Unsteady rotating mesh moment calculation

[IvorCecelja]

I am currently performing a simulation of a rotor in edgewise flow with an unsteady SA solver. My domain looks like the following:

So an inner cylinder that creates the rotating motion of the rotor and an outside "bullet" from which I apply edgewise flow on the curved surface. My rotor spins around the negative z axis and to better understand the moments I attached a close up image.

I extract the moments in x,y and z from the history file and notice that one of my moments oscillates around 0. This shouldn't be the case. I dug a bit further and plotted the forces in x, y and z as well as each moment per blade.

This is where I found out that one of the hub moments is behaving weirdly. Instead of being dominated by the Fz (thrust) force which has the largest magnitude and largest moment arm, it follows the shape of its own force. I attached the graphs of the moments below and their forces.

The results in the picture are not fully converged but the moment do not change much and the issue persists.

Desktop (please complete the following information):

• OS: [e.g., Linux (UBUNTU 24.04 LTS)
• C++ compiler and version: [e.g., g++ (GCC) 4.8.5?]
• MPI implementation and version: [e.g., OpenMPI 3.0.0?]
• SU2 Version: [e.g., v8.3.0]

[IvorCecelja]

I ensured the origin is at 0, 0, 0 and made sure this is also the case in configuration files. The rotation rate is defined correctly too. I ran with just inner iterations, just outer and a combination but the issue persists with the moments. If anyone could also advise on what should be used for an unsteady multizone problem in terms of inner/outer iterations that would be great.

[rois1995]

Could you please provide more explanations on what you are expecting vs what you obtain? Please also provide an image of the geometry.

[IvorCecelja]

First of all, thank you for getting back to me as I have been busting my head for a few weeks on how to resolve this.

STARCCM+ gives higher peaks and the moment waves are shifted by around 15 degrees to the right. Additionally, my biggest issue is that when time averaged, specifically the My moment is around 50Nm while it should be 150Nm. The Mx moment is quite good when time averaged though. I should also mention, I used other software from low to high fidelity which tend to go towards the STARCCM+ results. I tried manually calculating forces and moments in paraview and they match with SU2's output. So something is off in either my setup or SU2's solver.

I tried increasing my rotating domain to have a smoother transition across the interface and that the main flow physics stay inside the rotating domain but that only improved the My moment slightly.

The geometry is a standard rotor. I included pictures of how the domains look. The flow goes in positive x and the rotor spins in negative z direction.

I would also like to add a description of my config files:

In the subconfig files I use RIGID_MOTION for both and give the cylinder a rotation, while the bullet has no rotation. Additionally, I define my propeller as MARKER_HEATFLUX and the interface as MARKER_FLUID_INTERFACE, while the bullet has an inlet/outlet/farfield markers.

In the main config file I use URANS SA with a timestep (dual time stepping second order) of 1E-4 equating to 1.108 degree of rotation. I am unsure on what I should use for inner and outer iterations so I ended up using 20 inner and 1 outer but later switched to 10 outer and 12 inner iterations. Neither of which solved the problem. I have some the basic solver and setup up definitions. For the boundary conditions I have the MARKER_ZONE_INTERFACE and the MARKER_FLUID_INTERFACE. CFL number is at 2 and I use the adapting feature with EULER_IMPLICIT time discretisation.

[rois1995]

From your plots, the average My on the hub between 0 and 180 deg is 0 Nm, whereas the Mx is -140 Nm. I have a couple of questions:

1- How long is the window of your time averaging? Are you performing it over one whole revolution (could also be on just half of it)?
2- Are you using the same turbulence model in your StarCCM computations?
3- Have you tried reducing the time step?
4- Are you sure that the reference frame is the same between all the different codes?
5- I guess you are using the sliding mesh interface. What is the interpolation procedure at the sliding boundary?

[IvorCecelja]

1. My window of averaging is over 1 full revolution. I have done so after simulating 5 revolutions and the My and Mx moments barely change after the 1st revolution so It isn't the case of not being fully converged yet.
2. Yes I am.
3. I tried reducing the time step but it didnt change much. My time step is already lower compared to STARCCM+ at 0.0001s equating to 1.108 degrees per step.
4. The STARCCM+ code uses the stationary reference frame while I am not sure what SU2 uses. I tried assuming it uses the rotational reference frame and converting the graphs to a stationary frame but this gave weird results that aren't sinusoidal.
5. I used rigid_motion so I believe that is the sliding mesh interface. I didn't setup a interpolation procedure but I based on the output of the code, it calls the nearest neighbour by default.

[rois1995]

Sorry for the late reply. Unfortunately, I am not able to help you other than by suggesting refining the grid and seeing what happens.