Change FSILS: unexpected behavior in FSILS error condition back to a warning

[ktbolt]

Problem

In ns_solver if an ill-conditioned LHS is computed the a runtime exception is thrown

FSILS: unexpected behavior in FSILS (likely due to the ill-conditioned LHS matrix)

However, a simulation may start off with an ill-conditioned LHS but later settle down so throwing an exception and halting the simulation is too strict.

Solution

Have ns_solver print a warning if an ill-conditioned LHS is computed.

Additional context

No response

Code of Conduct

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

[ktbolt]

In svFSI if an ill-conditioned LHS was computed then a warning message was printed and the simulation was not halted. I'm now thinking that this was wrong to halt the simulation because the simulation may start off with an ill-conditioned LHS but later settle down.

This is what happens for the Cylinder Example run on a single processor

---------------------------------------------------------------------
 Eq     N-i     T       dB  Ri/R1   Ri/R0    R/Ri     lsIt   dB  %t
---------------------------------------------------------------------
libc++abi: terminating due to uncaught exception of type std::runtime_error: FSILS: unexpected behavior in FSILS (likely due to the ill-conditioned LHS matrix)

Changing throwing an exception to a warning

---------------------------------------------------------------------
 Eq     N-i     T       dB  Ri/R1   Ri/R0    R/Ri     lsIt   dB  %t
---------------------------------------------------------------------
[svMultiPhysics:FSILS] Warning: Unexpected behavior in FSILS (likely due to the ill-conditioned LHS matrix)
 NS 1-1  7.510e-01  [0 1.000e+00 1.000e+00 0.000e+00]  [6 0 75]
 NS 1-2  1.481e+00  [-8 3.692e-01 3.692e-01 4.217e-05]  [6 -14 75]
 NS 1-3  2.097e+00  [-31 2.769e-02 2.769e-02 7.485e-05]  [5 -13 71]
 NS 1-4  2.681e+00  [-58 1.129e-03 1.129e-03 5.820e-05]  [5 -13 69]
 NS 1-5  3.304e+00  [-85 5.526e-05 5.526e-05 9.675e-05]  [5 -12 71]
 NS 2-1  3.911e+00  [0 1.000e+00 3.986e+02 8.836e-05]  [3 -11 71]
 NS 2-2  4.661e+00  [-61 7.947e-04 3.168e-01 3.483e-05]  [6 -13 76]
 NS 2-3  5.345e+00  [-85 5.176e-05 2.063e-02 5.877e-05]  [6 -11 74]
 NS 3-1  5.804e+00  [0 1.000e+00 6.306e+02 4.529e-05]  [2 -19 60]
 NS 3-2  6.559e+00  [-68 3.765e-04 2.374e-01 4.156e-05]  [6 -13 76]
 NS 3-3  7.371e+00  [-93 2.116e-05 1.335e-02 4.204e-05]  [7 -12 78]

allows the simulation to settle down and complete.

[ktbolt]

And if the Cylinder simulation is run on multiple processors the unexpected behavior in FSILS does not occur.