[TurBoris]

Hi Everyone,

I want to simulate inlet distortion problem for a multi-stage compressor. This requires to run full annulus URANS simulation. However, this needs very high computation time. That is why, I want to model the first stage with full annulus URANS using Transient-Rotor Stator interface and steady-state approach for the further stages with periodic BC's and mixing plane interfaces. Is CFX able to run such a configuration ?

Thank you in advance.

[Gert-Jan]

You are mixing things up.

In a single simulation it is not possible to run Transient and steady state simultaneously.

It is possible to use Transient Rotor Stator interactions and for the first interface and mixing plane (with periodic BC's) for the others. But that is a full transient calcuation and has nothing to do with Steady State.

[Opaque]

It seems you have already concluded the influence of the distortion does not travel beyond the first stage to make an impact downstream; therefore, you can cut your model to solve the first stage, or where you feel is the most appropriate.

To do this, I would (not exact science here) run the steady-state case, and export the pressure on the upstream side of the mixing plane where you plan to cut your model. Recall you can export the exact profile, or one circumferentially/pitch averaged, your call.

Then, set up the new transient model with the stages you plan to keep, and use the exported profile for the outlet boundary condition. Your model is now transient and hopefully smaller and, you can capture the influence of the distortion in the first stage.

Hope the above helps

[TurBoris]

Originally Posted by Gert-Jan

You are mixing things up.

In a single simulation it is not possible to run Transient and steady state simultaneously.

It is possible to use Transient Rotor Stator interactions and for the first interface and mixing plane (with periodic BC's) for the others. But that is a full transient calcuation and has nothing to do with Steady State.

That is actually what I wanted to express above. I mean a periodic domain like in the steady-state analysis for the further stages and coupled it with a mixing plane interface to the full annulus first stage domain. No doubt that transient solver will work for the second stage as well. Thank you.

[TurBoris]

Originally Posted by Opaque

It seems you have already concluded the influence of the distortion does not travel beyond the first stage to make an impact downstream; therefore, you can cut your model to solve the first stage, or where you feel is the most appropriate.

To do this, I would (not exact science here) run the steady-state case, and export the pressure on the upstream side of the mixing plane where you plan to cut your model. Recall you can export the exact profile, or one circumferentially/pitch averaged, your call.

Then, set up the new transient model with the stages you plan to keep, and use the exported profile for the outlet boundary condition. Your model is now transient and hopefully smaller and, you can capture the influence of the distortion in the first stage.

Hope the above helps

Steady-state solutions with inlet distortion will yield in non-realistic flow field unfortunately. Thus, the flow variables that I will obtain at mixing plane cannot be used for the transient case.

For stage matching effects I need to include the further stages in the simulation but the main effects of distortion will be sensed at first stage. Normally the flow at the outlet of the first stage will not be uniform in circumferential direction due to the distortion, however these are going to be averaged by mixing plane and transferred to further stages. This averaging is the penalty of this model. Thank you.

[Opaque]

If you are running transient already, there is no point to usage mixing plane downstream. What will you gain? A tiny speed increase because of the sliding mesh work savings?

Either you go full-blown transient to get the maximum value of the expensive simulation, or reduce the model in a reasonable way. My suggestion is just one of them. It all depends on the level of fidelity you are looking for.

Once the flow is mixed up after the first stage, all the distortion influence is eliminated downstream. So the only value of the downstream components is to provide you with a pressure level for the outlet of the first stage, is it not? How far will that pressure level be from the steady solution?

Effectively, the downstream component becomes boundary condition of the third kind, i.e. mass flow_outlet = Coefficient * (P_outlet - P_oo). Sure the coefficient can be very complex function of machine parameters

My 2 cents