[CFD] The Transition SST (gamma - Re_theta) model
ฝัง
- เผยแพร่เมื่อ 29 ก.ย. 2024
- An introduction to the Transition SST model that is used to capture laminar-turbulent transition in modern, unstructured CFD codes (OpenFOAM, Fluent, CFX, Star). The following topics are covered:
1) 3:03 How is the Transition SST model different to the k-omega SST model?
2) 4:45 What is gamma?
3) 14:45 What is Re_theta?
3) 27:24 What boundary conditions should be applied to the model?
#transitionSST #fluidmechanics101 #gammaRetheta
===========================================
Some useful references:
===========================================
1) ANSYS FLUENT User Manual
4.7.2 Transport Equations for the Transition SST Model
www.afs.enea.it...
2) Robin Blair Langtry, 'A Correlation-Based Transition Model using Local
Variables for Unstructured Parallelized CFD codes', University of Stuttgart,PhD Thesis, 2006
3) P. Malan, K. Suluksna, E. Juntasaro, 'Calibrating the gamma-Re_theta Transition Model for Commercial CFD', 47th AIAA Meeting, 2009.
4) Langley Research Centre, NASA turbulence models database:
turbmodels.lar...
============================================
Want to learn more?
============================================
Grab a copy of my CFD Fundamentals Course (for beginners):
www.udemy.com/...
Learn how to write your own CFD code in MATLAB and python (for intermediates):
dr-aidan-wimsh...
Learn how I draw my figures and diagrams in Inkscape (for everyone):
dr-aidan-wimsh...
============================================
Did you like the video?
============================================
Download the lecture slides from my website:
www.fluidmecha...
Buy me a coffee to say thanks:
www.buymeacoff...
Support the channel on Patreon (and get useful extras for your CFD studies):
/ fluidmechanics101
============================================
Donations:
============================================
1) PayPal
www.paypal.me/...
==================================
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. - วิทยาศาสตร์และเทคโนโลยี
![[CFD] Turbulence Intensity for RANS](http://i.ytimg.com/vi/Xr7BzHImL68/mqdefault.jpg)
![[CFD] Turbulence Intensity for RANS](/img/tr.png)
![[CFD] Large Eddy Simulation (LES) 3: Sub-Grid Modelling](http://i.ytimg.com/vi/N81Io_yrOQU/mqdefault.jpg)
![[CFD] Eddy Viscosity Models for RANS and LES](http://i.ytimg.com/vi/SVYXNICeNWA/mqdefault.jpg)
Thank you so much Aidan...you are the kind of teacher everyone is longing for...
Thanks Aidan, but may i ask what is C1 and Cmu the SST parameter file in my OpenFOAM SST system?
Goat🐐🐐🐐
Hi Aidan. I am wondering if I could use SST transition model for high Reynolds flow too. seems the gamma can turn to 1 which converts the SST transition model to k-w SST anyway. Is that right?
Yep, you could use it for fully turbulent flow. However it is quite a bit slower, and less stable. So the extra effort might not be worth it, if the transition region is very small.
A sensible approach would be to try normal SST first, then transition SST and compare your answers. If the drag is basically the same, then you could just go with normal SST. If there are some differences, then transition SST is probably worth the extra effort
Big fan of your work.
Just a quick comment at 9:37 F1 is 1 near the wall which is also mentioned in the slide but in the narration, it was 0 for both regions.
Keep up the good work.
Ahh yes! I slipped up in my narration. It should be 1 near the wall and 0 away from the wall (as shown on the slide)
thank you so much!!! I wonder your great insight of the simulation!!
Thanks, Aiden! It is a fantastic video that helped me a lot to understand this model, but there's one question:
as it is mentioned in the papers, this model is not Galilean invariant, but I'll find many papers using this model to model moving body problems like Darrieus wind turbines with dynamic mesh approach, isn't that a mistake? I understand why it's not correct to use MRF, but I can't understand why it's okay to use dynamic mesh when using this turbulence model.
Hello Dr Aiden, another great video as usual. Do try to make a video on Detached eddy simulations and scale resolving options as seen in Ansys fluent, of-course if its in demand.
Yes, there is a high demand for DES, SAS and LES in fluent. I am going to start by finishing off the rans models with the k omega model (which will be out soon). Then i am going to move on to the LES type models 👍
Thank you so much. Be careful, the exponent in F3 is 8 and not 3 (see Menter, Langtry, et al. 2004, 2006)
Great video! I spent a couple of hours watching a half hour video.
Nice Work @Fluid Mechanics 101. I did use gamma-Re_theta model in my under grad studies for predicting transition on a Airship envelope. Also published. Thank you for introducing this in detail to everyone, as it is not a very famous turbulence model.
Thank you so much for your comprehensive explanations.
Excellent job.
Thanks again Prof Engin!
Nice work. I am working on train aerodynamics and your lectures were of great help (both basics and advanced topics). I am learning POD and DMD these days and would like to learn from you too regarding these. Keep up your good work. Stay healthy. Thanks again.
Thankyou! I need to understand and describe this model in dissertation, very helpful thanks dude :)
Incredibly well explained! I wish I would have found this video earlier. Thank you!
Very nice work. Thank you for the explanation. I want to suggestion a new subject for future videos. I want to explain more about the UDF in CFD programs. Thanks again.
Good suggestion. The difficulty is UDF is different for every CFD software (Fluent, Star, OpemFOAM), so this might be tricky. I am sure Fluent would be popular though and would help a lot of people!
Hi Aiden
Thanks a lot for excellent lectures. It helps me lot and learn lot. Could you share your softcopy of lectures. That will help to read those lectures in one bounch.
Also want some dynamic simulation such as stirring in mixing tank.
Best Regards
Prasanjit Das, Bangladesh
Hi Prasanjit, yes of course! You can get PDF slides of the lectures from my website (link in the description, www.fluidmechanics101.com/pages/shop.html) or from patreon (link also in the description). Thanks for the support 😊
Hi Aiden
Thanks! You will discuss in future. Please keeping uploading your lovely lectures and also want Ansys somulation.
I believe there is a typo in slide number 5. In the last three lines, I believe the variable should be gamma instead of P_k.
Yep, well spotted!
@@fluidmechanics101 By the way, this channel is probably one of the best channels for explaining Fluid Mechanics in TH-cam. Thank you very much for it.
Maybe you should consider explaining more advanced topics such as transition and non-modal stability theory? ;)
At 9:44, you say that F1 would tend to 0 near the wall in a laminar boundary layer. Can you clarify WHY this was happening in the first place? My best guess is as follows:
The definition of F1 from your video on the k-w SST model suggests that in the laminar BL, since k is very small, the argument to the tanh( ) function will be very small. Is this line of thinking correct?
Yep, that's the one 👍
Aidan, congratulations for you explanations! Could you suggest a link to good practices for modeling turbulent boundary layers of rough surfaces with RANS methods? Most models we have found rely on wall functions calibrated only for flat plates with zero pressure gradient. Thank you!
The only source I know of is the ANSYS CFX manual, which has a variety of treatments (including treatment for icing surfaces which have high roughness)
@@fluidmechanics101 Thank you very much. There is surprisingly little bibliography about how to mesh a boundary layer when the surface is rough.
Hi Aidan!
You cannot imagine how much these videos are helping me to understand some concepts with which I had been struggling for a lot of time. Thank you for all your effort!
I wanted to ask you about something I have noticed for some time but have not still been able to decipher: I am not able to understand why the second term of the material derivative takes the form that it takes in, for example, the scalar transport equations in the 4:00 of this video (or the Navier-Stokes equations in any other video). I know that, in the definition of the material derivative, the second term (the one due to spatial variations) is the dot product of the velocity and the gradient of the quantity of interest, which is a formulation that makes sense for me. However, that term appears here as the divergence of the quantity of interest times the velocity. It is as if the nabla operator and the velocity had changed their position in that term. Is there any underlying assumption that allows performing this change? What would be the intent of it?
Apart from that, I would like to suggest you a couple of videos that, at least for me, would be very interesting: one about boundary conditions in general (what are the implications of setting a pressure outlet, no-slip walls, etc. for the different variables in a RANS simulation), another about DES/LES and how to set up such a simulation compared to RANS, and one about meshing strategies and good practices when creating a mesh -just in case you are looking for topics of your next videos-.
Thank you again for sharing your knowledge in such a comprehensive manner. I got also the full version of your Fundamentals Course and it is helping a lot.
Kind regards.
Hi David, thanks so much for your support and the kind feedback. Yes, i spotted the same thing as you and it keeps cropping up! The material derivative of U for example gives partial U / partial t + (U dot nabla) U, but most places seem to write partial U / partial t + nabla dot (U U). To this day i havent been able to work out why! But my best guess is that it is something to do with the product rule and mass conservation. Since nabla dot (U U) = (U dot nabla) U + (nabla dot U) U, the second term is zero if the flow is incompressible. However I am not sure and am still looking for an answer. If i find one i will post it immediately, as this seems to be a definite gap in the process ....
Great, those are some fantastic video ideas. I will add them to my long list! My aim is to make a video on all these important topics as the imformation really seems to be missing from the internet and a good explanation is needed. However, it does take me a little while as the topics are sometimes complicated and confusing and i want to make sure we get it right, so it can be a good refefence.
I really hope you enjoy the course. I found it really useful to make and it really helps put a lot of the finite volume method in a consistent framework. I am working on more content and detailed courses still to come, so stay tuned!
All the best
Aidan
@@fluidmechanics101 the passage from infinitesimal non conservation (material derivative) form to infinitesimal conservation form can be explained like this: i indicate the x component of velocity vector V with u. rho du/dt + (rho V*∇). BUt we can write the first term as d(ρ u)/dt - u d(ρ)/dt; and the second term as ∇ * (ρ u V) - u ∇*(ρ V). We put them into the the original and then we notice that dρ/dt + ∇*(ρ V) = 0 from the continuity equation. Then we are left with d(ρ u)/dt +∇*(ρuV). Done! Just wanted to clarify this as a little thank u for this awesome video on this really hard turbolence model! :) Have a good day
Awesome explanation! Thanks Domenico
@@domenicobianchi8 Thanks for your explanation :)
Hi Aidan,
Can you please share some best practices in doing a transient simulation on an airfoil using the transition model? How to get the number of time step/size etc and proper way to do the mesh/time step independent study.
Have a go with some different time step sizes and number of iterations per time step. Compare the results. If the results dont change much then you can be confident your time step is sufficiently converged
@@fluidmechanics101 Thanks a lot Aidan!, So in other words I should just set up the monitor for lift, drag and maximum courant number in the system. I should then use adaptive time stepping to pull down the courant number. Once the courant number is close to one i can then maintain the same time step until lift and drag give a constant reading. Sounds good?
Sounds good to me 👍
@@fluidmechanics101 Hello Aidan,
Just another thought, I understand the importance of having a good initialization from a steady state simulation, but how do you define a good initial guess from the steady state? Should it be closer to the experimental data or should it taken when the residuals are behaving well? (Assume we can choose only one).
I would go with your initial guess as low residuals. Remember that you have to have an initial guess for all cells in the mesh and it can be tricky to know what a good guess would be in some areas. Either way, the CFD codes are pretty good and the initial guess is usually washed out in the first few hundred iterations of the solver, so as long as you are close you will be good to go
Hi Aidan, (its me again,lol)
I hope that you can please help me out with this one;
I have a pipe system with very low inflow velocity (supposedly Laminar). Then wall heating is applied after some distances downstream.
Due to the sudden change, I am not really sure if the system is entirely laminar anymore.
Is it okay if I use the transition model instead?
I would do a quick comparison with: fully turbulent (k omega SST), fully laminar (no turbulence model) and transition (transition SST). Have a look at the results and see if there are any differences. Also it is worth looking at the ratio of eddy viscosity / dynamic viscosity in your fully turbulent calculation. If the ratio is
@@fluidmechanics101 The system if for Chemical Vapor Deposition (CVD) and the Re is 743. From what's given to me (CFD, laminar), it seems that there's some circulation as the flow reaches the heated walls. This is why Im having some doubts using the fully laminar model.
Ah I see, yes the flow will accelerate when it reaches the heated section. You need to reach Re = 2500 to get turbulence for internal flow (so roughly 3 times your current flow speed). Yep this is going to be a tricky one to model. How expensive is the model to run? Can you just try laminar flow first and see what happens? Then try with turbulence and transition models
@@fluidmechanics101 I haven't tried running yet so I can't really tell how expensive it is. Currently just want to get some feel on the physics behind it to make sure I'm modeling it correctly haha. But surely I'll go with your suggestion. Will share my findings here later on. Perhaps somebody will find it useful. Thanks Aidan!
Hi, thanks for your video. I have a question. I'm working on a channel with ribs for the Reynolds number from 10 to 6000 in fluent. Is it need to use of SST model for the range of 2300 to 4000? Or i can use of turbulence model like k-e
Thanks a lot
It would be nice if u can make videos for LES models.
I know! I really need to make some LES videos. As most people tend to use RANS thats why im making these first
hope you issue a video about LES, thank you very much.
Its on the way 😄
Hi Aidan,
Im wondering why is the simulated result using transition SST got me drag values which are twice as large compared to the experimental data for re ranging from 50k-100k? What can i do to solve this?Changing TU, turbulence viscosity ratio gave minimal change. Should i change the empirical data "model constant"?
Can you compare profiles of wall shear stress with the experimental data? This will tell you if you are predicting transition in the correct location
@@fluidmechanics101Noted Aidan,
So does it cd is by the separation whereas cl is not. Also, i have not seen any paper published yet regarding naca 0015 with low reynolds of 50k,80k and 160k roughly with cf and cp plots. Only very old ones with aerodynamic coefficients
Have you tried the book ‘Theory of lifting sections’ by Abbott and Von Doenhoff? I think this book has lots of aerodynamic data you can use
@@fluidmechanics101 Noted with thanks Aidan
Thanks a lot. It cleared a lot of my doubts. The content is excellent, please keep uploading more videos like this.
Thanks Aditya!
@@fluidmechanics101 so yesterday i had my project presentation, and the information i gained from the video was instrumental in my understanding of the model. Thanks again for the video
Sir, as I understand, there are Flenght and Rec which are constant values for a certain airfoil.
What's the difference b/w dissipation rate (epsilon) and specific dissipation rate (omega)?
Hi, How does the Transition SST differ from K Omega SST with intermittency turned on?
They are the same 🙃 some CFD codes will just say ‘turn intermittency on’ rather than describing it as a different model
@@fluidmechanics101 Thanks Aidan!
@@fluidmechanics101 One more thing, how come I get very different results when using ansys mesher (unstructured) and ICEM regardless of good overall quality/skewness on both. Surprisingly, finer mesh despite good overall quality etc gives worst results. Is there a trick to this?
Do they both have the same y+, growth ratio normal to the wall and spanwise resolution? Transition SST is very sensitive to your near wall mesh (inflation layers)
@@fluidmechanics101 yeah i did both for y+ : 0.5 and 1. Spanwise i had 400 points. When both styles of mesh had this level of refinement, the bl failed to reattach. I wonder why?
@Fluid Mechanics 101 The stream edge definition for the SST Transition model in CCM+ is defined as $WallDistance>0.005?1:0 which means that boundary layer is 5 mm thin everywhere. Can we change 0.005 with the BL thickness of our problem of interest?
Hi Mazhar, sorry i am not a CCM+ expert. Dont think i can help you. Have you tried the CFDonline forums?
@@fluidmechanics101 Yes. Posted it there.
Me: ¨How do I change the oil in my car?¨
Him: ¨You solve a transport equation¨
There is always a transport equation hiding in plain sight
This is too accurate. I LOL'd when he said we now have to solve a transport equation even for the Reynold's number 🤣!
Thanks a lot, very clear explanations!
The best