nonNewtonianIcoFoam: errors near boundaries
 

hey folks,

i am currently using nonNewtonianIcoFoam to simulate the transient blood flow in an artery. BCs for the outlets are constant pressure, and for the inlet i prescribe a velocity profile. For the viscosity I use CrossPowerLaw, but the results with BirdCarreau are very similar.
Two things about the results are very strange:

1) When the inlet velocity is decreasing very rapidly, the viscosity field shows very large gradients at the boundary of the domain (see attached picture). This effect goes away as soon as the velocity is increasing again, but it clearly can't be right.
I thought it could have something to do with discretization, so I tried using the upwind and the QUICK schemes for div(phi,U) and laplacian(nu,U). This had no visible effect.
I also decreased the time step (although it was already very small before: 1e-4s) to 5e-5s. Again, this had no effect.

Apparently the reason for these large nu values are overly small gradients in the velocity field near the boundaries. How can these be avoided?

2) Even when the above effects do not appear, the range of viscosity values seems way to large to me: approximately 3e-6 to 5e-5 (or, multiplied by density: 0.003 Pas to 0.05 Pas). According to literature, this is usually not the case in arteries such as the one i'm modelling.
I realized that when I do steady-state simulations, i.e. constant inlet velocity, the values are in a much more realistic range after around 10 time steps.



I already did a couple of simulations with solvers based on icoFoam using the same mesh, BCs, and discretization schemes, and every thing worked fine. But maybe the mesh is not adequate near the boundaries for non-uniform viscosity?



any ideas/hints why the results are so strange?


thx in advance..