Discussion Forum

Dear Team, I am trying to solve a fluid-structure interaction-based problem (Vibrating Beam in Fluid Flow) using COMSOL Multiphysics 5.5. I also followed the below tutorial. www.comsol.com/model/vibrating-beam-in-fluid-flow-9408.

Following the above model, I am able to validate most of the data points of the following references: 1. Turek, Stefan, and Jaroslav Hron. Proposal for numerical benchmarking of fluid-structure interaction between an elastic object and laminar incompressible flow. Springer Berlin Heidelberg, 2006. 2. Furquan, Mohd, and Sanjay Mittal. "Multiple lock-ins in vortex-induced vibration of a filament." Journal of Fluid Mechanics 916 (2021): R1.

Next, I wanted to run this model for different parameter spaces similar to my experiment. But, I failed to simulate and got the time step divergence error. I have been trying to resolve it for a long time (more than 3 months) by changing grid sizes, time steps,s and also the different models. A representative .mph file I have attached here. The errors for different model setups are mentioned below.

1. For the attached file using Seggregated solver, I found the error: Recoverable error in residual, but solver could not recover. Time: 376.464283915321 s. Last time step is not converged.

2. I studied the blog: www.comsol.com/support/knowledgebase/1262. I followed different Time stepping methods (BDF, intermediate, strict, manual), Events for Modeling Step Changes (step function for inlet velocity), Solver Relative Tolerance, Initial Time Step, Nonlinear Solver Settings, Scaling, Consistent Initialization (faced issues using 2 steps or 2 studies, ist stationary and then time-dependent, hence left). But all the cases, I got error: Nonlinear solver did not converge. Maximum number of segregated iterations reached. Time: 322.91395075694669 s. (different for different cases) Last time step is not converged. or Recoverable error in residual, but solver could not recover. Time: 376.464283915321 s. (different for different cases) Last time step is not converged.

3. Next, I found a large deformation of moving mesh, and the mesh distorted more. The quality of the mesh elements decreased. In some cases, inverse mesh elements were observed. So, I used automatic re-meshing. And got the error for all different cases (including changing the limit condition of mesh quality and distortion) : The following feature has encountered a problem:

- Feature: Automatic Remeshing (sol1/t1/arDef) Failed to find consistent initial values. Last time step is not converged. I also tried adaptive meshing but, my system was not good enough to simulate that. Hence I left adaptive meshing.

4. I tried a fully coupled solver rather than a segregated solver. I got errors (using with and with out re-meshing and other different settings): Repeated error test failures. May have reached a singularity. Time: 262.82708100085091 s. Last time step is not converged. or Nonlinear solver did not converge. Maximum number of Newton iterations reached. Time: 249.81531460024419 s. Last time step is not converged.

For your information, in our experimental case, it is a wall boundary condition. But, I found if I use an open boundary then there was not any divergence error for the parameter space of the attached setup file (although there was large deformation and distortion of the moving mesh elements). But, if I again decrease the bending stiffness of the solid (modulus of elasticity) then again it shows 'Last time step is not converged'.

Looking forward to your kind response to resolve this issue.

Thanks and Regards, Shubham Giri Ph.D. student