[mabinty]

Dear all,

I am simulating a kind of beam fixed on top and loaded on the bottom in longitudinal direction of the beam.

The used solver is elasticSolidFoam (foam-extend 3.1) but the same behaviour is observed with elasticIncrSolidFoam. When the non-orthogonal correction is switched off, such as

 laplacianSchemes
{
    default         none;
    laplacian(DU,U) Gauss linear uncorrected;
}

snGradSchemes
{
    snGrad(U)       uncorrected;
}

the simulation converges quite fast. At the moment I switch on non-orthogonal correction ("corrected" instead of "uncorrected" for laplacian(DU,U) and snGrad(U)) the solver crashes after several hundreds of iterations with a floating point exception (log.elasticSolidFoam see attachments: Time 1 with, Time 2 without non-orthogonal correction).

The mesh was generated with snappy (checkMesh log see attachment). The average non-orthogonality seems to be OK and its correction should not introduce too much instability (at least for my experience with CFD) so I currently do not know where the real problem lies.

I also tried to play with the settings in the system/fvSolution dictionary such as with the methods for the explicit term divSigmaExp, or switching on/off the aitkenRelaxation. No improvements!

I greatly appreciate your comments!

Cheers,
Aram