[Far]

Dear All

I would like to know that the mixing plane model available in CFX (also in Fluent) , it is though good enough for rotor stator system, is also valid for contra rotating rotors (one in +ve direction and other one in -ve direction). Will I have to change some sources terms in mixing plane model to make it work for this type of simulation?

You guidance will be highly appreciated.

PS. Few years back I had also tried this on coupled simulation of high pressure (+ve rotation) and low pressure turbine (-ve rotation), but failed to get the reasonable solution. Although when solved separately, I got good results.

Thanks.

[ghorrocks]

Whether the mixing plane model is appropriate has little to do with whether the blades contra-rotate. The decision is based on whether the inherent assumption of the model, that is that the interface can be averaged to give an even flow to each stage, is suitable for your model.

BTW: In CFX it is called a stage model.

[Far]

Traditionaly the purpose of mixing plane, frozen rotor and transient methods is to connect the rotating reference frame to stationary reference frame. And these models were designed around this requirement.

The upstream part of the two domains (stationary or rotating) needs only one boundary condition at the connection of the two domains. That is the static pressure. The other boundary conditions are extrapolated from the calculation domain

The downstream part of the two domains needs four boundary conditions and information about the turbulence quantities. The circumferentially averaged quantities are the total pressure, total temperature, flow angles, and turbulence quantities. The absolute value of the velocity vector is extrapolated from the upstream domain interface. And this is problem, because these conditions are set from the point of view of rotating or stationary reference frame and not designed keeping in mind the contra rotating rotors (i.e. two rotating domains).

Therefore question is that
1. this model is valid for contra rotating rotors?

2. if not valid, then can we model this case by giving the one rotor as speed of rpm1+rpm2 and other rotor as stationary?

Reference 1 : http://www.cfdthermo.tkk.fi/Teaching...pilapublic.pdf

Reference 2 : http://www.turbostream-cfd.com/media...-asme-2009.pdf

Page : 5 Heading : Multistage and unsteady simulations

For steady-state calculations, multistage simulations are made possible by circumferentially averaging the flow entering a blade row, see Denton and Singh [14], Dawes [15], Denton [11]. The technique is applied at a mixing plane, the position of which is arbitrarily chosen by the user. The non-uniformflow upstream of this plane is mixed out so that it becomes pitchwise, but not spanwise, uniform downstream of the mixing plane. The process is conservative, but, like anymixing process, it is irreversible (see Fritsch and Giles [16]). In addition, care must be taken to avoid locating the mixing plane too close to leading or trailing edges, and thereby enforcing a non-physical circumferentially uniform flow.

Reference 3: http://books.google.com.pk/books?id=...denton&f=false

[ghorrocks]

I cannot see why it is not valid for joining two rotating domains - but I am no expert in the area and have not studied the equations in detail.

I can only recommend you try it using a benchmark case and check it is OK.

But you definitely cannot get around it by putting one domain as rpm1+rpm2. The whole point of a rotating frame of reference is that rotational terms are added (centipetal, coriolis) and if you put int he wrong rotation speed these terms will be wrong.

[csmistry]

The concept for stage( Mixing plane) and frozen rotors are truly giving different results... but my experience suggest for contra rotating fan concept better to adopt frozen rotor option.
1. In mixing plane approach it was found the data from domain-1 is not truly (Fully) transfer to domain-2. which can also be cross checked my total pressure contour before mixing plane and after mixing plane( joining plane).

2. For rozen rotor this is not the case the data from domain-1 is fully transfer to domain-2 (Considering circumferential averaging).

3. The flow physics i feel what CFX is not capturing is " THE DATA TRANSFERS FROM DOMAIN-1 TO DOMAIN-2 BUT THE TRANSFER OF FLOW PARAMETERS FROM DOMAIN-2 TO DOMAIN-1 IS NOT POSSIBLE"

can you please explain the reason behind?

[Far]

But stage (mixing plane) plane model does reduce the problem to steady state with single passages from both components. Which is clearly not the option in frozen rotor, where for each relative position you have different results. Therefore you have to use the transient sliding mesh method where you need the equal pitch for both components and which is not always possible due to combination of odd (prime) and even no of blades in turbomachinery.

However new methods can overcome this situation.

1.Phase Shifted Periodic Boundary Conditions - Time Transformation Method

2.Phase Shifted Periodic Boundary Conditions - Fourier Transformation Method

[Far]

Originally Posted by csmistry

BUT THE TRANSFER OF FLOW PARAMETERS FROM DOMAIN-2 TO DOMAIN-1 IS NOT POSSIBLE"

can you please explain the reason behind?

from domain 2 to domain 1 only thing needed is the static pressure, which is independent of reference frame and hence can be easily applied as downstream boundary condition for upstream component.

[csmistry]

Yes it is true but for contra rotating case with frozen rotor you need to go through trial and error for pitch setting. for particular pitch setting there is not much variation in total pressure rise was found at the exit of rotor-2.

[D.B]

Hi Far,
I know this is not your intended solution but still I will try. Is it possible to introduce a very small stationary domain between the two rotating domains. In this way no false information is transferred from one rotating domain to another.
Also I feel that transferring flow from one RPM R1 to another at -R2 RPM might introduce some numerical issues and errors might go up. In this case amongst the available models introducing a small stationary domain might be a good idea. But this is all based on speculation and only actual simulations will give a good idea as to which is better.

[Far]

It is the practical idea. But one needs to compare it to full 3D transient simulation to quantify the errors.