[g7sznAc4]

Hello,

I'm trying to use Fluent to simulate a fairly simple geometry in 2D axisymmetric mode. I have a long duct train feeding a choked nozzle. The duct train consists of a bellmouth inlet (with total pressure/temperature applied at the inlet face) followed by a constant area pipe, diffuser, more pipe, and a choked ASME nozzle. The nozzle exhausts into a quiescent flow with farfield boundary conditions (I'll attach a picture of the geometry).

I'm trying to get Fluent to match experimental boundary layer rake data in each of the constant area pipes, so that I can then use Fluent as a design tool to develop some updates to the duct train.

However, Fluent is providing very poor predictions for the boundary layer thickness and shape (I'll attach a plot, created at the second/middle blue line in the above picture). The experimental flow is far more developed than the CFD flowfield. I've examined the simulation's sensitivity to many parameters (grid, turbulence model, inflow turbulence intensity, wall roughness) but varying those do not have a significant effect on the solution.

I expect RANS to provide a better estimate for a simple developing pipe flow. I have plenty of experience with CFD but not much with Ansys, so I'm wondering if this is simple user error - maybe there's an important setting I'm unaware of. Any thoughts? I'm hoping an experienced Fluent user might have an idea as I'm out of them.

Some additional relevant info: y+ ~ 1, k-omega SST turbulence model, adiabatic walls, solution is grid converged, density-based solver with default method settings.

Thanks.

[vinerm]

The boundary layer from the experiments appear to be too thick for air at normal temperature. Is it some other gas? Or is it that the temperature is higher? I suppose you have already checked if the flow rate is correct or not since the condition is far-field and there is no explicit condition on the flow rate. What's the Mach number?

[g7sznAc4]

It is air at close to room temperature, however the pipe is large - around 20 ft long. The flow rate is fairly accurate - CFD says 139 lbm/s, experiment measured ~140 lbm/s. The Mach number is around 0.5 at the station where total pressure is plotted.

[vinerm]

Then there are three things

1. If not yet done, enable Viscous Heating in Turbulence Model
2. Either use Sutherland law or other function for molecular viscosity; otherwise former point has no effect, of course
3. Use pressure-based solver. Though density-based solver can be used across whole range, pressure-based is recommended to be used for cases with M < 2.

[g7sznAc4]

Thanks for your help vinerm - I will try your suggestions.