MRF and gravity
 

Hi, is there any way/workaround to switch on the gravity not aligned with the axis of rotation? I simulate a bearing segment (1/6) with horizontal axis from which oil discharges to a drain (full model).

I assume you are running a transient simulation correct?

If so, I think it should work w/o any additional modification.

If your rotation axis does not align with the gravity vector, then you need to do transient analysis. This is because the problems are inherently instable. Unless your rotation is very strong. Then you can ignore gravity, just by switching it off.

Therefore, sorry, there is no workaround.

Thank you both a lot for reply. I am running a steady state. In the drain domain (full 360° domain), the rotation is not strong, therefore gravity is important. Nevertheless, thank you, at least it is clear I need to perform transient analysis or to create full domain of the bearing and cancel MRF. I hoped there is some workaround to switch on the gravity in one domain via subdomain or something like that.


I ve just tried to run transient analysis with transient rotor stator interface. The analysis does not run either. If I switch off buoyancy, it works. What are the restrictions for gravity and MRF? I cannot find it in ANSYS/CFX documentation.

Doesn't it tell you the restrictions for gravity and MRF when you try to run that combination?

Computation just suddenly crashes. The error is shown below:




COEFFICIENT LOOP ITERATION = 1 CPU SECONDS = 4.532E+02
----------------------------------------------------------------------
| Equation | Rate | RMS Res | Max Res | Linear Solution |
+----------------------+------+---------+---------+------------------+
| U-Mom-Bulk | 0.00 | 4.6E-04 | 4.1E-02 | 2.0E-03 OK|
| V-Mom-Bulk | 0.00 | 6.2E-04 | 9.8E-02 | 1.4E-03 OK|
| W-Mom-Bulk | 0.00 | 1.7E-04 | 1.8E-02 | 5.5E-03 OK|
| Mass-oil | 0.00 | 5.9E-05 | 5.9E-03 | 1.1E-02 ok|
| Mass-air | 0.00 | 4.2E-05 | 4.9E-03 | 42.5 3.9E-03 OK|
+----------------------+------+---------+---------+------------------+

+--------------------------------------------------------------------+
| An error has occurred in cfx5solve: |
| |
| The ANSYS CFX solver has terminated without writing a results |
| file. Command on host pc (PC-NamePC) exited with return code |
| 0.

I would have thought it would show an error message for that. Oh well, as previously by several people it is obvious that you cannot do a steady state model with MRF and a gravity vector (unless it is parallel to the rotation axis). So regardless of whether there is an error message or not, you cannot do it.

Yeah but I try to run it as transient.. The error is from transient case which does not work either.

Post your whole output file. Not in the query but as file (see "Go Advanced")

Please attached find the output file of the transient analysis including gravity and transient rotor stator interface. Any comments are appreciated.

You appear to have two domains connected by a transient rotor-stator GGI. But both domains are set as stationary. Are you sure that is correct? You do not appear to be using MRF at all, everything is in the stationary frame of reference.

Also, it looks you have set all physics 'on'. Don't try to hit the top in one go. That is not going to work and will be very frustrating. I would start as simple as possible, then get a reasonable result., increase complexity, show progress, etc. etc.
I don't know you case, but you have a significant pressure drop, high rotation speed and run with Total energy. I would reduce these and increase step by step.

Thank you both for your comments.

Yes there are two stationary domains connected via transient rotor stator. In the upstream domain (60deg segment), there are rotating walls. Recently, I analysed only this domain without downstream domain (full 360deg). Therefore the 60deg domain is still stationary to be consistent with the previous analyses. I expect this should not be any issue, however. But I will try to set the domain as rotating and set the non-rotating walls as counter rotating, thanks for suggestion.



As for the complexity, the settings was found as a result of sensitivity study on the 60deg domain only. Hence I would like to keep it the same also when adding the 360deg domain.

Why don't you share a picture (with BC's) of the case you are running. This could help a lot to understand your case.

I attached the figure. There are two domains:
- Segment domain: Several walls are rotating around Z axis.
- Full domain: all walls are stationary.



Axis of rotation is Z, I need to switch on gravity in -X direction.

If gravity is relevant, I don't think you can model it like this You need to model the segment over 360° as well.
If you are convinced you can model it like this, then gravity will be irrelevant and it can be ignored. Then don't include it.

Possibly, it is a mix. Then I would first do a steady state without gravity and then restart transient with gravity on. But in full 3D.

Thank you for comment. Gravity is important in the 360deg domain. Sure in the 60deg segment domain it is irrelevant, but in the 360def domain it is important to consider. Based on the first comments of Gert-Jan and Opaque, I understood that gravity should work in transient simulation but it does not. You suggest it should work if I create even the segment domain to be full 360deg. But there will still be the interface - therefore, based on the information from this discussion and from the internet, it is still unclear, if it does not work because of periodicity or the interface.

If I should model full model - both domains as 360deg - I would glue them together and include rotation of walls, hence there would be no interfaces. But this is something I would like to avoid and based on the comments it seemed that switching on transient solver only shall make the analysis run.

You might think you can model this using periodic boundaries, since you expect equal flow in all 6 segments, because of the high rotation, making gravity irrelevant. But, CFX won't let you since the periodic boundaries do not align with gravity at all.

Either:
1) run steady state without gravity, use periodic boundaries, and frozon rotor on the interface with a pitch of 60/360
2) run transient with everything full 360° and transient rotor stator interactions on on the interface and 'None' for the pitch.

Note: there seems to be an inlet, walls, and other things in the segment. Are these all rotating with the segment?

Thank you for the explanation with periodic boundaries and gravity vector. So, the periodic boundaries are the main culprit.


No, Inlet and Other do not rotate. Only several walls are rotating in the segment (simulating shaft).
I think separation of the model to two domains will be the solution. Assuming no significant upstream influence of the full domain, possible workaround is to export the velocity/temperature profile from the segment outlet (now the interface) from steady state, copy it 6 times and apply it to 360deg domain as an inlet and switch on gravity.

Why does this model require multiple domains? If all the domains are stationary then it could be modelled as a single domain. Then the GGI is not required.

Also - if the shaft rotation generates a tangential wall velocity only then you do not need multiple frames of reference to model this. It is simply a tangential velocity on the wall boundary of a stationary domain.

Thank you for comment, Glenn.



The upstream domain is periodic, therefore I model just a segment. Flow transfer between 60deg segment to 360deg domain can be done only by MRF or not? Is there another way to include a segment and to "copy" the velocity field to full 360deg domain?

You are talking about one big stationary domain without segment (zou assume 360deg to 360deg glued mesh - one domain) but I need 60deg to 360deg to reduce mesh elements.

You are not doing a MRF simulation, as everything is in the stationary frame of reference. That is a single frame of reference, not multiple.

But if you want to model only a 60 degree segment and have it mapped to a whole 360 ring then I think you need to do what you have done. (But it is not MRF, it is just using the GGI and periodicity to copy the 60 degree segment around).

Sorry for wrong term, yeah it is not MRF in fact. There is no easy way out. When using periodicity, gravity does not work. Setting gravity to selected domain only would help, it could be topic for CFX developers.

I think they will just recommend you to buy a larger computer and a full commercial license.

Can't you model the 60° segment alone, then export a boundary profile on the interface and use that in a separate simulation on the complete circumference with gravity? You only need to copy the profile 5 times to reposition it over the complete 360-interface.

Exactly. And they will release a new version with new ribbon.




This is exactly what I am gonna do - my comment from 11.1. 15:51.


Thank you all for discussion!

You can set gravity for a single domain (disable constant physics in the beta features), but you will be on your own.

Recall that pressure is shared across the interface; therefore, it is a tricky situation having different definitions of pressure on both sides, is it not?

Thank you for sharing this possibility. I tried it but it does not work. The computation crashes in the same way as when gravity is switched on in both domains (the same behaviour as shown in the output file I uploaded some days ago.