¿Es correcto usar el mismo mallado del dominio computacional adyacente a la pared, para tubo hidráulicamente liso que para hidráulicamente rugoso, solo cambiando el parámetro de rugosidad en la pared?. Is it correct to use the same meshing of the computational domain adjacent to the wall, for hydraulically smooth pipe and for hydraulically rough pipe, only changing the roughness parameter on the wall?
Hi Lina, physically the roughness elements should always be smaller than the cell (ks < yp). If the roughness elements are larger than the cell then really we should be making our mesh around the actual elements themselves! I hope this makes sense :)
Hi Aidan, thank you so much for your videos. I've watched many of them and they are super helpful! About this topic, I have a question: we need y+ = y*u_tau/nu to calculate flow velocity, but how do we get u_tau (friction velocity) in the first place without knowing the flow velocity?
You can either use a local iteration, starting with an initial guess of y+ = 11 and the Newton-Raphson algorithm, or you can just use the wall shear stress from the previous iteration. The first of these approaches is used by OpenFOAM in 'nutUWallFunction.C'
Hi Aidan, great video. A quick question: assuming we have flow over a sphere below the crictical Reynold's number, and assuming we split this sphere into a top half and a bottom half and add a roughness value to the top hemisphere to get rubulent flow on this half. Normal aerodynamic theory would suggest the turbulent flow should separate later, as shown in your video. However, I find that the turbulent side actually separates earlier than the laminar side. Any thoughts on this?
Ah yes. Spheres and cylinders are really hard to get right. They are one of the hardest problems for CFD! So don't worry if you don't quite get the answer you expect. You will need to use a really fine mesh to even get close to the right answer for a smooth sphere. You will probably need y+ ~ 1 and at least 200 cells around the diameter. In terms of roughness modifications, I'm not sure how to do it when you resolve the mesh into the viscous sub-layer. Have you tried contacting ANSYS or Siemens support?
Hi Aidan, thanks for the great content. May I ask what beamer theme and color theme are you using for your presentation? Also, what do you use to draw your figures? Tikz?
I use boadilla theme with a blue colour palette. I have tried to choose the best settings for UA-cam (I now use wide-screen with a slightly larger font) so that people can watch the videos on their phones more easily. For figures, I draw everything inkscape and then import the figures as PDFs into latex 👍 I am always trying to improve the slides so that people can see them more easily on UA-cam from their phones and I will probably keep improving them in future as well 😊
@@fluidmechanics101 Thanks a lot. Love your presentations. Indeed, sometimes I watch them from desktop, sometimes from mobile. The experience is execellent both ways.
Great explanation and presentation Dr. Wirmhurst. Many important questions find their answers through your videos. Please let me know which program are you using to create these amazing sketches?
8:51 It's unclear whether Ks (equivalent sand-grain roughness height) refers to the diameter of the sand grains used, or to the height of protrusion of those grains. Surely, the grains must have some depth of embedment into the substrate. Is this depth ignored?
Hello Aidan, I was looking for some insight into the effect of roughness on channel flow and found your video to be quite helpful. However, I have a question with respect to your statement that the flow gets broader when the roughness is increased (higher velocities close to the wall and lower velocities in the outer flow, slide #5). This is in contrast to my intuition and to other sketches of velocity profiles that compare smooth and rough turbulent flow that I could find online at a first glance. Unfortunately, I haven't found a proper and detailed explanation in a textbook on this issue so far. -- Am I overlooking something here? I would be glad I you could take a second look on this and either verify or rectify this statement. Thanks!
Yes, this can be a bit confusing. The case for rough channel flow and rough flow over a flat plate lead to different shaped velocity profiles as you increase the roughness. Remember that for a flat plate, the freestream boundary condition has U_infty far from the plate (regardless or the roughness) so the profile only really changes close to the wall as you increase the roughness. However, for a channel flow, as the profile changes near the flow it also has to change in the freestream, so that the mass flow rate is conserved. This is why the profiles are different. I hope this helps 😊
@@fluidmechanics101 Hello Aidan, thank you for your quick response! If I may, I'd like to ask a follow-up question because I still feel a bit uncomfortable with the effect of an increase in roughness on the velocity profile in turbulent channel flow. For example, in Nikuradse's work "Laws of Flow in Rough Pipes" the following statement can be found on page 24: "As the relative roughness increases, the velocity distribution assumes a more pointed form." - which should correspond to a velocity decrease close to the wall, or am I on a wrong track here?
hi, excelent video!.. thank you.. one doubt i have is that from Eqn9 to Eqn10, is the RHS of Eqn9 fully multiplied by 1/k?.. i.e. U_plus = 1/k*(log() - delB)?.. because i feel like without that, Eqn10 cannot be achieved.. plz explain..
Please do videos on finite element methods concepts too. I bought your courses in udemy too. They are very good. If you can provide some insights into FEM concepts through courses or through youtube, it would be very helpful
The wall function approach takes into account roughness by shifting down the loglaw curve. It obviously has an upper limit on the roughness height. Do you have any suggested approach to model the very rough wall surface?
Yes, I came across this during my PhD. There seems to be a limit for how rough you can make the surface. A trick which you can use when the wall cant get any rougher is to swap the boundary condition from ' no slip wall with roughness' to 'applied shear stress'. You can then ramp up the shear stress as high as you want to mimic really large roughness. This is a trick which is often used in Tidal energy calculations to mimic large boulders and rocks on the sea bed. I would have a look at the thesis by Conor Fleming 'Tidal Power Extraction in the Offshore Environment' for details (Google search)
Sorry for this dumb question. At 7:31 you said as U+ increase, wall shear stress increase and that gives broader profile. Do you mean U+ decrease, wall shear stress increase? Could you explain a little.
Yes! U+ is the dimensionless velocity at the cell centroid. If you think of a linear velocity profile between the cell centroid and the wall (where the velocity is zero), if you increase the velocity, the gradient increases. Because the wall shear stress is viscosity * velocity gradient, this increases the wall shear stress. It might help if you drew a picture of it. That normally helps me
Hey, how is roughness accounted for if we don't use wall functions? Let's say I want to model viscous sublayer with linear profile where wall functions are not used.
Hello ,Great Video As always (y) and i have a question too "as Always" , All the video you ve been talking about roughness but only took in consideration the fact that we are In log-Law region which means this rough wall function is only available for 30
Yes, this is an issue which i havent managed to figure out yet. I think we can only use the roughness model with y+ in the range of 30 < y+ < 200. Im not sure what we can use with y+ < 1. Let me get back to you on that one!
Thanks a lot! Very helpful video. Do anybody know how to get Ks+ distribution on the wall in Fluent and CFX? Is it included in variables list (like y+) or I need to use expression to calculate Ks+?
You can model your geometry as a flat plate. Then use a correlation for the skin friction coefficient (cf). The wall shear stress is then cf * 1/2 rho U^2 👍
Sorry for another dumb question. 7:16 you said "shear stress here represented by u sub tau". u sub tau is velocity not wall shear stress, correct? Frankly, ut =(wall shear stress/rho)^0.5 which means ut increase wall shear stress increase.
Yep, you are correct. Recall that we are trying to find a velocity scale that is representative of the flow at the wall (for normalisation). We can't use the freestream velocity or the pipe centreline velocity or a cross sectional velocity of some type because we are close to the wall. The velocity = 0 at the wall, so we can't use the wall velocity either. The square root of the wall shear stress (divided by density) has the correct units and turns out to be a good scale for us to do our normalisation. We could have chosen something else (like the square root of turbulent kinetic energy) but all that would happen is the experimental data would just be plotted slightly differently and our correlations would be slightly different. I hope this helps
Hi Aidan. Thank you for your super helpful videos. I pointed out a mistake in the derivation of E' (i.e. the modification to E in the log-law equation when taking into account the roughness effect). I guess the correct relationship is E' = E/exp(k*Delta_B), rather than E' = E/exp(Delta_B) (I'm not quite sure if you've derived it yourself or borrowed it from a reference, but you've probably missed multiplying the Von Kármán constant k). I leave a link for you to see my derivation, so please correct me if I'm wrong. drive.google.com/file/d/1s_IQTNcHAnEHTCsr_pG4GBeMVoyVqBcF/view?usp=sharing
Hi Aidan, really appreciate your videos - very clear explanations that have helped me refresh my CFD theory knowledge. Do you have any recommendations for implementation of roughness without wall functions (i.e. with the k-w SST model?) Ive found a few sources that touch on the issue and mention a wall shift but im still unclear. For example, if I were comparing two cases of smooth and rough pipe flow and I have a y+ ~ 1 for the smooth case, what happens if I then apply roughness? It doesn't make sense to me to have y+ < ks+ but I have been told that (in fluent) the treatment is automatic. Many thanks.
Hi Phil, this is something I am still not clear about either. It is not clear what to do if y+ ~ 1 and the roughness elements are the same order height as your first cell! Sorry about this one but i dont know the answer yet!
Hello Phil, yes in Ansys Fluent it is possible to calculate a case with k-w SST model and take surface roughness into account. And, it works somehow, but I still did not figure out how. To compare the CFD results with y+ lower than 5 or around the value of 1, I conducted several experiments for the flow in a pipe and I got very close to the CFD results with roughness. Currently, I am struggling with comparisons in subsonic and supersonic flow, becauce it seems to have a different behaviour in each of these cases. Just a little remark.
Great! Thanks for your additions Dmitry. I hadn't really looked into the treatment of omega with roughness effects. I'm sure this will be useful for people using models for roughness
![[CFD] What is the difference between y+ and y*?](http://i.ytimg.com/vi/nSdVaF3JnI0/mqdefault.jpg)
![[CFD] What is the difference between y+ and y*?](/img/tr.png)
![[CFD] Enhanced Wall Functions in ANSYS Fluent](http://i.ytimg.com/vi/h5OiFpu0L4M/mqdefault.jpg)
![[CFD] What is Mesh Non-Orthogonality?](http://i.ytimg.com/vi/Dg2PwX6yxYY/mqdefault.jpg)
![[CFD] What Wall Functions Do I Need for Turbulent Kinetic Energy?](/img/n.gif)
I have search the exact difference between Ks and Cs and finally I get it. Great work, excellent channel. Greetings from Colombia.
These are so good, I always seem to find the perfect one for what I’m working on
¿Es correcto usar el mismo mallado del dominio computacional adyacente a la pared, para tubo hidráulicamente liso que para hidráulicamente rugoso, solo cambiando el parámetro de rugosidad en la pared?. Is it correct to use the same meshing of the computational domain adjacent to the wall, for hydraulically smooth pipe and for hydraulically rough pipe, only changing the roughness parameter on the wall?
When a fine near-wall mesh is created to achieve y+=1 for use with SST or k-w,
what to do when rough walls are applied?
Thank you for this video Aidan! It is very explanatory video. I would like to hear from you how AMG solver works :) Thank you a lot!
fun fact: Nikruadse carried out his experiments back in 1933! Rest in peace! (Thanks Aidan - highly appreciated!)
Great working, well done. It was really helpful.
Great explanation in this video as well, thank you very much!I was wondering if it makes sense to have a rough wall function with ks>yp. Thank you.
Hi Lina, physically the roughness elements should always be smaller than the cell (ks < yp). If the roughness elements are larger than the cell then really we should be making our mesh around the actual elements themselves! I hope this makes sense :)
Ah, I've seen your Masters Thesis Lina. Really nice to have such a CFD community :)
@@fluidmechanics101 finally i can explain to my supervisor why I need the rough bed haha
Hi Aidan, thank you so much for your videos. I've watched many of them and they are super helpful!
About this topic, I have a question: we need y+ = y*u_tau/nu to calculate flow velocity, but how do we get u_tau (friction velocity) in the first place without knowing the flow velocity?
You can either use a local iteration, starting with an initial guess of y+ = 11 and the Newton-Raphson algorithm, or you can just use the wall shear stress from the previous iteration.
The first of these approaches is used by OpenFOAM in 'nutUWallFunction.C'
Hi Aidan. Thank you for this useful video. I have a question. What happens when the size Ks is bigger than the size of the adjacent wall cell (Yp>Ks).
The roughness correction tends towards an asymtopic value, so there is a limited increase in the wall shear stress as you make KS larger and larger
Hi Aidan, great video. A quick question: assuming we have flow over a sphere below the crictical Reynold's number, and assuming we split this sphere into a top half and a bottom half and add a roughness value to the top hemisphere to get rubulent flow on this half. Normal aerodynamic theory would suggest the turbulent flow should separate later, as shown in your video. However, I find that the turbulent side actually separates earlier than the laminar side. Any thoughts on this?
Ah yes. Spheres and cylinders are really hard to get right. They are one of the hardest problems for CFD! So don't worry if you don't quite get the answer you expect. You will need to use a really fine mesh to even get close to the right answer for a smooth sphere. You will probably need y+ ~ 1 and at least 200 cells around the diameter. In terms of roughness modifications, I'm not sure how to do it when you resolve the mesh into the viscous sub-layer. Have you tried contacting ANSYS or Siemens support?
Hi Aidan, thanks for the great content. May I ask what beamer theme and color theme are you using for your presentation? Also, what do you use to draw your figures? Tikz?
I use boadilla theme with a blue colour palette. I have tried to choose the best settings for UA-cam (I now use wide-screen with a slightly larger font) so that people can watch the videos on their phones more easily. For figures, I draw everything inkscape and then import the figures as PDFs into latex 👍 I am always trying to improve the slides so that people can see them more easily on UA-cam from their phones and I will probably keep improving them in future as well 😊
@@fluidmechanics101 Thanks a lot. Love your presentations. Indeed, sometimes I watch them from desktop, sometimes from mobile. The experience is execellent both ways.
Great explanation and presentation Dr. Wirmhurst. Many important questions find their answers through your videos. Please let me know which program are you using to create these amazing sketches?
I create all my sketches in inkscape. Inkscape is free and you can use it on linux mac or windows. I would highly recommend it 😊
8:51 It's unclear whether Ks (equivalent sand-grain roughness height) refers to the diameter of the sand grains used, or to the height of protrusion of those grains. Surely, the grains must have some depth of embedment into the substrate. Is this depth ignored?
Hello Aidan, I was looking for some insight into the effect of roughness on channel flow and found your video to be quite helpful. However, I have a question with respect to your statement that the flow gets broader when the roughness is increased (higher velocities close to the wall and lower velocities in the outer flow, slide #5). This is in contrast to my intuition and to other sketches of velocity profiles that compare smooth and rough turbulent flow that I could find online at a first glance. Unfortunately, I haven't found a proper and detailed explanation in a textbook on this issue so far. -- Am I overlooking something here? I would be glad I you could take a second look on this and either verify or rectify this statement. Thanks!
Yes, this can be a bit confusing. The case for rough channel flow and rough flow over a flat plate lead to different shaped velocity profiles as you increase the roughness. Remember that for a flat plate, the freestream boundary condition has U_infty far from the plate (regardless or the roughness) so the profile only really changes close to the wall as you increase the roughness. However, for a channel flow, as the profile changes near the flow it also has to change in the freestream, so that the mass flow rate is conserved. This is why the profiles are different. I hope this helps 😊
@@fluidmechanics101 Hello Aidan, thank you for your quick response! If I may, I'd like to ask a follow-up question because I still feel a bit uncomfortable with the effect of an increase in roughness on the velocity profile in turbulent channel flow. For example, in Nikuradse's work "Laws of Flow in Rough Pipes" the following statement can be found on page 24: "As the relative roughness increases, the velocity distribution assumes a more pointed form." - which should correspond to a velocity decrease close to the wall, or am I on a wrong track here?
Hi Aidan! Does the 'log' in your ppt represent 'ln' instead of log(10, c)?
Yep, should be ln
hi, excelent video!.. thank you.. one doubt i have is that from Eqn9 to Eqn10, is the RHS of Eqn9 fully multiplied by 1/k?.. i.e. U_plus = 1/k*(log() - delB)?.. because i feel like without that, Eqn10 cannot be achieved.. plz explain..
Yes, I think this was a typo. Thanks for spotting it!
Great video Aidan! Do you also have the any reference of how deltaB is calculated in openFoam?
It is calculated in exactly the same way as in Fluent, so i normally just reference the Fluent manual 👍
Very useful.
Please do videos on finite element methods concepts too. I bought your courses in udemy too. They are very good. If you can provide some insights into FEM concepts through courses or through youtube, it would be very helpful
Yep, i will see what i can do 😄
Hello, it is an amazing well explained video. However, I have a question and it is about the log function you are using. Is it loge or log10?
I will have to check. I am pretty sure it is loge 👍
The wall function approach takes into account roughness by shifting down the loglaw curve. It obviously has an upper limit on the roughness height. Do you have any suggested approach to model the very rough wall surface?
Yes, I came across this during my PhD. There seems to be a limit for how rough you can make the surface. A trick which you can use when the wall cant get any rougher is to swap the boundary condition from ' no slip wall with roughness' to 'applied shear stress'. You can then ramp up the shear stress as high as you want to mimic really large roughness. This is a trick which is often used in Tidal energy calculations to mimic large boulders and rocks on the sea bed. I would have a look at the thesis by Conor Fleming 'Tidal Power Extraction in the Offshore Environment' for details (Google search)
Thanks sir
Sorry for this dumb question. At 7:31 you said as U+ increase, wall shear stress increase and that gives broader profile. Do you mean U+ decrease, wall shear stress increase? Could you explain a little.
Yes! U+ is the dimensionless velocity at the cell centroid. If you think of a linear velocity profile between the cell centroid and the wall (where the velocity is zero), if you increase the velocity, the gradient increases. Because the wall shear stress is viscosity * velocity gradient, this increases the wall shear stress. It might help if you drew a picture of it. That normally helps me
Hey, how is roughness accounted for if we don't use wall functions? Let's say I want to model viscous sublayer with linear profile where wall functions are not used.
Hello ,Great Video As always (y) and i have a question too "as Always" ,
All the video you ve been talking about roughness but only took in consideration the fact that we are In log-Law region which means this rough wall function is only available for 30
Yes, this is an issue which i havent managed to figure out yet. I think we can only use the roughness model with y+ in the range of 30 < y+ < 200. Im not sure what we can use with y+ < 1. Let me get back to you on that one!
@@fluidmechanics101 Thank you very much for the informations .
@@fluidmechanics101 is this the case? My project is similar and involves modelling roughness at y+
Thanks a lot! Very helpful video. Do anybody know how to get Ks+ distribution on the wall in Fluent and CFX? Is it included in variables list (like y+) or I need to use expression to calculate Ks+?
Hi Aidan, I have to ask, how do you estimate wall shear stress in order to determine the Ks+? Thanks!
You can model your geometry as a flat plate. Then use a correlation for the skin friction coefficient (cf). The wall shear stress is then cf * 1/2 rho U^2 👍
Sorry for another dumb question. 7:16 you said "shear stress here represented by u sub tau". u sub tau is velocity not wall shear stress, correct? Frankly, ut =(wall shear stress/rho)^0.5 which means ut increase wall shear stress increase.
Yep, you are correct. Recall that we are trying to find a velocity scale that is representative of the flow at the wall (for normalisation). We can't use the freestream velocity or the pipe centreline velocity or a cross sectional velocity of some type because we are close to the wall. The velocity = 0 at the wall, so we can't use the wall velocity either. The square root of the wall shear stress (divided by density) has the correct units and turns out to be a good scale for us to do our normalisation. We could have chosen something else (like the square root of turbulent kinetic energy) but all that would happen is the experimental data would just be plotted slightly differently and our correlations would be slightly different. I hope this helps
Hi Aidan. How do it calculate E? Does it use U+ and y+ of the "smooth wall" -▲B in the log law equation modified for roughness?
E is a constant 🙂 you can look it up with a quick Google search
@@fluidmechanics101 thank you !!
Hi Aidan. Thank you for your super helpful videos. I pointed out a mistake in the derivation of E' (i.e. the modification to E in the log-law equation when taking into account the roughness effect). I guess the correct relationship is E' = E/exp(k*Delta_B), rather than E' = E/exp(Delta_B) (I'm not quite sure if you've derived it yourself or borrowed it from a reference, but you've probably missed multiplying the Von Kármán constant k). I leave a link for you to see my derivation, so please correct me if I'm wrong.
drive.google.com/file/d/1s_IQTNcHAnEHTCsr_pG4GBeMVoyVqBcF/view?usp=sharing
Ah yes, i think i missed the k .... 🤦♂️ well spotted!
Hi Aidan, really appreciate your videos - very clear explanations that have helped me refresh my CFD theory knowledge. Do you have any recommendations for implementation of roughness without wall functions (i.e. with the k-w SST model?) Ive found a few sources that touch on the issue and mention a wall shift but im still unclear. For example, if I were comparing two cases of smooth and rough pipe flow and I have a y+ ~ 1 for the smooth case, what happens if I then apply roughness? It doesn't make sense to me to have y+ < ks+ but I have been told that (in fluent) the treatment is automatic. Many thanks.
Hi Phil, this is something I am still not clear about either. It is not clear what to do if y+ ~ 1 and the roughness elements are the same order height as your first cell! Sorry about this one but i dont know the answer yet!
Hello Phil, yes in Ansys Fluent it is possible to calculate a case with k-w SST model and take surface roughness into account. And, it works somehow, but I still did not figure out how. To compare the CFD results with y+ lower than 5 or around the value of 1, I conducted several experiments for the flow in a pipe and I got very close to the CFD results with roughness. Currently, I am struggling with comparisons in subsonic and supersonic flow, becauce it seems to have a different behaviour in each of these cases. Just a little remark.
Great! Thanks for your additions Dmitry. I hadn't really looked into the treatment of omega with roughness effects. I'm sure this will be useful for people using models for roughness
very good as always!
Sir how can we generate surface roughness in circular microchannel?