[mikeh]

I am having difficulty setting up a FLUENT13 simulation of a supersonic nozzle. I am relatively new to CFD and have had some success with simple convergent-divergent nozzle geometries, however, I find it very difficult to get convergent repeatable results with even small changes to the problem setup.

My problem is basically a convergent divergent Venturi tube with a fixed inlet pressure. The nozzle exhausts into atmosphere at supersonic velocities and I am particularly interested in modelling the flow in the output domain.

From tutorials for 2d-axisymetric nozzles I have had limited success in modelling the flow within the nozzle without including the output domain in the simulation, but simply putting a pressure outlet boundary at the nozzle end.

Presently I am attempting to use 3D solid model of my part reduced to a 45° sector and using periodic rotational boundaries in the simulation. I have enclosed my entire cylindrical nozzle in a large cylindrical domain that is many times the diameter and length of the part with one end aligned to the nozzle inlet port.

My simulation setup is:

• Density based solver, 3D, steady state
• Standard k-e model
• Default mesh with inflation at the part boundaries (looks reasonable to me). 352k cells.
• Fluid is ideal gas air
• Pressure inlet defined (with a supersonic guess pressure) with 1% turbulent intensity
• All far-field boundaries defined as "outlet-vent" at 100kPa, turbulent intensity set to 0.
• Solution method: Explicit, Roe-FDS, Least squares, second order flow, first order turbulence.

The simulation residuals are quite unstable and eventually it will diverge. If stopped at a minima the results look somewhat reasonable.

I would greatly appreciate any advice on how to get my simulation to converge more reliably without oscillation. I am hoping that I can find a setup that will allow me to vary nozzle geometry without disturbing the sim parameters.