Aidan, I have no word good enough to thank you. I appreciate your generosity and the amount of time and effort you put into these videos to transfer your knowledge to us. Whatever you choose for the subject of the next video, I will be blessed by the opportunity you give me.
I am going through all your videos one by one. Concept that you are expalining in detail and with reasoning is outstanding. I have started one project and your all videos are so much helpfull to build my case. Today I am pledging to give my entire one month salary to this channel. Hope I can buy a coffee for you with that. Lots of love and respect.
Watching this at 2:30am writing up a project for my grad class on turbulence. Your channel is amazing, just so great dude! Thanks a ton. Doing turbulent flow in a duct over a stair step ala Driver and Seeger.
Aidan, many thanks to you for helping me better understand CFD for the last couple months. I´ll have my master thesis defense in 1 week and your videos were indispensable and vital for going forward in my simulations. Thank you a lot! Greetings from Austria
Very enlightening talk with clear message and explanations on wall functions in ANSYS Fluent! Thank you! If you're up for ideas then here are 2: 1. Fluid Structure Interaction / Fluid Particle Interactions - Maths behind it, implementation in commercial software, any tips/details about using them people are often not aware of, etc (this one is something I'm personally interested in). 2. Viscosity models - Newtonian vs Non-Newtonian, Temperature dependence, kinetic theory, etc (this is one I think would be more helpful as a general necessary topic in fluid mechanics).
I'm going to have to say: I absolutely LOVE how you glitch in the beginning of your videos. It looks like I'm about to watch some lost scientist from a sci-fi movie!
As a beginner into ANSYS Fluent, I found your videos very useful! It really enlightened me on CFD and it would be wonderful if you could also make a video on "Enhanced Wall Treatment" Thank you in advance!!!!
Aidan, I absolutely LOVE your videos. Thank you for existing. May I request you to do a video on Unsteady RANS (Transient Analysis) as well? I'm working on the CFD analysis of vortex generators in automobiles, and I'm using Fluent. URANS may be an economical technique to capture this type of flow with fair accuracy. I would really like your explanation on it, if possible. Thanks a ton again, cheers! :)
Thanks for your video. But, at Cebeci book "Analysis of Turbulent flows with Computer Programs" 3rd Ed, Page 119. there is a formula for buffer region. The formula by thompson is 4
Keep up the great work !! Your videos are very informative and understandable, thank you for providing all this content. Please make a video for enhanced wall treatment!
It would be amazing if you made some sort of a video called ''guide for beginners in CFD'', where you could describe briefly what CFD can do (various applications) and the ''paths'' that someone can take on order to understand and use CFD in those applications (some book recomendations would be quite usefull). I'm currently a mechanical engineering undergrad and I have to say I have concerns regarding the whole CFD thing and the steps you have to take in order to use it correctly. By the way, great job, keep posting those amazing videos!
Great video! Great that you mentioned how it's actually implemented. BTW Kader blending is responsible for this bump in the profile, that's why I opted for Reichardt blending in our solver, which improves on this issue. Two layer treatment (Patel&Rodi) goes hand in hand with this topic in Ansys Fluent, hope it is on to do list as well! Keep up the good work, I'm really enjoying this video series!
Ahhh so it is Kader blending that is responsible! This had me scratching my head for a long time! Might I ask, what is Reichardt blending? Do you have a paper / reference? I would love to have a look! Thanks for the feedback, much appreciated
@@fluidmechanics101 Also check out lines 933-953 here (I comment it out, because I switch between two variants): github.com/nikola-m/freeCappuccino-dev/blob/master/src/finiteVolume/fvEqnDiscretization/TurbulenceModels/k_omega_sst.f90
@@fluidmechanics101 I think it might be this one: Reichardt, H. "Über eine neue Theorie der freien Turbulenz." ZAMM‐Journal of Applied Mathematics and Mechanics/Zeitschrift für Angewandte Mathematik und Mechanik 21, no. 5 (1941): 257-264.
Hi Adrian, thanks a lot for you fantastic videos. They are much appreciated. I wonder if you could explain the difference between enhanced wall functions and enhanced wall treatment. Is there a video where you explain the difference? Cheers.
I haven't made a video on this yet, but roughly speaking: Enhanced Wall Functions - Fluent terminology for the k-omega (and k-omega SST models) Enhanced Wall Treatment - Fluent terminology for the k-epsilon model
thank you for explaining the y+ clearly to us. i have a question that the enhanced wall function has a better agreement than standard wall function ,why ansys fluent not cancel the standard wall funciton option?
This is gold! I have a doubt, it might sound silly but please correct me if I am wrong. So, if the cell is well within the viscous sub layer, the CFD code does not use wall function since it the behavior is linear. But if the cell near wall lies in the log law region, CFD codes use wall functions to account the non-linear behavior of the flow field. As you have said the above condition is true for flow over a flat plate. What considerations should be taken for complex geometries. (with respect to y+ and turbulence model, I am still confused ) How CFD codes behave for complex geometries when it comes to y+ or thermal boundary layer. What value should one select for y+ for complex geometries. Thank you in advance Dr. Aidan. Big Fan. Some day I wish to be like you in terms of CFD Engineer.😎 Thank you for inspiring.
Yes you are 100% correct. For complex geometries ... noone really knows. However, regardless of how complex the geometry, if we get reallllly close to the wall then there will be a viscous sublayer with linear behaviour. Hence, the usual approach is to have a small y+ (around 1 or even lower) if the flow is complex
I am currently working on the next one in the LES series! It has been really helpful to a lot of people, so i am definitely going to focus on the LES videos from now on 😊
So, will you make a video about EWT?) It was mentioned in this video like a Chekhov's gun and never shot. I try to model a slug flow at Re~1*10^5 and use it but I wonder if it's an overkill. Would like to listen to your lecture about it, if it fits your plans, of course.
😅 yes this is another video on my very long list ... I am planning on making a video as soon as I get around to it. For now, I think you should use it if ANSYS recommends it for your case. That's the best you can really do without knowing how some functionality works in detail
Hi Aidan, thank you for your CFD series episodes. They are very helpful. Can you please explain how a two-layer model is combined with enhanced wall functions to form enhanced wall treatment?
Please can you make a video compressible flow in supersonic velocity. What could the best model to used among the different RANS models, and models other than RANS.
Dear Dr. Aidan, I ran into a question about wall functions. The Spalding wall function is a very good fit for turbulence analysis, but it seems ANSYS Fluent doesn't utilize it for the turbulent boundary layer. I don't know about Open FOAM since I don't use it, but why Fluent doesn't have the Spalding wall function option? Is there any particular reason? I believe the Spalding wall function could be helpful, especially in the buffer region. I appreciate your time reading this comment. Yours sincerely, Soroush Asadian
I can't speak for fluent, but the Spalding wall function is explained quite nicely in the book by Greenshields and Weller for OpenFOAM (Notes on Computational Fluid Dynamics; General Principles). If you Google the book, you can access the webbook for free. Essentially the difficulty is that the Spalding wall function gives y+ as a function of U+, so you need to invert it. I think (but can't remember exactly) that this is expensive to do, so the standard wall functions are normally preferred
By the way, is it possible that you prepare a video about the cell aspect ratio regarding how it affects the CFD results, its maximum/minimum limit, etc.? I have found many explanation regarding the y+ values recommendation and its effect on the CFD results but have not found that many on the aspect ratio. This is also, I think, related to what the recommended stream-wise cell size is. Again, thank you very much for all your videos.
Adian, I am not sure you can use standard wall function with a high (turbulence) Reynolds number model (e.g., standard kEpsilon) on a mesh y+ that is less than five (5). The high (turbulence) Reynolds number models are not designed to be integrated down to the viscous sublayer, even if the standard wall function is correctly provided the wall turbulence viscosity formula based on y^+ switch. Please try the combination of standard wall function+standard kEpsilon+yPlus
Yes you are right. The standard k epsilon model is not designed to be applied here and will give inaccurate answers. However, when writing a CFD code you have to assume that the user is inexperienced and could give you an innappropriate mesh or a mesh with regions where y+ is ok and regions where it is not. To make sure you code is general and can account for all cases without diverging (or giving weird solutions) it is important to have a general treatment that can account for all cases. It is then up to the reader to check that their results are accurate (or not). I hope this helps!
@@fluidmechanics101 I agree with you Adian. It is absolutely needed to have a wall function with this y^+ switch. I think it also must be stressed that the correct wall boundary condition/wall function is not a sufficient condition to have an accurate solution. In the example that I presented, the accuracy of the solution is limited by the incorrect application of the mother turbulence model. The reason I bring this up is that I have seen many incorrect implementations such as the example that I mentioned in the peer-reviewed papers. Your channel will make a substantial positive impact on the CFD communities and therefore I just can not help to mention this rabbit hole. BTW, another excellent video!!!
Hello Dr. Aiden, I have 1 question. If code is just modifying nu-w (modified viscosity), and we care about shear stress, forces etc then why anybody should go for y+ < 2 when they mesh? Is the core reason for these efforts on better wall functions is that they depend on the velocity which we don't know prior hence more flexible wall function is better (enhanced wall function)? I hope I am talking sense here.
Sure. Remember that these wall functions are based on data for a flat plate boundary layer. If you don't have a flat plate, then the models are less accurate. So the way around this is to have y+ < 2, as the viscous sub layer is pretty universal (unless you have supersonic flow) and you are more likely to get the right answer, as you aren't using a model anymore
Aidan, thanks for the great video, which is very helpful to understand how wall functions are implemented behind the codes. I am wondering if having y+ in the viscous sublayer is better than having it in the log-law region since the latter has larger y+ value considering similarly good performance of wall functions in those two regions, is that true? Another thing is, why is Spalding's wall function not applied? I watched another video of yours and saw Spalding's wall function fit well in all three regions. Thanks.
Yep, if you can get y+ in the viscous sub layer this is will always give you more accurate results (just make sure that your turbulence model is appropriate here. The historic k epsilon model couldn't do the viscous sub layer but most modern versions can with appropriate damping functions. I usually go with k omega SST). Good question, I'm not sure why Spaldings function was never adopted! As this decision probably was made back in the 1970s it is difficult to find out
Great work as you just don't show how to use the software but the theory behind that. Coming from background that uses software for day to day work, you need to be able to understand what is going on behind the software. I like how you use the term CFD code to reflect software as it is a bunch of code that perform computation. Honestly speaking I have never used CFD but can relate them to fluid mechanics what you learn in school. Anyway I am just wondering whether this wall function or wall stress is kind related to those friction factor correlation that we typically learn in school like colebrook etc. Are these empiral correlation are even implemented in CFD software or CFD treates the wall stress differently?
Thanks for your kind words, I really appreciate it ☺️ in answer to your question, CFD codes don't use the friction factor correlations that you are referring to, as you are calculating the local flow features throughout your domain of interest. The friction factor correlations that you refer to can be calculated as an output of a CFD simulation, in much the same way that you can derive them from an experiment. Use the CFD (or your experiment) to calculate the total pressure drop across your domain and then normalise it appropriately, and that will give your friction factor. The 'wall function' empirical correlations refer to the local flow very close to the wall (they are smaller than a mesh cell), whereas the friction factor correlations refer to the global pressure drop across an obstacle. I hope this makes sense, and thanks for supporting!
Thanks for this video! So for my case I had to simulate a propeller. Because of license limitations I can't have a very fine inflation layer, so thought of increasing the y+ value. Would you recommend keeping the y+ values somewhere within the 15-30 range with enhanced wall function, or use std. wall function with y+ > 30? In the previous wall function video you said that y+ > 30 doesn't bode well for curvatures and pressure gradients.
Hmmm that's a good question. Could you try both and compare the results? That way you can quantity the difference between the two methods. If the difference is small then you can happily use either and don't need to worry about the choice of model
@@fluidmechanics101 I don't understand why you just recommend using EWF with y + less than 2 / greater than 30. Because in my opinion, EWF was born to correct the wall shear stress in the buffer layer and there're not many changes if y +
Well, the EWF for momentum and temperature look like they are pretty good but the behaviour of turbulent kinetic energy is more difficult to model in the buffer region, so I doubt the overall modelling of the buffer region is that good. At the end of the day, this is all modelling magic and the only way to be sure is to compare with experimental measurements / DNS
So... - Classic WF works well for y+ < 5 and y+ > 30. Between 5 < y+ < 30, it is not a good match - Enhanced WF works well for y+ < 2 and y+ > 30. Between 2 < y+ < 30, it is a not-so-good-but-better match That is a weird thing and the advantage of Enhanced WF is not such a big deal as I understand it. Also, I'm a bit uncomfortable with Wall Functions. If you have a case where you won't use y+ < 2, is there any benefit in setting y+~30 rather than ~300? Or are you "less wrong" with y+~30? Let's take the case of a diffuser with flow detachment. Using k-epsilon with WF has been proved to fail right. Because "The production of turbulent kinetic energy is over-predicted in locations with strong velocity gradients (for example near a stagnation point). Therefore flow separations are not (or insufficiently) recognized." (Gulich, Centrifugal Pumps). I believe this is a default with all turbulence models with WF. So in the end, if WF fails at predicting such thing as a detachment, it is rarely applicable in real-life engineering applications, if the goal is to resolve proper flow features. It may give good results for integral values because statistically, several errors are canceling each other but it's luck somehow and using the wrong formula to achieve a good answer. My message sounds a bit confusing, but I am a bit confused also :D. The more I do CFD, the less I trust CFD. Thanks for your work though :)
Yep i think you are pretty much there. The reason to go with y+ close to 30 is the outer region of applicability of the wall function depends on the freestream Reynolds number. This is why you will sometimes see 30 < y+ < 200 and sometimes 30 < y+ < 300. Best to go close to 30 if you are using this method 👍
Hi Aidan.... Now I am model and it is about a carparking jet fans system.... Can you tell me if I can use the Eulerian model to simulate it or lagrangian model... in the end I have to find the ppm (particles per million) of the CO (carbon dioxide) in the air that produced from an automobile. Thank you for your time and have a nice day 😊
An adverse pressure gradient is a term that aerodynamacists used to describe pressure which increases in the streamwise direction. An aerofoil for example has a favourable pressure gradient on the leading edge, as pressure decreases when the flow moves around the maximum thickness of the aerofoil. On the trailing edge the pressure increases as the flow moves along the aerofoil. This is an adverse pressure gradient and is important because it is one of the main causes of boundary layer separation at high angles of attack. I hope this helps!
Enhanced wall treatment is for k epsilon model for an internal pipe flow having turbulent nature shall have y plus in between sir? What is the maximum y plus value in simulations?
Hi Aidan, thank you for the video, I really appreciate it, like all the others. I have a question though. I am using CFX and k-omega SST, to simulate centrifugal fans at low velocities. Having a yPlus equal to 1 is not affordable for my cases so I need to live with higher values. Therefore my question is this, correct me if I am wrong: when y Plus is lower than 5 you get a wall resolved solution and you do not apply a wall function and it is the best case (k-omega); when y Plus is higher than 30 and lower than 300, CFD code uses a wall function as in k-eps But what happens if I have yPlus in between 5 and 30? Should I avoid this situation? I am asking you this because some ANSYS experts told me that I can be fine with a y Plus lower than 300, and a good blending is performed even when yPlus is in the range between 5 and 30. So I try to stay with a yPlus lower than 300, but it happens in my cases to have yPlus in this range between 5 and 30 and this situation is not very easy to be avoided. Thanks in advance.
@@fluidmechanics101 Perfect thank you for the answer. Anyway, I had a look at the CFX documentation, and an Automatic Wall Treatment is applied, when using the SST model. This is a yPlus insensitive approach, which performs a good blending between the Low-Re formulation (wall resolved solution) applied for yPlus lower than 2 and the Wall Function formulation applied for yPlus higher than 20. I was asking this because sometimes I find separation on the blades of the fan and this situation needs to be handled with care. I suppose that also with this CFX treatment, the velocity profile of the wall function is not imposed on the mesh, but it is just applied to correct the wall shear stress, maintaining a linear variation between the velocity of the centroid of the first cell and the velocity at the wall (=0), as you explained very well in your videos. Thanks for the answer and for the precious guidelines!
@@maurotortora5317 hello sir, i am simulating a centrifugal fan with fluent ,i am using k -omega sst model, i would like to ask you about the seperation in the blades how can you find out if there is a separation in the blade.thanks in advance.
There are a variety of ways of looking for separation. You can look at streamlines, wall shear stress or plot some velocity profiles normal to the wall. Either way, you need to do some investigation in the post processor!
Hello everyone, I run a cyclone simulation with Fluent (mesh 935k) with RSM. It runs fine and I get a solution that resembles experimental data. Y+ is 29. When I check enhanced wall treatment I get an error at 25th iteration: # Divergence detected in AMG for k: protective actions enabled! # Divergence detected in AMG for k, temporarily solve with BCGSTAB! Divergence detected in AMG solver: k Divergence detected in AMG solver: k Divergence detected in AMG solver: k Divergence detected in AMG solver: k Divergence detected in AMG solver: k Divergence detected in AMG solver: k Divergence detected in AMG solver: k Divergence detected in AMG solver: k Error at host: floating point exception Error at Node 0: floating point exception Error at Node 1: floating point exception Error at Node 2: floating point exception Error at Node 3: floating point exception Error at Node 4: floating point exception Error at Node 5: floating point exception Error at Node 6: floating point exception Error at Node 7: floating point exception Error: floating point exception Error Object: #f Any ideas? Thnx in advance!
Hi Matt, it is unlikely that the divergence is caused by the enhanced wall treatment alone. Have you checked all aspects of your simulation? Some basic ideas to check: what is your mesh orthogonality? What have you used for the initial condition? Have you checked your boundary conditions / user defined functions? You could always run the case with some basic settings to get initial convergence, the turn on EWT and run another 100 iterations to get the solution you want.
@@fluidmechanics101 Thnx for your instant answer. It is a cyclone simulation and the mesh orthogonality is bad in the area where the tangential inlet meets the cyclone cylinder, but hasn't caused problems with other models. I would like you to clarify me something. I run, for example 10000 iterations RSM with std_wall. If I set the enh_wall and run another 100 or 500 iterations, the results will represent the enh_wall_fn? I mean I don't have to run allover from the beginning with enh_wall_fn? Furthermore, it looks funny, but it isn't, your tutorials have been proved more useful than any help I can get from engineers in Greece (referring to CFD always). Thanks again!
Yep, you don't need to start again. You can just run another 500 iterations and you should have your solution. This trick of getting convergence with an easy simulation (1st order upwind, standard wall functions etc) first and then turning on more detailed models is really useful in CFD and has lots of uses, particularly in detailed models. It is super easy to do in Fluent as well with the TUI as you can run a set number of iterations, then turn on a model and then run 500 more iterations
![[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)
![[CFD] How are Wall Functions Modified for Roughness?](http://i.ytimg.com/vi/vYbRUmVTmGM/mqdefault.jpg)
Aidan, I have no word good enough to thank you. I appreciate your generosity and the amount of time and effort you put into these videos to transfer your knowledge to us. Whatever you choose for the subject of the next video, I will be blessed by the opportunity you give me.
I am going through all your videos one by one. Concept that you are expalining in detail and with reasoning is outstanding. I have started one project and your all videos are so much helpfull to build my case. Today I am pledging to give my entire one month salary to this channel. Hope I can buy a coffee for you with that. Lots of love and respect.
Thank you for your kind words. I really hope you find everything you are looking for and come out the other side a more confident CFD engineer 🙂
This is not possible. No way . I was just surfing online yesterday to understand enhanced wall functions. What a timing Aidan!!!!!!
Amazing!
Love the inquisitive approach and answering the why question, e.g. why we don't just do the polynomial fit to the DNS. Pleasure to watch.
Watching this at 2:30am writing up a project for my grad class on turbulence. Your channel is amazing, just so great dude! Thanks a ton. Doing turbulent flow in a duct over a stair step ala Driver and Seeger.
Good luck! I hope you get the project done
Aidan, many thanks to you for helping me better understand CFD for the last couple months. I´ll have my master thesis defense in 1 week and your videos were indispensable and vital for going forward in my simulations. Thank you a lot!
Greetings from Austria
Aidan, in each topic you discuss; you share extremely precious information. Looking forward to new videos. Gramercy !
Thanks for the detailed explanations. These questions have been confusing me for quite a lot of time. You explained the difference so well.
Very enlightening talk with clear message and explanations on wall functions in ANSYS Fluent! Thank you!
If you're up for ideas then here are 2:
1. Fluid Structure Interaction / Fluid Particle Interactions - Maths behind it, implementation in commercial software, any tips/details about using them people are often not aware of, etc (this one is something I'm personally interested in).
2. Viscosity models - Newtonian vs Non-Newtonian, Temperature dependence, kinetic theory, etc (this is one I think would be more helpful as a general necessary topic in fluid mechanics).
I'm going to have to say: I absolutely LOVE how you glitch in the beginning of your videos. It looks like I'm about to watch some lost scientist from a sci-fi movie!
😂😂
As a beginner into ANSYS Fluent, I found your videos very useful! It really enlightened me on CFD and it would be wonderful if you could also make a video on "Enhanced Wall Treatment" Thank you in advance!!!!
Aidan, I absolutely LOVE your videos. Thank you for existing. May I request you to do a video on Unsteady RANS (Transient Analysis) as well? I'm working on the CFD analysis of vortex generators in automobiles, and I'm using Fluent. URANS may be an economical technique to capture this type of flow with fair accuracy. I would really like your explanation on it, if possible. Thanks a ton again, cheers! :)
Thanks for your video. But, at Cebeci book "Analysis of Turbulent flows with Computer Programs" 3rd Ed, Page 119. there is a formula for buffer region.
The formula by thompson is 4
Thanks for pointing this out Murat!
Keep up the great work !! Your videos are very informative and understandable, thank you for providing all this content. Please make a video for enhanced wall treatment!
It would be amazing if you made some sort of a video called ''guide for beginners in CFD'', where you could describe briefly what CFD can do (various applications) and the ''paths'' that someone can take on order to understand and use CFD in those applications (some book recomendations would be quite usefull). I'm currently a mechanical engineering undergrad and I have to say I have concerns regarding the whole CFD thing and the steps you have to take in order to use it correctly. By the way, great job, keep posting those amazing videos!
Excellent explanation. Thanks a million Dr. Aidan.
Great explanation. Crystal clear.
Yes, it was very useful. Please cover enhanced wall treatment
Thanks for the detailed explanations. Its very informative.
Perfect video! Adian, please carry on! :)
Great video! Great that you mentioned how it's actually implemented. BTW Kader blending is responsible for this bump in the profile, that's why I opted for Reichardt blending in our solver, which improves on this issue. Two layer treatment (Patel&Rodi) goes hand in hand with this topic in Ansys Fluent, hope it is on to do list as well! Keep up the good work, I'm really enjoying this video series!
Ahhh so it is Kader blending that is responsible! This had me scratching my head for a long time! Might I ask, what is Reichardt blending? Do you have a paper / reference? I would love to have a look! Thanks for the feedback, much appreciated
@@fluidmechanics101 Also check out lines 933-953 here (I comment it out, because I switch between two variants): github.com/nikola-m/freeCappuccino-dev/blob/master/src/finiteVolume/fvEqnDiscretization/TurbulenceModels/k_omega_sst.f90
@@fluidmechanics101 I think it might be this one: Reichardt, H. "Über eine neue Theorie der freien Turbulenz." ZAMM‐Journal of Applied Mathematics and Mechanics/Zeitschrift für Angewandte Mathematik und Mechanik 21, no. 5 (1941): 257-264.
Thank you so much for sharing your knowledge. It would be very helpful if you could do a video on DES and IDDES. Thank you.
Hi Adrian, thanks a lot for you fantastic videos. They are much appreciated.
I wonder if you could explain the difference between enhanced wall functions and enhanced wall treatment.
Is there a video where you explain the difference? Cheers.
I haven't made a video on this yet, but roughly speaking:
Enhanced Wall Functions - Fluent terminology for the k-omega (and k-omega SST models)
Enhanced Wall Treatment - Fluent terminology for the k-epsilon model
@@fluidmechanics101 realized that I misspelled your name. Very sorry for that. Thank you for your answer. Best regards
thank you for explaining the y+ clearly to us. i have a question that the enhanced wall function has a better agreement than standard wall function ,why ansys fluent not cancel the standard wall funciton option?
It's for backwards compatibility (in case you want to compare against a historic case)
This is gold!
I have a doubt, it might sound silly but please correct me if I am wrong.
So, if the cell is well within the viscous sub layer, the CFD code does not use wall function since it the behavior is linear.
But if the cell near wall lies in the log law region, CFD codes use wall functions to account the non-linear behavior of the flow field.
As you have said the above condition is true for flow over a flat plate.
What considerations should be taken for complex geometries. (with respect to y+ and turbulence model, I am still confused )
How CFD codes behave for complex geometries when it comes to y+ or thermal boundary layer.
What value should one select for y+ for complex geometries.
Thank you in advance Dr. Aidan.
Big Fan. Some day I wish to be like you in terms of CFD Engineer.😎
Thank you for inspiring.
Yes you are 100% correct. For complex geometries ... noone really knows. However, regardless of how complex the geometry, if we get reallllly close to the wall then there will be a viscous sublayer with linear behaviour. Hence, the usual approach is to have a small y+ (around 1 or even lower) if the flow is complex
@@fluidmechanics101 I saw your video on inflation layer and it pretty much cleared my doubt!
Thank you Dr. Aiden
Aidan, you're awesome, man.
Very nice video sir full of knowledge please continue LES series , it's request as I badly need your LES video
I am currently working on the next one in the LES series! It has been really helpful to a lot of people, so i am definitely going to focus on the LES videos from now on 😊
@@fluidmechanics101 thanks sir
@@fluidmechanics101 Great
So, will you make a video about EWT?) It was mentioned in this video like a Chekhov's gun and never shot. I try to model a slug flow at Re~1*10^5 and use it but I wonder if it's an overkill. Would like to listen to your lecture about it, if it fits your plans, of course.
😅 yes this is another video on my very long list ... I am planning on making a video as soon as I get around to it. For now, I think you should use it if ANSYS recommends it for your case. That's the best you can really do without knowing how some functionality works in detail
@@fluidmechanics101 Thanks, will look forward for your works)
Hi Aidan, thank you for your CFD series episodes. They are very helpful. Can you please explain how a two-layer model is combined with enhanced wall functions to form enhanced wall treatment?
Coming soon! This is for the k-epsilon based models
It's an informative video, could you please make a video on the K-e model with enhanced wall function approach?
Coming soon!
My doubts cleared
Great video! thanks a ton!
Please make a video on polyhedral vs hexahedral mesh. In sort, fluent meshing vs ICEM CFD meshing.
Will do! I think this would be a great topic!
Please can you make a video compressible flow in supersonic velocity. What could the best model to used among the different RANS models, and models other than RANS.
Dear Dr. Aidan, I ran into a question about wall functions. The Spalding wall function is a very good fit for turbulence analysis, but it seems ANSYS Fluent doesn't utilize it for the turbulent boundary layer. I don't know about Open FOAM since I don't use it, but why Fluent doesn't have the Spalding wall function option?
Is there any particular reason?
I believe the Spalding wall function could be helpful, especially in the buffer region.
I appreciate your time reading this comment.
Yours sincerely,
Soroush Asadian
I can't speak for fluent, but the Spalding wall function is explained quite nicely in the book by Greenshields and Weller for OpenFOAM (Notes on Computational Fluid Dynamics; General Principles). If you Google the book, you can access the webbook for free.
Essentially the difficulty is that the Spalding wall function gives y+ as a function of U+, so you need to invert it. I think (but can't remember exactly) that this is expensive to do, so the standard wall functions are normally preferred
@@fluidmechanics101 Thank you so much
Very very helpful ☺️🙏
By the way, is it possible that you prepare a video about the cell aspect ratio regarding how it affects the CFD results, its maximum/minimum limit, etc.? I have found many explanation regarding the y+ values recommendation and its effect on the CFD results but have not found that many on the aspect ratio. This is also, I think, related to what the recommended stream-wise cell size is. Again, thank you very much for all your videos.
Yep, i will see what i can do 😊 i agree that aspect ratio is more difficult to find good information than y+
@@fluidmechanics101 Thanks a lot. Looking forward to your new videos😊
Adian, I am not sure you can use standard wall function with a high (turbulence) Reynolds number model (e.g., standard kEpsilon) on a mesh y+ that is less than five (5). The high (turbulence) Reynolds number models are not designed to be integrated down to the viscous sublayer, even if the standard wall function is correctly provided the wall turbulence viscosity formula based on y^+ switch. Please try the combination of standard wall function+standard kEpsilon+yPlus
Yes you are right. The standard k epsilon model is not designed to be applied here and will give inaccurate answers. However, when writing a CFD code you have to assume that the user is inexperienced and could give you an innappropriate mesh or a mesh with regions where y+ is ok and regions where it is not. To make sure you code is general and can account for all cases without diverging (or giving weird solutions) it is important to have a general treatment that can account for all cases. It is then up to the reader to check that their results are accurate (or not). I hope this helps!
@@fluidmechanics101 I agree with you Adian. It is absolutely needed to have a wall function with this y^+ switch. I think it also must be stressed that the correct wall boundary condition/wall function is not a sufficient condition to have an accurate solution. In the example that I presented, the accuracy of the solution is limited by the incorrect application of the mother turbulence model. The reason I bring this up is that I have seen many incorrect implementations such as the example that I mentioned in the peer-reviewed papers. Your channel will make a substantial positive impact on the CFD communities and therefore I just can not help to mention this rabbit hole. BTW, another excellent video!!!
Thank you, this is very useful.
perfecto. i really liked the video. it helped.
Hello Dr. Aiden, I have 1 question. If code is just modifying nu-w (modified viscosity), and we care about shear stress, forces etc then why anybody should go for y+ < 2 when they mesh? Is the core reason for these efforts on better wall functions is that they depend on the velocity which we don't know prior hence more flexible wall function is better (enhanced wall function)? I hope I am talking sense here.
Sure. Remember that these wall functions are based on data for a flat plate boundary layer. If you don't have a flat plate, then the models are less accurate. So the way around this is to have y+ < 2, as the viscous sub layer is pretty universal (unless you have supersonic flow) and you are more likely to get the right answer, as you aren't using a model anymore
@@fluidmechanics101 thanks a ton sir!
Aidan, thanks for the great video, which is very helpful to understand how wall functions are implemented behind the codes. I am wondering if having y+ in the viscous sublayer is better than having it in the log-law region since the latter has larger y+ value considering similarly good performance of wall functions in those two regions, is that true? Another thing is, why is Spalding's wall function not applied? I watched another video of yours and saw Spalding's wall function fit well in all three regions. Thanks.
Yep, if you can get y+ in the viscous sub layer this is will always give you more accurate results (just make sure that your turbulence model is appropriate here. The historic k epsilon model couldn't do the viscous sub layer but most modern versions can with appropriate damping functions. I usually go with k omega SST). Good question, I'm not sure why Spaldings function was never adopted! As this decision probably was made back in the 1970s it is difficult to find out
Thanks for your reply.@@fluidmechanics101
Great work as you just don't show how to use the software but the theory behind that. Coming from background that uses software for day to day work, you need to be able to understand what is going on behind the software. I like how you use the term CFD code to reflect software as it is a bunch of code that perform computation. Honestly speaking I have never used CFD but can relate them to fluid mechanics what you learn in school. Anyway I am just wondering whether this wall function or wall stress is kind related to those friction factor correlation that we typically learn in school like colebrook etc. Are these empiral correlation are even implemented in CFD software or CFD treates the wall stress differently?
Thanks for your kind words, I really appreciate it ☺️ in answer to your question, CFD codes don't use the friction factor correlations that you are referring to, as you are calculating the local flow features throughout your domain of interest. The friction factor correlations that you refer to can be calculated as an output of a CFD simulation, in much the same way that you can derive them from an experiment. Use the CFD (or your experiment) to calculate the total pressure drop across your domain and then normalise it appropriately, and that will give your friction factor. The 'wall function' empirical correlations refer to the local flow very close to the wall (they are smaller than a mesh cell), whereas the friction factor correlations refer to the global pressure drop across an obstacle. I hope this makes sense, and thanks for supporting!
Thanks for this video!
So for my case I had to simulate a propeller. Because of license limitations I can't have a very fine inflation layer, so thought of increasing the y+ value. Would you recommend keeping the y+ values somewhere within the 15-30 range with enhanced wall function, or use std. wall function with y+ > 30? In the previous wall function video you said that y+ > 30 doesn't bode well for curvatures and pressure gradients.
Hmmm that's a good question. Could you try both and compare the results? That way you can quantity the difference between the two methods. If the difference is small then you can happily use either and don't need to worry about the choice of model
@@fluidmechanics101 Hmmm ok, seems like no other option. Have to try both.
Thanks for your video. So what is the difference between EWF and SWF if y+ less than 5 ? Which is better to capture seperation/detachment ?
Both should be fine. You could always do a side by side check to see if it makes much difference at all
@@fluidmechanics101 I don't understand why you just recommend using EWF with y + less than 2 / greater than 30. Because in my opinion, EWF was born to correct the wall shear stress in the buffer layer and there're not many changes if y +
Well, the EWF for momentum and temperature look like they are pretty good but the behaviour of turbulent kinetic energy is more difficult to model in the buffer region, so I doubt the overall modelling of the buffer region is that good. At the end of the day, this is all modelling magic and the only way to be sure is to compare with experimental measurements / DNS
So...
- Classic WF works well for y+ < 5 and y+ > 30. Between 5 < y+ < 30, it is not a good match
- Enhanced WF works well for y+ < 2 and y+ > 30. Between 2 < y+ < 30, it is a not-so-good-but-better match
That is a weird thing and the advantage of Enhanced WF is not such a big deal as I understand it.
Also, I'm a bit uncomfortable with Wall Functions.
If you have a case where you won't use y+ < 2, is there any benefit in setting y+~30 rather than ~300? Or are you "less wrong" with y+~30?
Let's take the case of a diffuser with flow detachment.
Using k-epsilon with WF has been proved to fail right. Because "The production of turbulent kinetic energy is over-predicted in locations with strong velocity gradients (for example near a stagnation point). Therefore flow separations are not (or insufficiently) recognized." (Gulich, Centrifugal Pumps).
I believe this is a default with all turbulence models with WF.
So in the end, if WF fails at predicting such thing as a detachment, it is rarely applicable in real-life engineering applications, if the goal is to resolve proper flow features. It may give good results for integral values because statistically, several errors are canceling each other but it's luck somehow and using the wrong formula to achieve a good answer.
My message sounds a bit confusing, but I am a bit confused also :D. The more I do CFD, the less I trust CFD.
Thanks for your work though :)
Yep i think you are pretty much there. The reason to go with y+ close to 30 is the outer region of applicability of the wall function depends on the freestream Reynolds number. This is why you will sometimes see 30 < y+ < 200 and sometimes 30 < y+ < 300. Best to go close to 30 if you are using this method 👍
thank you so much...
When using SST k-omega model and keeping y+ less than 3, does wall functions play any role in calculating the solution. Please share your views.
If y+ is less than 3 you should be in the viscous sub-layer and wall functions shouldn’t affect the solution 👍
@@fluidmechanics101 thank you
Hi Aidan.... Now I am model and it is about a carparking jet fans system.... Can you tell me if I can use the Eulerian model to simulate it or lagrangian model... in the end I have to find the ppm (particles per million) of the CO (carbon dioxide) in the air that produced from an automobile.
Thank you for your time and have a nice day 😊
Hi Dr.Aidan, what do you mean by adverse pressure gradients?
An adverse pressure gradient is a term that aerodynamacists used to describe pressure which increases in the streamwise direction. An aerofoil for example has a favourable pressure gradient on the leading edge, as pressure decreases when the flow moves around the maximum thickness of the aerofoil. On the trailing edge the pressure increases as the flow moves along the aerofoil. This is an adverse pressure gradient and is important because it is one of the main causes of boundary layer separation at high angles of attack. I hope this helps!
Muchas gracias.
Enhanced wall treatment is for k epsilon model for an internal pipe flow having turbulent nature shall have y plus in between sir? What is the maximum y plus value in simulations?
Great work and thank you very much. If you could sir talk a little bit slowly 😊that can help to more understand
I will try my best 😊
You could reduce the playback speed to 0.75x
Amazing, thank u!
hello.. please make video about RANS equation
Hey hi, this is very informative . Can make one video on different boundary conditions used in cfd in details.
great job
Hi Aidan, thank you for the video, I really appreciate it, like all the others.
I have a question though. I am using CFX and k-omega SST, to simulate centrifugal fans at low velocities. Having a yPlus equal to 1 is not affordable for my cases so I need to live with higher values.
Therefore my question is this, correct me if I am wrong:
when y Plus is lower than 5 you get a wall resolved solution and you do not apply a wall function and it is the best case (k-omega);
when y Plus is higher than 30 and lower than 300, CFD code uses a wall function as in k-eps
But what happens if I have yPlus in between 5 and 30? Should I avoid this situation?
I am asking you this because some ANSYS experts told me that I can be fine with a y Plus lower than 300, and a good blending is performed even when yPlus is in the range between 5 and 30.
So I try to stay with a yPlus lower than 300, but it happens in my cases to have yPlus in this range between 5 and 30 and this situation is not very easy to be avoided.
Thanks in advance.
Yep, your ANSYS experts are correct. Try and stay below 300. If you can avoid going between 5 and 30, perfect 👍 y+ of 1 is great if you can afford it
@@fluidmechanics101 Perfect thank you for the answer. Anyway, I had a look at the CFX documentation, and an Automatic Wall Treatment is applied, when using the SST model. This is a yPlus insensitive approach, which performs a good blending between the Low-Re formulation (wall resolved solution) applied for yPlus lower than 2 and the Wall Function formulation applied for yPlus higher than 20. I was asking this because sometimes I find separation on the blades of the fan and this situation needs to be handled with care.
I suppose that also with this CFX treatment, the velocity profile of the wall function is not imposed on the mesh, but it is just applied to correct the wall shear stress, maintaining a linear variation between the velocity of the centroid of the first cell and the velocity at the wall (=0), as you explained very well in your videos.
Thanks for the answer and for the precious guidelines!
@@maurotortora5317 hello sir, i am simulating a centrifugal fan with fluent ,i am using k -omega sst model, i would like to ask you about the seperation in the blades how can you find out if there is a separation in the blade.thanks in advance.
@@aissaamour6276 hello, you just need to plot the streamlines on a plane at constant span
There are a variety of ways of looking for separation. You can look at streamlines, wall shear stress or plot some velocity profiles normal to the wall. Either way, you need to do some investigation in the post processor!
what is the difference between enhanced wall treatment and enhanced wall function?
Enhanced wall functions is for k omega and k omega SST models. Enhanced wall treatment is for k epsilon model
@@fluidmechanics101 yeah, I have seen that in Fluent. But is there any difference in Mathematics point of view? Why the different names?
Up in equation 3 and U in equation 7 are the same?
Yes, there are some differences in the mathematics. I haven't had a chance to make a video on this yet though 🙂
What is the difference between enhanced wall function and enhanced wall treatment??
EWF is for k omega SST, EWT is for k epsilon
@@fluidmechanics101 ok thanks Mr. Aidan
Could you make in the future a video about EWT please?
Hello everyone, I run a cyclone simulation with Fluent (mesh 935k) with RSM. It runs fine and I get a solution that resembles experimental data. Y+ is 29. When I check enhanced wall treatment I get an error at 25th iteration:
# Divergence detected in AMG for k: protective actions enabled!
# Divergence detected in AMG for k, temporarily solve with BCGSTAB!
Divergence detected in AMG solver: k
Divergence detected in AMG solver: k
Divergence detected in AMG solver: k
Divergence detected in AMG solver: k
Divergence detected in AMG solver: k
Divergence detected in AMG solver: k
Divergence detected in AMG solver: k
Divergence detected in AMG solver: k
Error at host: floating point exception
Error at Node 0: floating point exception
Error at Node 1: floating point exception
Error at Node 2: floating point exception
Error at Node 3: floating point exception
Error at Node 4: floating point exception
Error at Node 5: floating point exception
Error at Node 6: floating point exception
Error at Node 7: floating point exception
Error: floating point exception
Error Object: #f
Any ideas? Thnx in advance!
Hi Matt, it is unlikely that the divergence is caused by the enhanced wall treatment alone. Have you checked all aspects of your simulation? Some basic ideas to check: what is your mesh orthogonality? What have you used for the initial condition? Have you checked your boundary conditions / user defined functions? You could always run the case with some basic settings to get initial convergence, the turn on EWT and run another 100 iterations to get the solution you want.
@@fluidmechanics101 Thnx for your instant answer. It is a cyclone simulation and the mesh orthogonality is bad in the area where the tangential inlet meets the cyclone cylinder, but hasn't caused problems with other models. I would like you to clarify me something. I run, for example 10000 iterations RSM with std_wall. If I set the enh_wall and run another 100 or 500 iterations, the results will represent the enh_wall_fn? I mean I don't have to run allover from the beginning with enh_wall_fn? Furthermore, it looks funny, but it isn't, your tutorials have been proved more useful than any help I can get from engineers in Greece (referring to CFD always). Thanks again!
Yep, you don't need to start again. You can just run another 500 iterations and you should have your solution. This trick of getting convergence with an easy simulation (1st order upwind, standard wall functions etc) first and then turning on more detailed models is really useful in CFD and has lots of uses, particularly in detailed models. It is super easy to do in Fluent as well with the TUI as you can run a set number of iterations, then turn on a model and then run 500 more iterations