2-way FSI fails during first time step

    • UniversityofQLD

      Hello everyone,

      I have been trying for a long time to get my 2-way FSI (Mechanical and CFX) model of an idealised pipe (which is supposed to simulate an artery in the future) to work but it keeps crashing with errors before finishing the first time step. I cannot get even an extremely simplified model to work for any material softer than a Young’s modulus of 50 MPa.

      I have looked at all relevant tutorials (best practices for FSI, oscillating plate, etc). The model I am actually trying to solve is quite complex with a coded Windkessel outlet, initial stress, hyperelastic soft material and time-dependent velocity inlet. It works completely fine when solving the individual solvers and when solving a 1-way FSI (CXF->Mechanical). However, I can simply not solve a 2-way FSI simulation with this geometry (which is literally just an idealised pipe) no matter how much I simplify it, except when solving for water in a relatively hard pipe.

      95% of the time I will get an “excessive deformation” error (and the model looks "blown out" in the structural results view) but depending on the support type and meshing technique I occasionally also get “negative volume” or “unable to converge on a solution for the non-linear problem as constrained” errors. In cases where the material is still pretty hard it might not crash but it cannot converge. I have tried building the model from scratch several times and I am now working with extremely simple models (e.g. water with constant velocity flowing through a isotropic linearly elastic pipe without turbulence) but I can still not solve a 2-way FSI with a softness anywhere close to human arterial tissue.

      I have tried the following things (and checked if the individual solvers and 1-way FSI work):


      -         GEOMETRY

      -         Structural: Simple idealised pipe. Internal diameter: 5.5 mm, thickness: 1 mm, length: 100 mm

      -         Fluid: Cylinder with same diameter as internal diameter of pipe

      -         I have tried importing the geometry as different file types (SolidWorks part, stl, binary Parasolid)

      -         I have checked the geometries and they completely match (no scaling issues during import)


      -         MATERIALS

      -         The only material I can solve the 2-way FSI simulation for is structural steel

      -         I tried working with linearly elastic materials with different Youngs Modulus. Since I am working with small pressures and velocities, the pipe is not deforming visibly even for a Young’s modulus of 100 MPa in the mechanical solver (deformation around e-8 m ) but again blows up in the 2-way FSI when applying the same pressures (deformation around e-2)

      -         Tring different Poisson Ratios


      -         MESHING

      -         Working with fine meshes generated in Ansys Meshing (good values for min angle, quality, skewness and orthogonality). Tetra or Hex. Homogenous or with inflation near data transfer surface

      -         Alternatively working with fine tetrahedral meshes generated in ICEM (homogenous mesh density or becoming finer around the data transfer surface or inlet+outlet. Smoothed to ensure good values for min angle, quality, skewness and orthogonality.

      -         Trying different mesh motion settings (e.g. having the inlet and outlet faces “stationary” or “unspecified”)



      -         Ramping a small pressure or velocity from 0 in CFX (e.g. v=0.5*t). Still crashes in the first timestep even though there are basically no forces applied to the system.

      -         Various types of support (fixed on the ends or cylindrical or displacement or elastic etc)

      -         Outlet as opening with no pressure, small pressure or Windkessel function

      -         Trying with or without initial stress mapped onto geometry.

      -         Trying to solve just with CFX and checking that all pressure and velocity values in different locations have the expected magnitude and direction. This was also still the case for 1-way FSI results.


      -         SOLVER

      -         Trying the “System Coupling” and “Fluid-Structure Interaction” interfaces

      -         Using the Ramping option in System Coupling for the data transfers

      -         Using Quasi-Newton Stabilisation in GUI

      -         Trying different Windows workstations

      -         Building the model in different versions of Ansys (R2022 R1 and R2020 R2)

      -         Trying out small of timesteps of e-4 and less but stability is not improved and convergence is not accelerated

      -         Trying direct and iterative solvers for Mechanical and first and second order backwards Euler for CFX


      My external collaborators are solving very complex FSI models with CFX and Mechanical such as heart valves and patient-specific stented arteries so I am very confused that even when following their tips and advice I cannot even get such a simple model to work. Any advice would be highly appreciated. Thank you!

    • Stephen Orlando
      Ansys Employee
      I suspect the issue is due to the numerical instability between the fluid and structural solver. Using Quasi-Newton can help, but not in all cases. You can also try using the Solution Stabilization option in Fluent. This requires manually tuning to find the correct value.
Viewing 1 reply thread
  • You must be logged in to reply to this topic.