[knixxor]

Hey guys,

I'm struggeling with the implementation of humid air as a fluid to simulate a 3 stage low pressure compressor. I have used air ideal gas so far, but to get more accurate results and to continue my studies it is necessary to change ideal gas to real gas.

Unfortunately, I feel a bit overloaded since there is a large amount of different materials, models and parameters available on ANSYS CFX.

Could one of you please answer the following questions? I would be rather happy

1. What is the general approach to simulate humid air, in order to set up a certain level of rel. humidity?

2. How has Ansys implemented "humid air" as a fluid into its code?

3. Which kind of multiphase model I have to use for humid air? (or is it even a multiphase flow?)

I hope you can follow my questions

Knixxor

[ghorrocks]

1) Use a multicomponent fluid. Air as one component, water vapour as the other.
2) It hasn't. But multicomponent fluids cover the required physics so no special model is required.
3) It is not a multiphase flow. All components are gaseous so there is only a single phase.

I would be rather happy

I hope those answers provide the required amount of happiness

[knixxor]

Thank you very much ghorrocks!

[knixxor]

I have tried several combinations for the multicomponent fluid, but with no success. I get either an error message right at the beginning or an overflow error after some timesteps.

Could anyone of you have a look on my configuration and give me some tips?

general properties

Mesh with sufficient density (grid sensivity test already performed)
Turbulence model: SST
Heat Transfer: Total Energy (incl. viscous work term selected)
Pressure based inlet/outlet conditions

Fluid properties

Mixture Option: Fixed composition mixture
Child 1: Air ideal gas
Child 2: Water vapour

Thermodynamic state: gas
mixture properties: ideal mixture

mass fraction
Air:0.996
Water: 0.004
(round about 40% rel. humidity @ 1013 hPa. and 15 deg. C)

For this configuration I always get an error message after 140-160 (overflow error, Mass residual rate ist > 99.9)

Then I changed the mass fraction (60% and 80% rel. H and even 0% (which means the mass fraction of water is 0.0) but with the same result.

Other changes (with the same output > error) I tried:
- changing air from ideal gas to Air at 25C
- changing water vapour to water ideal gas and then forced it to be in gas state
- changing reference point values (according to booster inlet)

I'm running out of ideas and I don't understand the behaviour of the solver for multicomponent fluids. Can you help me with this?

Best regards
Jens

[ghorrocks]

Does your simulation converge nicely when you run a single component fluid, maybe air ideal gas?

Overall I would recommend looking at the standard issues when you have numerical stability problems: double precision solver, smaller time step, better initial conditions, improve mesh quality. Do these simple things before you consider anything more complex.

Also, unless you have viscous work doing soemthing significant then turn it off. If you don't need the model then you don't need to activate the option. Having said that this option is pretty benign and I have never known it to cause problems, but as a general rule you should only use the physical models necessary in the simulation.

[knixxor]

Hey, thanks for your reply.

When using a single component fluid, everything works fine. I get a converged solution (RMS residual of P-Mass < 1e-05, U,V,W-Velo. < 1e-04). Imho the mesh is fine enough, it's round about 700.000 elements per passage, which means 5 milion elements in total.

I also took care about the numerics:

pysical timestep: 0.1/w (w= rotational speed)
High resolution advection scheme
first order turbulence numerics
...

As I said, I'm running out of options and I really do not understand the solver's behaviour...

I will try to simulate the same conditions without the visc. work term, thank you for this tip!!

Kind regards
Jens

[ghorrocks]

As I said in the last post, there are many things to do to respond to convergence difficulties:

Overall I would recommend looking at the standard issues when you have numerical stability problems: double precision solver, smaller time step, better initial conditions, improve mesh quality. Do these simple things before you consider anything more complex.

And just because your time step and mesh is OK for simple ideal gas models does not mean it is OK for more complex physics. You may need a finer time step or better mesh quality to run the more complex model.