLOBAL=YES
NT
*EL FILE
HFL
*BOUNDARY

pipe_flow_mdot,1,1e-5

*** pipe
*FILM
7 ,F5FC, 46,70.123
6 ,F5FC, 46,70.123
8 ,F5FC, 47,70.123
5 ,F5FC, 46,70.123
9 ,F5FC, 47,70.123
4 ,F5FC, 46,70.123
10 ,F5FC, 47,70.123
3 ,F5FC, 45,70.123
2 ,F5FC, 45,70.123
1 ,F5FC, 45,70.123
*END STEP

Hi John,

How cool is this.!!!. Thanks for sharing. It’s kind of an HEX tube temperature profile.

I’m obtaining similar result but would like to comment some points with you.

-I have noticed that the final temperature depends on how many surfaces are assigned to each Temperature node. ¿Could this be the source of discrepancy?
You are assigning (3,4,3).

I have refined and discrepancy has drop to 20% which is still large. Each node has two surfaces now.

By other hand, if we compare Q’s assuming the same temperature on top and bottom of the solid bar , we are indirectly assuming perfectly homogeneous transverse heat transfer in the equilibrium.Not sure about that considering the nature of the heat flow on the source (longitudinal) and the small number of elements.

I keep investigating. Maybe it is a bug, we never know.

1 Like

@Disla thanks for taking a look, and yes this stuff is cool. This method is used by aerospace to model bypass cooling air in turbine blades, disks and ducts, and since Guido works at MTU Aero it’s no surprise that this is part of the Calculix tool set. We use this a lot as a “poor man’s conjugate heat transfer”. At some point I will share the python API scripts that generate the network and I’m hoping to get our friend Victor to build this in as a command set. But in the mean time I would like to minimize the heat balance discrepancy. As you say, the error gets reduced a bit with mesh density, but I was hoping to get better than 20%. If you make any discoveries, let me know. Thanks!

Can’t you use CalculiX + OpenFOAM with preCICE for CHT ?

I have always found preCICE interesting, and may find a use for it, but the pipe network approach is very effective and very fast for what we need. When you understand the flow side well (i.e. know the heat transfer coefficients) and only need to consider the energy transfer to/from the fluid, the network is a great tool.

1 Like

HI John

I have obtained a 0.17% of deviation with the following distribution:
Use the same number of elements than surfaces and two nodes for each surface .

1,F5FC,45,70.123
1,F5FC,93,70.123
2,F5FC,93,70.123
2,F5FC,95,70.123
3,F5FC,95,70.123
3,F5FC,97,70.123
4,F5FC,97,70.123
4,F5FC,99,70.123
5,F5FC,99,70.123
5,F5FC,101,70.123
6,F5FC,101,70.123
6,F5FC,103,70.123
7,F5FC,103,70.123
7,F5FC,105,70.123
8,F5FC,105,70.123
8,F5FC,107,70.123
9,F5FC,107,70.123
9,F5FC,109,70.123
10,F5FC,109,70.123
10,F5FC,47,70.123

Mdot 1.00E-05 Kg/s
Cp 4186
AT 0.173 ºC 100.000 ºC 99.827 ºC
Q 0.007242 W

Surface Integral Top - 0.007254 In
Surface Integral Bottom 0.007254 Out

Deviation 0.17 %

1 Like

Nice! I’ll give that a try, sounds like you have the handle on it! Have a great weekend.