[Kerouac]

Hi everyone!
I'm trying to simulate the melting and flow of a non-Newtonian fluid in a duct with Fluent 19.2 since three months.. and I'm finding great problems with the numerical convergence.
In particular I've to deal with solidification + laminar flow with viscous heating + energy, as physical models. The constitutive equation for fluid viscosity is as in the attached image.

viscosity.png

Where : T_g=373.15K. These are data for ABS resin, yet published in a number of papers.
For solidification/melting physics I've used A_mush=1e3 and I've not included pull-velocity calculation.
My mesh is done with hexa/poly-elements and, because of the fact that I'm simulating the flow of this fluid in a duct, the mesh is done on the first plane (inlet) and then gradually extruded through the entire domain.
The upper wall is moving, with a value of temeprature T=503K, while the other walls are stationary.
My study is steady state and is done with pressure-based solver. Both proceding with first order and second order schemes I'm not able to find a solution.
If I use less great values of viscosity (for example, if for the variable mu0 I set a lower 'threshold' for the if in the UDF) the solution converges, but clearly I can't motivate physically this choice..
Is it possible that Fluent can't handle too high values of viscosity?
The entire study is performed reducing up to 0.1 the residuals for all variables, but I can't get a solution.
Inlet b.c is that of a give mass-flow rate and outlet is 'outflow' boundary condition. I've also tried with 'outlet' b.c. but it did not converge.
I will provide more details, if you ask me or if I've not been clear in what I've provided.
Best regards,
Kerouac