[Scaty]

Hi everybody,

I am a beginner in CFD and have a question about turbulence models in general.
As far as I know, I only have to switch on a turbulence model if I expect turbulence in my calculation. Indicator for this is the Reynolds number.
But where do I calculate it? As the velocity and the diameter change over a model, the Reynolds number changes as well.

Let’s imagine a vortex flow separator with one inlet and two outlets.
Do I simply calculate the number at the inlet (where I know the velocity as a boundary condition) and if I have turbulence there I must include a turbulence model?

And what happens if my flow at the inlet is laminar, but in the separator turbulence could possibly appear, but I cannot calculate the Reynolds number as I don’t know the velocity in the separator before running the numerical calculation (for which I must know if I need a turbulence model)?

And do I need a turbulence model if anywhere in my separator turbulence plays a role?

You see, a lot of basic questions that hinder me to progress with my solution.

I am happy for any explanation.
Thanks in advance!

Scaty

[NotDrJeff]

Hello Scaty,

I'm certainly no expert, but here's a few thoughts. (I also can't comment on vortex flow separators specifically since I don't know what they are )

1. Most flows that are encountered in reality are turbulent. Laminar flows are the exception not the rule. So be prepared to use a turbulent model in most situations.
   
2. When starting to simulate a new problem, you should always consult work that has already been done. This should give you an idea of whether a flow is turbulent. (Also remember that the "critical" Reynolds number is different depending on the [arbitrary] length and velocity scales chosen, so you need to compare with existing work). Previous work is also the starting point for choosing which turbulence model to use.
   
3. If you really don't know whether a flow is turbulent or not, remember you can use trial and error. Using the values at the inlet can give you an initial guess. If you think the flow might be laminar, then perform a laminar simulation and compare your results with experiment (CFD ideally always needs some verification against experiment). If they don't match, then could this suggest that the flow transitions to turbulence somewhere in the domain.

[LuckyTran]

Originally Posted by NotDrJeff

1. Most flows that are encountered in reality are turbulent. Laminar flows are the exception not the rule. So be prepared to use a turbulent model in most situations.

Sorry but this is akin to saying that the most common fluids are air and water when the universe is mostly composed of hydrogen gas, a much lesser amont of helium gas, and some other rarely occurring elements. If you ask an astrophysicist, they'll tell you that most flows in the universe are inviscid.

[NotDrJeff]

Originally Posted by LuckyTran

Sorry but this is akin to saying that the most common fluids are air and water when the universe is mostly composed of hydrogen gas, a much lesser amont of helium gas, and some other rarely occurring elements. If you ask an astrophysicist, they'll tell you that most flows in the universe are inviscid.

Yeah, you're absolutely right. My bad! I'm a mechanical engineer by trade so that's the perspective I was coming from.

[NotDrJeff]

Originally Posted by LuckyTran

If you ask an astrophysicist, they'll tell you that most flows in the universe are inviscid.

Still thinking about this and wanted to clarify. I'm still trying to figure all this stuff out myself, so please correct me if I'm wrong!

"Inviscid" does NOT mean laminar. Inviscid flow means the viscous forces are negligible compared to inertial (and other) forces. This is actually a definition of turbulent flow! Of course, it is possible to have inviscid flow that is not turbulent in certain circumstances. This is called potential flow (I think).

I'm not an astrophysicist, so I can't comment on whether the potential flow equations are useful in those scenarios. I do know that potential flow can be used in (for example) bulk flow calculations of flow around bodies , but this needs to be combined with turbulent boundary layer theory at the surface of the body.

Here's a quote I found in Tennekes and Lumley, A First Course in Turbulence (1972) p. 1:

"Most flows occurring in nature and in engineering applications are turbulent...The photospheres of... stars are in turbulent motion; interstellar gas clouds (gaseous nebulae) are turbulent; the wake of the earth in the solar wind is presumably a turbulent wake... In fluid dynamics laminar flow is the exception, not the rule: one must have small dimensions and high viscosities to encounter laminar flow."

[agd]

Referring back to the original question, unless you are writing your own solver then you are going to be limited in what you can actually do. If the turbulence model you are using includes a tripping function the you can try to rely on this to get turbulence started in the flow field correctly - but most turbulence tripping functions are not very reliable for general problems since they are typically tuned for a particular set of conditions. If you have a solver that includes a transition model then you may have more luck in capturing mixed flow behavior (laminar/turbulent in the same flow field). In the old days the basic rule of thumb was that if most of the flowfield was turbulent then run the problem as turbulent, since there were few ways to accurately determine transition.


You can always try running your problem fully turbulent and then fully laminar to see what differences there are.


Compute your Reynolds number based on the relevant physical length scale for your problem. It sounds like for your problem that is either going to be the inlet diameter or perhaps something to do with the vortex separator geometry.



If you don't know the velocity at the inlet how are you going to set your boundary and initial conditions?

[FMDenaro]

One of the misunderstanding is about the direct link of turbulence and a certain Reynolds number.

The very simple case of the flow over a flat plate show you have not a unique value of the Re number. Turbulence develops after a certain position is reached.

Consequently, one can generally think about different cases, a fully developed turbulence (that is since from the inflow) or a developing flow where turbulence is generated by something to be described in the computationbal domain. The formulations used to solve such problems are different, not in the turbulence model but in the proper equations (and variables) to solve.
In any case, the Re number depends on the scales you are referring.

[FMDenaro]

Originally Posted by NotDrJeff

Still thinking about this and wanted to clarify. I'm still trying to figure all this stuff out myself, so please correct me if I'm wrong!

"Inviscid" does NOT mean laminar. Inviscid flow means the viscous forces are negligible compared to inertial (and other) forces. This is actually a definition of turbulent flow! Of course, it is possible to have inviscid flow that is not turbulent in certain circumstances. This is called potential flow (I think).

I'm not an astrophysicist, so I can't comment on whether the potential flow equations are useful in those scenarios. I do know that potential flow can be used in (for example) bulk flow calculations of flow around bodies , but this needs to be combined with turbulent boundary layer theory at the surface of the body.

Here's a quote I found in Tennekes and Lumley, A First Course in Turbulence (1972) p. 1:

Turbulence in inviscid flows is indeed studied. It is more a theoretical framework to study the behavior of some solutions and the inviscid energy transfer.

[Scaty]

Thanks everybody for your answers!
I got it solved to a reasonable level of precision.