Dirichlet 3D–0D genBC Coupling Diverges in CFD

[zinanhu0810]

Description

When the 3D model is coupled to a 0D genBC using a Dirichlet boundary condition in a CFD simulation, the simulation diverges immediately.

Here is the usage I have in xml file:

   <Add_BC name="lumen_inlet" >
      <Type> Dir </Type>
      <Time_dependence> Coupled </Time_dependence>
      <Profile> Parabolic </Profile>
      <Impose_flux> true </Impose_flux>
      <Effective_direction> (0, 0, 1) </Effective_direction>
   </Add_BC>

Simulation diverges immediately:

---------------------------------------------------------------------
 Eq     N-i     T       dB  Ri/R1   Ri/R0    R/Ri     lsIt   dB  %t
---------------------------------------------------------------------
 NS 1-1  5.100e-01  [0 1.000e+00 1.000e+00 5.551e-04]  [4 -16 55]
 NS 1-2  1.017e+00  [1 1.139e+00 1.139e+00 2.022e-04]  [5 -17 62]
 NS 1-3  1.467e+00  [16 6.686e+00 6.686e+00 8.851e-04]  [4 -18 58]
 NS 1-4  1.920e+00  !41 1.163e+02 1.163e+02 7.885e-04!  [4 -17 58]
 NS 1-5  2.737e+00  !58 8.396e+02 8.396e+02 6.908e-04!  [8 -7 76]
 Error! NAN encountered..Dir
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_OVERFLOW_FLAG
 NS 1-6  3.903e+00  !45 1.984e+02 1.984e+02 4.206e-03!  !9 -3 83!  WARNING: The linear system solution has not converged
 Error! NAN encountered..Dir
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_OVERFLOW_FLAG
 NS 1-7  5.023e+00  !33 4.891e+01 4.891e+01 2.628e-03!  !9 -4 83!  WARNING: The linear system solution has not converged
 Error! NAN encountered..Dir
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_OVERFLOW_FLAG
 NS 1-8  6.217e+00  [19 9.818e+00 9.818e+00 3.852e-03]  !9 -3 84!  WARNING: The linear system solution has not converged
 Error! NAN encountered..Dir
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_OVERFLOW_FLAG
 NS 1-9  7.390e+00  [8 2.556e+00 2.556e+00 3.756e-03]  !9 -3 84!  WARNING: The linear system solution has not converged
 Error! NAN encountered..Dir
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_OVERFLOW_FLAG

I’m going to set up a simple test case with a prescribed flow rate in genBC and check whether this issue also occurs in svFSI.

Reproduction

Attached is a simulation folder that couples the inlet and outlet to a closed loop lumped network.

Conduit_genBC.zip

Expected behavior

The simulation should converge and produce reasonable results.

Additional context

No response

Code of Conduct

• I agree to follow this project's Code of Conduct and Contributing Guidelines

[ktbolt]

I've reproduced this behavior, will have a look in the code.

[ktbolt]

@zinanhu0810 The SUBROUTINE FINDF in src/USER.f is not for a simple cylinder problem

      INTEGER, PARAMETER :: nbr= 24    !SET NUMBER OF PULMONARY BRANCHES#####
      INTEGER, PARAMETER :: ninf= 4    !SET NUMBER OF INFERIOR BRANCHES#####

! Ventricle.  normalized elastance, these functions peak at 0.3/T

!  "Resting" elastance shape
      fe=(/  (0.223791662, 0.000000000),

[zinanhu0810]

It’s a slightly complicated LPN, but there are only two coupled surfaces in the setup. The pulmonary branches with nbr = 24 are treated as 0D nodes only. I can also create a much simpler LPN to debug the issue.

[ktbolt]

@zinanhu0810 That's OK, I can use this one, just wanted to confirm that this GenBC code worked correctly for svSolver.

[aabrown100-git]

Just pointing out this test case: https://github.com/SimVascular/svMultiPhysics/tree/main/tests/cases/fluid/pipe_RCR_genBC. This has a Dirichlet BC that could be modified to be a 0D-coupled Dirichlet BC.

[ktbolt]

@aabrown100-git Thanks, the GenBC code in pipe_RCR_genBC is much simpler !

[ktbolt]

@zinanhu0810 The failure occurs during the RK4 integration after several iterations, have certain RK terms f that are becoming very large and then Infinity here

f(5) = (x(4) - x(5)*Rthao - x(6)) / Lthao

I think I will need a simpler example to track down what the problem might be.

[zinanhu0810]

Attached is a simple example with a constant inlet flow rate and an RCR outlet prescribed via genBC. The simulation runs smoothly, and the inlet velocity profile shows correct flow rate and a parabolic shape. Maybe the divergence reported above stems from the genBC equation and parameters.
Conduit_genBC.zip

[ktbolt]

@zinanhu0810 So it seems that data is being passed correctly between GenBC and svMultiPhysics which is the main thing we wanted to verify.

The genBC USER.f file in Conduit_genBC.zip is 400 lines of messy code, commented-out lines, not clear what all is going on. And it is the RK4 integration that is generating Infinity and NaN values.

So should we continue looking into this ?