[DrZoidberg]

Hi all,



I am wondering if there is a faster way to get to a converged solution for a high-pressure duct flow that enters the environment. Usually, hybrid or FMG initialization may give a good guess to start from in low Ma regimes. However, for my geometry and BC it always leads to divergence. Getting a useful inviscid solution is troublesome either.



The setup:
Air-ideal gas, sutherland visco model, kwSST with compressibility and production limiter

Pressure inlet (1e7Pa, 800K, turbulence intensity 5%) for the duct (zigzag shape but constant cross section)

Pressure outlet (1e5Pa, 300K, turbulence intensity 1e-5, far away from duct end, which is a straight nozzle withlut angles or curvature)



My solution so far:
None of the common initializations worked for this case. As recommended by FLUENT manual for such hard cases, I init for inlet conditions and decrease the outlet pressure via a multi-step function (steps of -5e5Pa for timescale/iteratios 0.01/2000, 0.1/1000 1/500). Steps must be decreased even more, the closer it is to p_outlet=1e5Pa. This works, but is quite tedious and requires scheme scripts. It took almost a week for the 2d case on a structured mesh with 800k elements (orth quality >0.65, AR in BL for y+=1 is 2000).



Sometimes pressure ramps in the form of expressions are recommended. I tried linear, tanh-like steps but it always diverges. I guess the problem is, for using such expressions for the pressure values, that the timescale factor is fixed and the solver has no chance to propagate the changes through the domain.