[CFD] Multiple Reference Frame (MRF) Approach for Turbomachinery
HTML-код
- Опубликовано: 6 июл 2024
- A brief introduction to Multiple Reference Frame (MRF) approach, which is used to simulate the motion of rotors/impellers and blades. The following topics are covered:
1) 10:56 What additional source terms are added in the Single Reference Frame (SRF) approach?
3) 17:34 What additional source terms are added in the Multiple Reference Frame (MRF) approach?
4) 19:22 How is the face volume flux modified in the MRF approach?
Frozen Rotor and Mixing Plane approaches will be considered in a separate video.
==================================
Derivation of the SRF and MRF equations:
1) OpenFOAM Wiki
Derivation of the SRF and MRF equations
openfoamwiki.net/index.php/Se...
============================================
Want to learn more?
============================================
Grab a copy of my CFD Fundamentals Course (for beginners):
www.udemy.com/course/computat...
Learn how to write your own CFD code in MATLAB and python (for intermediates):
dr-aidan-wimshurst-s-school.t...
Learn how I draw my figures and diagrams in Inkscape (for everyone):
dr-aidan-wimshurst-s-school.t...
============================================
Did you like the video?
============================================
Download the lecture slides from my website:
www.fluidmechanics101.com/pag...
Buy me a coffee to say thanks:
www.buymeacoffee.com/NKEZrhvg0
Support the channel on Patreon (and get useful extras for your CFD studies):
/ fluidmechanics101
============================================
Donations:
============================================
1) PayPal
www.paypal.me/fluidmechanics101
==================================
Disclaimer
==================================
The methods, algorithms, equations, formulae, diagrams and explanations in this talk are for educational and demonstrative purposes only. They should never be used to analyse, design, accredit or validate real scientific / engineering / mathematical structures and flow systems. For such applications, appropriate trained, qualified and accredited (SQEP) engineers / scientists should be consulted along with the appropriate documentation, procedures and engineering standards. Furthermore, the information contained within this talk has not been verified, peer reviewed or checked in any way and is likely to contain several errors. It is therefore not appropriate to use this talk itself (or any of the algorithms, equations, formulae, diagrams and explanations contained within this talk) as an academic or technical reference. The reader should consult the original references and follow the verification and validation processes adopted by your company / institution when carrying out engineering calculations and analyses. Fluid Mechanics 101 and Dr. Aidan Wimshurst are not accountable or liable in any form for the use or misuse of the information contained in this talk beyond the specific educational and demonstrative purposes for which it was intended. 
Наука
![[CFD] Aspect Ratio Warnings in CFD](http://i.ytimg.com/vi/oa3sxesYsKc/mqdefault.jpg)
![[CFD] Aspect Ratio Warnings in CFD](/img/tr.png)
![[CFD] Hexcore Meshes for CFD](http://i.ytimg.com/vi/59xfM0ayFE0/mqdefault.jpg)
Aidan, your skill of delivering a topic is absolutely genius. Very few can do this. Thanks a lot again.
Another beautifully explained video, demystifying the complicated domain of CFD. Thank you.
You have done a great job at demystifying a lot of clutter about this topic for me. I can't thank you enough.
Really enjoys every moment i listen to your explanation.. Thanks Doctor.
One of the best CFD masters
Wonderful,
What a lucid way of presenting a topic in simple, detailed and streamlined way.
Thank you so much.
Great video as always Aiden. Thanks a lot!
Definitely, videos on mixing plane and on frozen rotor would be much appreciated.
This video explained the difference so perfectly. Thank you
Great video Aiden So Happy that you treated this subject very interesting .looking forward for your next one. Thanks a lot
Incredible quality. Thank you!
Amazing video. I am so happy with the video quality and the content. Very informative and easy to understand. Keep up the good work 🙂
Looking for this part from time.
Thanks
A fantastic talk, thank you. I have learned a lot form the talk.
Great lesson, Dr Aidan!
I would love a video talking about the Mixing Plane approach
Thanks dr for bringing helpful lessons 🙂🙂
Thanks Aidan, nice explanation😄
this exactly what I need, thanks.
nicely made video and lesson. well done
GREAT VIDEO.... THANKS A LOT!
nice explanation ...thank you
its very interesting since i'm working on turbomachinery simulation شكرا
Brilliant!
Nice job! I really appreciated this video. It would be great if you could explain more in detail the Frozen Rotor approach, thanks!
Yes, i will hopefully get round to this video soon. The frozen rotor and mixing plane are quite tricky, as it all depends on how you treat the variables as they cross the interface between the stationary and rotating domain. Coming soon!
Thanks, Dr Aiden, It would be also helpful to talk about the limitations of the MRF and perhaps point out some of the questions asked when applying such approaches, for example, at what distance one should place that Moving region around the blade to get the optimum accuracy? which is difficult to answer when taking the mixing tank example, in particular for the ones with baffles.
same question for what distance and size of moving region to get optimum accuracy
Great question. There is obviously an optimum between accuracy and speed of calculation. Usually I make the diameter of the 'coin' about 1.3 times the diameter of the rotor itself. The lateral length will depend slightly on what the structure of your wake looks like and whether you need to resolve it. This probably requires some testing 👍
I would love a mixing plane explanation! I am modelling an axial flow stator/ rotor problem for my 4th year project
Thank you.
Very very appreciated! thank you. I am waiting for ALE formulation 😎😎😎😎😎
But for the Sliding mesh section, I guess the NS equations are written in different formulation to account for the sliding region? Could you please point me to the NS formulation for Sliding mesh?
Thank you for the talk. I need more information about Frozen rotor and mixing plane approches. Can you help me to find any ressources ? Thank you
can you make a video explaining sliding mesh to study the rotating heat transfer pipe? thank you
Excellent explanation for the concept....Thank you so much for effort 😁
Can we get Frozen rotor concept ???
Thanks, Dr Aiden，great work of you. Let me ask you something please. I model a propeller based on the slidingMesh, using the fieldAverage functionObject to get the uPrime2Mean(so as to get the TKE), apparent value jump appears in the AMI interfaces. This maybe due to that , fieldAverage is based on the rotating cell, not on the fixed spatial point, so the Uprime2Mean cannot be used in the rotationg region. The question is: how can I get the right TKE(on a cuttingPlane or the whole region) in the rotating region? Do you have any references to do so?
Thank you for this video, it is just at the right time! How to choose the size (diameter) of the MRF or Sliding mesh zone around a rotating object?
depends on your problem
Great video. I am trying to use SRF with Fluent but when post-processing I am getting a swirling motion usptream of my impeller as well. I don't know how to "filter it out" so I am using the MRF method (basically splitting my domain where my impeller is).
It might be because you need to apply the inlet boundary conditions in the rotating reference frame when you use the SRF. But yes, just go with the MRF, that will do the trick 👍
hi,
Like I said in the last comment, I am implemented the MRF in my own solver, I found that if we divided the convention term in equation of relative frame but solving the absolute velocity will cause some problem even we do some flux correction when I choose the upwind scheme for the stability reason. The term /nabal U_I U_I makes me confused to use U_I as the judgement of the upwind direction, correctly, the it should be the U_r as the judgement of the upwind direction in the MRF zone.
The aim of this commit is just for fun and discussion , u do a very good job for the the vedio. I really
appreciatE that.
jialiang.zhou
Perfect! I wish to also point out how the slides are full of information while being visually clear and full of space. Beamer helps but the figures are on point.
Stupid thoughts looking at this presentation:
1a) Simple pipe, I solve it with the SRF approach.
1b) Same pipe, I solve it stationary, with rotating walls.
Are the results the exact same? I'd say yes, but never tested.
Would the conclusion be the same in case of the MRF approach? (a portion of the pipe solved with MRF approach in case a), and with rotating walls in case b) )
Hi, Your Videos have nice explanations. Could you please make a video on Multigrid method. Thank you.
Thanks Aiden for the video. It is quite clear to follow. I am trying to use the same approach in generating a MRF in Ansys CFX to simulate flow over a wind turbine blade, just the same model you used for the explanations. However, I am a beginner in CFX and was wondering if you have tutorial videos that covers generating a MRF frame for simulating flow of a wind turbine blade?
Thanks
This is reasonably straightforward to do in ANSYS CFX. I think there is a tutorial in the manual that you can follow? The main thing to do is have a domain that contains the wind turbine blade. Then you need to specify the domain as rotating in CFX PRE. When you set the rotational speed, CFX will calculate and apply the source terms in that domain (giving an MRF approach)
please do video on mixing plane and frozen rotor
Need help in understanding the need of this source term for Volume flux corrections on the faces of the cell.
(Thanks.)
Thanks for the great explanation. Would this be equivalent to adding a source boundary condition to the leading edges of the blade, and adding a sink boundary condition to the trailing edges? Would that be a valid approximation of blade motion?
Yes, actually that is a nice way of thinking about it! (With the source slightly upstream of the blade, where the boundary of the coin region is)
Excellent lecture Aidan! I love the drawings, are you using Tikz? Thanks very much for sharing.
Hi Martin, i find it easier to do all my drawings in inkscape and save the image as a pdf (then import it as a figure into latex). I always found TikZ a bit too tricky for me 😅
@@fluidmechanics101 The quality you obtain is great; I think you do not need Tikz. Your lectures are excellent and the pictures help a lot to understand. Thank you !
Hello, amazing video, i have a question, for example we have rotating domain around a rotor and we make frame motion for this domain. We also have fluid with the same material (for example air) around the rotating domain. So we have fluid-fluid interface for the rotating domain and the fluid domain around it, how to define the interface so the interface not act as a wall (so the rotating fluid can flow through the interface and we have continuous velocity profile at the interface)? Thank you
The name of the interface will depend on the software you are using. It may be called 'arbitrary mesh interface (AMI), CGI or just 'interface'. OpenFOAM for example has changed the name of its mesh interface algorithm a few times, so worth checking the user manual 👍
hi sir first i'd like to thank u for this explanation ,and i want to ask u how can simulate a tesla turbine machine with ansys so can u help me
Hi Dr. Wimshurst, thanks for the great video. I’m working on a ducted turbine CFD simulation using MRF method and I have a question about a size of the MRF zone.
There is quite a difference in terms of a generated torque between setting the MRF zone only around the rotor (inside the duct) and setting a bigger zone including the duct (non-rotating patch for the duct). It seems that the smaller MRF zone loses the wake information behind the rotor and inside the duct, affecting the degree of interaction between the duct and the rotor.
I wonder how the information is transferred across the interface and what MRF zone size would be reasonable. I’ve been searching for the answer but there was no clear answer. It would be really helpful if you share your insight and knowledge.
Yes, I found something similar. As the wake flow passes out of the MRF zone across the downstream face, the wake structure is smeared and loses a lot of its structure. So your observation is consistent with mine. What I would do is use the MRF solution as an initial condition for a transient simulation that physically rotates the MRF zone (a sliding mesh approach). You will probably get a converged solution within 4 - 5 rotor rotations (depending on your tip speed ratio) and this solution will be a lot better. As long as the mesh transition across the MRF interface is not too drastic and your downstream mesh is not too coarse you should get a pretty good downstream mesh structure that you are interested in.
@@fluidmechanics101 Thanks for the answer! Have you tried a big MRF zone (or the entire domain to be the MRF zone) together with the nonRotatingPatches option (OpenFOAM)? This approach still has the wake information but overestimates the velocity, pressure, torque, etc. Do you have any idea about this?
Thanks Aiden . Its a wonderfull video. But i am still waiting for k - omega model. Hope you make a video for that.
😄 i am researching the k omega video right now! It should be out in june/july
@@fluidmechanics101 ok thanks.👍waiting for it😁
Hi Adian Sir, in equation 7, global velocity U is unknown or relative velocity Ur is unknown? because we are replacing the Ur with Uf in equation 16 so i think Ur is unknown initially?
Looking back at this talk, some of the notation is quite confusing! Both U and Ur are unknown initially. We are choosing which variable to solve for (the other one we can get by post processing). So the key point is that we want everything expressed in terms or U or Ur (not a mix). When we do the finite volume integration we then have the face values instead of the cell centroid values, so we make a substitution here. For the MRF approach we choose to solve in terms of U (not Ur) and so it is easier to think that we are trying to remove Ur from the equations (the opposite is true in the SRF approach)
Thanks for the great explanation.
Do the Moving frame of reference still need to use small time step (360 per revolution)?
Often you can do the moving reference frame as a steady state calculation, so you can use much larger 'time steps' than resolving the rotor rotation 👍
Transient doesnt work in Moving frame of reference, Trust me, i have tried
Yes. Moving reference frame is for steady state only. If you have a sliding mesh (full transient) then a time step that gives somewhere between 0.5 and 5 degrees of rotation per time step is probably a good guess
Does having a larger surface area between rotating and non rotaing mesh increase computational time? as it needs to add extra source term and correct the flux on more faces
Yes it will. But it is probably not that significant if the mesh is stationary, because the interpolation factors only need to be calculated once
There is a typo in the summary slide @25:16 for the source term as omega x r, but earlier @17:47 it is correct as omega x U
Well spotted! Thank you
Hi Dr. Aden,
Is there any difference between Multiple Reference Frame and Movinv Reference Freme they two called as MRF and I reckond that they can be used interchangeably, can't we?
Yea both are the same thing really 👍
Hello Dr. Wimshurst.
Again, your video is very interesting, thank you so much to clarify this complex notion.
Can you do a video for the MPM (Mixing Plane Model) for turbomachinery applications? And can you give some advices for running simulations with this model, for compressible cases (centrifugal compressor for example) because, I have encountered so many convergence problems with FLUENT solver. I have already tried to reduce the relaxation factors or CFL number, initiate with the area averaging method in the mixing plane, use the FMG initialization, ... but I don't success to converge any simulation.
Hi Antony, generally you want to try and get a stable solution first and then slowly add the physics complexity in later. I would switch the discretisation schemes to first order and try standard initialisation instead (fluent often has difficulties with hybrid and FMG). Also, have you checked your mesh?
@@fluidmechanics101 Yes, I also already tried to perform the simulation with the first order numerical schemes. But, it is true that I have never tried to use standard initialisation, because in FLUENT's trainning, they generally suggest to use hybrid or FMG for better and fast convergence.
Concerning the mesh, I'm using TurboGrid to do the mesh so the quality is by default generally good and satisfying quality criterias.
Hi Dr.Aidan,
Wonderful tutorial as usual.
Please englighten me though, in a sliding mesh environment for a wind turbine case for example;
I can construct a structured mesh in the rotating zone and the whole zone will rotate as it slides at the interface. Therefore the quality of mesh is surely good enough.
However for the MRF approach, if I understand correctly, the mesh in the rotating zone is by itself sliding against the blades. So I guess I cant use structured meshing? What about the quality and skewness etc?
I'm not sure if I quite understand your question. You can any type of mesh you want in the coin region and outside the coin region. You can even mix and match! It is quite popular to have structured mesh inside the coin, to maintain high cell quality and then unstructured outside to reduce the overall cell count
@@fluidmechanics101 I see, I initially thought that we will impose something like a dynamic mesh in the coin region. So from a meshing point of view, everything is essentially static, right?
Yep, everything is static from a meshing perspective. You will need a separate 'fluid zone' so that the code knows where to apply the MRF source terms. You can essentially use the same mesh as the sliding mesh approach. But with this approach the coin is stationary and we apply a source term inside it (and a flux correction on the interface)
Hello Dr Aiden, is there any condition to determine the size of the rotary domain?
If you have an open rotor (wind turbine, propeller) then you could try setting the diameter to 120% of the rotor diameter. I don't know any formal guidance but you want to avoid having the domain interface being too close to the rotor tip, as it could interfere and lead to errors
@5:36 by using MRF / or in the frame of rotating , u said the flow field will be steady ,I didn't get that how fluid flow conditions( steady or unsteady) is related with reference frame
Maybe think about the case of a wind turbine. If the wind flow approaching the turbine is steady (not changing on time) then the flow approaching the blade does not change in time, even though the blade is rotating. However on a gusty day, where the wind flow is changing, the flow approaching the blades will be unsteady 👍
How far one should place an interface? for example here between rotor and stationary chamber? does that change the accuracy of result?
20% of the rotor diameter is usually fine. The upstream flow field, rotor thrust and power are rarely affected significantly, but the downstream flow field often is. So it is a good idea to have a look at the flow contours and check that they look sensible
I have a doubt, the MRF uses a stationary mesh and corrects using the flux terms or does it uses a moving mesh just like the other method sliding mesh
With MRF the mesh is stationary. It uses an added source term and corrects the flux on the interface between the stationary and MRF zones. With sliding mesh, the mesh is moved physically each iteration
Hello, Aidan, in CFD, how can I decouple the pressure and mass flow rate when setting the boundary conditions. what I mean is how can set different mass flow rate at the same pressure in boundary inlet? Thanks.
Hi lily, i dont think you will be able to specify the pressure and the mass flow rate an your inlets. I would probably specify the mass flow rates and let the cfd code calculate the pressure at the inlets. That should give you a stable solution
thanks, Adian,.
Hello, Dr Aiden. My name is Luis Sosa I send you a warm greeting from Mexico City. I have just one question, hope you can help me, please. Why are you using the incompressible N-S equation if the flow is compressible on turbomachinery??
Ah good point! My background is in wind turbines and tidal turbines which are incompressible, so i guess it's just a habit of mine 😂
Hi, I was developing the derivation of the Navier-Stokes equations and I found that for the MRF calculation it considers that the viscosity is constant, does this mean that it is only applicable to Newtonian fluids?
Hmm good question. I don't think so. I think MRF should be perfectly fine for non-newtonian fluids as well (we probably just need to check the equations)
@@fluidmechanics101 thanks for the quick answer, I checked again the equations considering the variable viscosity and if it is possible with non-newtonian fluid :-)
hi,
Firstly, thank you for making this excellent video about MRF.
Just for interesting, if we can use the SRF equation for rotating region and the normal NS equations for the non-rotating, but in the interface, we need to applied a velocity transform to just one reference frame in order to compute the interface velocity? Is it this scheme workable?
Another questions is from the apply the MRF method to the SIMPLEC solver. If my understanding is right, in the SIMPLEC solver, the momentum equations will become the MRF form while the pressure correction equation will still the same as the extra term given by the rotating motion will be cancel since the subtraction of divergence-free velocity and the non-divergence-free velocity.
Thank you for your time for reading and if you will reply me it will be wonderful.
jialiang.zhou
hi,
is that the extra source term in 25:19 is incorrect? The source term will be capital omega X U in the absolute frame?
Yep, should be a capital. Well spotted!
Thank you for the video. I think the velocity triangle at 10:45 may be incorrect. U and U_r should be swapped and the direction of (Omega x r) should be reversed
Very clear and useful explanation. Thank you.
I probably should have explained that the velocity triangle is in a reference frame moving with the blade. Thats why it looks like it is the wrong way around. Sorry if this was confusing, the rest of the talk should be straightforward to follow even if the diagram itself is a bit puzzling 👍
Very nice. Look lecture on particle flow
Thank you for the video , i would like to contact with u to ask about a simulation that i'm doing in ansys fluent .if there is no problem with it
Sure, send me an email: fluidmechanics101@gmail.com and I will see if I can help.
Thank you sir ,I have sent you an email
I tried using MRF and see the velocoty profile. The velocity grafient between the rotary domain and the stationary domain are very high, is it okay? I do a simulation for a rotating wheel, I applied MRF on the spokes and inner rim, and rotating wall on outer rim and tires.
From your description it sounds like this is because of the rotating wall you have applied at the interface between the stationary and rotating zones?
@@fluidmechanics101 I think this happens because when we use frame motion, we rotates all the fluid insde the rotary domain but in reality maybe the effect of the wheel rotation not rotate all fluid inside the domain, but i am not sure. Thank you for your reply.
I am trying LES simulation and i need rotation condition. Can i use MRF approach? Can i use MRF approach for unsteady analysis
You can use MRF for unsteady calculations but .... It will depend on the period of your unsteadiness and how close this is to the period of your rotating body.
As you have an unsteady simulation anyway, why not just rotate the mesh and do sliding mesh anyway? Then you can be a lot more confident in your answer
@@fluidmechanics101 Thank you very much for your reply
Can MRF be applied to non-symmetrical rotating body about an axis?
You sure can. I would probably do a comparison with a full sliding mesh transient anyway, just to check
@@fluidmechanics101 thank you!!
Can you say how Ansys CFX node based and Ansys Fluent cell centroid based apporach different? why we use Ansys cfx for MRF problems?
You can use either fluent or cfx for MRF problems. Many people prefer CFX for turbomachinery applications as it has many automated features that make things easier to post process (multiple blade rows in compressors for example), which was a traditional strength of the code. Nowadays I would go with whatever you feel comfortable with and have access to
Thank you and you are second reason to love CFD . I appreciate all your work@@fluidmechanics101
Oh shucks ☺️ I'm just trying to make the confusing world of CFD a little bit clearer, one video at a time
I don't understand why the steady state solution to a stationary mesh with rotational velocity component on a blade surface gives a different result to a steady state solution with rotating mesh and no blade surface velocity component.
Shouldn't these give the same result?
What am I missing here with my intuition?
edit: I guess this is analogous to having the fluid domain boundaries or the surfaces of the airplane in the middle have velocity components in a "linear" case. Shouldn't these also give the same result?
CFD Harry Potter!
Yep, CFD is popular in gryffindor house
@@fluidmechanics101 Haha happy to hear that! And happy new year😄!