Hi,

I want to model combustion with acoustics. Is this possible?

Regards,

Cujo

Combustion is possible. Acoustics not directly, as far as I know, unless you can make the link between flow and turbulence with acoustics yourself (although I think you are looking for that link).

Astrid

The only thing you have to do is add a relationship between density and pressure, I guess? When using a single fluid, one can select "General Fluid" and implement this relationship and the speed of sound. But when I use combustion, this seems impossible, all combustion species are ideal gasses. Can I change them to "General Fluids"?

Regards,

Cujo

Why can't you just do transient combustion calculations with the ideal gas equation of state? This supports acoustics.

The real problem is how to translate flow pressure variations into sound pressure levels is it not?

Neale

Hi Cujo,

Firstly, Ideal gas does give you a relationship between pressure and density. If you want to use a different relationship, then perhaps you would enter an expression for this in the General Fluid form.

Actually simulating acoustic phenomenon is beyond the capability of any CFD code I am aware of. Accurate transient simulation of pressure fluctuations requires very high order discretization schemes (6th order schemes are used in acoustic analysis codes), which is far more than is required to accurately simulate fluid flow. Not to mention that you would require an extremely small timestep to resolve the pressure fluctuations, resulting in an extremely long simulation for the combustion phenomena.

Regards, Robin

Hi,

It is not so unrealistic as Robin thinks. Simply sound is generated by flame fluctuations, so you need to descibe those fluctuations and find a relationship with sound generated. No need of compressibility in this case. Much larger source of sound comes from pressure waves amplified by flame, but it could be also handled. Sad true is, that you will not get any feedback from pressure waves on flame, as all those models are postprocessing cfd results.

To have pressure wave -flame interaction (different instabilities), that is bit different story. The best idea is to have moving grid, to resolve flame zone. However still very expensive calculations, or 1 or 2d simple runs. But i think possible even in CFX5.5.1, with comb.scalars desribed in command lang.

Matej

As soon as density is a function of pressure then d(rho)/dp_s is defined, and the code should support acoustics.

There is no need for 6th order advection discretisation to do this accurately. You do need an advection scheme that minimises phase errors well though. This could be done with a 2nd order spatially accurate scheme. The NAC scheme in CFX-5 has fairly decent phase error properties, so it should do alright.

More important than the spatial accuracy of your advection scheme, as you have pointed out, is what timestep you run. You need to run a timestep which can accurately resolve the sound pressure wave propagation. In the end this means courant numbers on the order of unity, which, as you mention, could get quite expensive.

The original poster still needs to figure out how to translate pressure fluctuations in to sound pressure levels though.

Neale

Why can't you get feedback from pressure waves into the flame? If sound pressure waves are interacting/bouncing around off the geometry and so on, then presumably these could also interact with the flame itself and possibly amplify it.

Neale.

Neale,

Of course you can have interaction of pressure with flame, but not with the way of acoustics modelling I reffered to. In this model, you have only "weak commpresible" fluid, and you model acoustics generated or amplified by flame. It's only model, no real modelling of acoustics.

When you have fully compressible fluid, mesh fine enough to describe flame zone, time step according to Courant criterium, fast and big cluster, enough time, you can have anything )

Matej

I just returned from the European CFX conference in Strassbourg. I was told that acoustics are planned for CFX 5.7 which will be released in the 4th quarter of 2003.

Astrid

Hi Neale,

Sound waves are not being advected through the domain, so the advection scheme really has little effect, it is the Pressure Diffusion equation which needs to be of higher order.

As for what is coming in future releases, it is doubtfull sound waves will be modelled directly (for the reasons already stated). Rather, the approach is to relate the turbulent and transient flow spectrum to sources of sound.

Robin

Sound waves are a hyperbolic (wave based), not diffusive, process. Pressure diffusion has nothing to do with acoustics either, and is not necessary. i.e., the Euler equations support acoustics with ideal gas equation of state. There are no diffusion terms in the Euler equations. Diffusion only acts to "damp" acoustic processes, it is not responsible for their propagation speed, or amplification.

I agree, simulating pressure variations directly is costly hence "fudge factor" correlations are used to relate sound pressure level to turbulence fluctuation/intensity levels, which is at best ad hoc.

Neale