Hydraulic pipe system with branch split

Windows 10, CCX 2.16

CalculiX has interesting hydraulic network capabilities, including branches (redirecting flow to two other pipes). There are Calculix examples for gas flow (Branchsplit2) and they work. However, there are difficulties for liquid flows (water). The following input shows a small test example with only one branch split, one entry and two exits.

** Branch Split with IDELCHIK1
1, 0.09, 0., 0.
2, 0.10, 0., 0.
3, 0.15, 0., 0.
4, 0.20, 0., 0.
5, 0.25, 0., 0.
6, 0.30, 0., 0.
7, 0.31, 0., 0.
8, 0.20, 0.05, 0.
9, 0.20, 0.10, 0.
10, 0.20, 0.11, 0.
*Element, type=D, Elset=EALL
1, 0, 1, 2
2, 2, 3, 4
3, 4, 5, 6
4, 6, 7, 0
5, 4, 8, 9
6, 9,10, 0
*Elset, elset=Branch
*Elset, elset=INOUT
**Cp Eta Temp
4217.,1750.E-6, 273.
2, 3, 5, 0.008, 0.008, 0.002, 0, 90
0.101, 0.0505, 0, 0, 1, 1, 0
2, 3E5
4, 2E5
1, 1,1, 1.0
6, 2,2, 1E5
9, 2,2, 1E5

The launcher window says: Error in liquidpipe: loss coefficients for branches do not apply for reversed flow. Due to initial and boundary conditions on pressure DOF 2 there is a pressure gradient downstream that should avoid reversed flow.

If anyone could help, it would be great.
Best regards,

Clearly, the solver claims that there is a reversed flow in your model which is not allowed for branch split type of fluid section.

That’s the system you want to simulate, right ? To make sure that initial and boundary conditions are correct, I would analyze this using standard CFD approach.

Did you check it with gas properties instead of water ?

thank you for your quick response. Yes, your plot with the element numbers is correct. Pressure boundary conditions are necessary only for the outlets at nodes 6 and 9 (atmospheric pressure = 1 bar = 1E5 Pa). Initial conditions are helpful for accelerating convergence, they were specified at nodes 2 and 4 in a meaningful way. Trying other combinations at these nodes, all of them result in “reversed flow”. The aim of this test analysis is to find out the pressure drop at the branch split and to gain confidence for larger problems. Theoretical solutions for the 90 degree branch split are available (I.E. Idelchik Handbook of Hydraulic Resistance, Diagram 7.21). It was hoped that alternatives like CFD can be avoided by using 1D-analysis with CalculiX as a first approximation.
I tried the same problem with gas (air properties) → divergence - too many cutbacks.
Thanks and greetings,

the branch split elements cannot be used on their own without any adjacent elements such as pipe elements (cf. Figure 102 in the ccx_2.19 documentation). If a pipe 1 splits into pipe 2 and pipe 3, define:

  • a pipe element for pipe 1
  • a branch element followed by a pipe element for pipe 2
  • a branch element followed by a pipe element for pipe 3.
    Both branch elements have the same properties.
    Please look at examples branchsplit3 for branch details and example piperestrictor for hints how to use genuine liquid elements in combination with gas elements of type liquid. Notice that genuine liquid elements use the hydraulic radius defined as A/P, whereas gas elements use the hydraulic diameter defined as 4A/P (A: area; P: wet perimeter). The definition A/P for radius may seem confusing, I may change this in the future.

I modified your example and can send it to you if you send an E-mail to dhondt@t-online.de.

Best Greetings,


1 Like

Thank you, these are the crucial requirements for branch split (or branch joint) elements to work. Note: For larger hydraulic networks, e.g.

the necessary initial pressures ‘pinit’ for the points P1 to P4 must be estimated qualitatively correct by the user, here thus p1 > p2 > p3 > p4 > p0, otherwise the user gets “reversed flow” and termination. This can be difficult with very large networks.
CalculiX provides in addition to the dat file a net file for the branch elements, which would be interesting. However, the Reynolds numbers and zeta values (loss coefficients) given there are all zero. Conclusion: The hydraulic elements are a highlight in CalculiX.
Thanks and greetings,