I have a question regarding adding a boundary layer to the moving wall which I would consider as ground. Why do we add a Boundary layer on the ground? Because in reality, the ground would be stationary contrary to what we simulate.
Also, I have read in the literature that when cars are actually run in a wind tunnel, extra methods are used to remove or avoid Boundary Layer.
I also asked one of our Professor who has a good knowledge of Fluid Dynamics. He also suggested that there is no need to apply a boundary layer on moving ground in simulation
I believe it’s fair to say that it’s more important to set layers when we have a no-slip wall than a moving wall, simply because the gradients are going to be much different:
E.g. if we use large values this will be very obvious. If we have a freestream velocity of 1000 m/s and a no-slip wall, then we have a massive acceleration in the near-wall region. If we don’t have layers here, the parameters will be completely off close to the wall.
If we have a moving wall, then everything will be close to 1000 m/s already and the gradients won’t be so steep. Regardless, I don’t really see any harm in adding more resolution to the near-wall.
So, from your post, I have understood that we should add Boundary Layer on the moving as it will add resolution to that wall. Does that mean the boundary layer (the vicious layer, etc) won’t actually form on the moving wall? (please correct me if I am wrong)
This is a good question. I would assume that a boundary layer is always present on a surface (as in the air particles touching the surface have a frictional force). Once something passes over a surface (a car moving over the ground) the shear force between air particles creates resistance or drag on the car.
For a moving floor in a simulation i am not sure if this is designated as a non-slip surface or not. I would again assume yes so any forces acting on the surface can be considered.
You must also remember that the air underneath the car will be accelerated and squished between the car and floor, which means there is a pressure difference between your pre-set moving floor speed and the air speed next to it.
i hope this helps you understand the need for a boundary layer on a moving floor.
Hi there @dschroeder, thanks for stopping by and helping in this.
As you said,
I have another question brewing in mind and that is how can we calculate the thickness of boundary layer? As for the moving ground, we don’t have a reference length required to calculate thickness.
Another thing you said
I guess someone from SimScale staff can answer this doubt of ours because I am curious about that too.
Also, regarding the sims i am running, I am having some problems with convergence for which I am making another post. Hope you can help there too…
Quite interesting topic, I admit. I will definitely use BLs on moving wall (or rotating wall), as with car wheels we have high pressure gradients on or near interfaces. For that reason we need better mesh in those zones. That means BL in those regions are of similar finesse as BL for wheels. Other part of moving wall could have ‘extended’ mesh cells as pressure gradients are low (and we can save simulation time by reducing mesh).
I do not have strict rule for that, but I experimented with ‘Ahmed body’ simulation and ‘Tesla truck’ project I did some time ago. Hence it is just a heuristic resulting from many rounds of simulation (TET mesh).
My opinion on this is that there is not definitive answer because there must always be a balance in simulation between accuracy and speed. You obviously dont want, and most of the time cant, run a massive 30 million cell mesh quickly. So you need to reduce mesh quality in some areas so that the simulation doesnt take 30hours or fails altogether.
My advice is to just use the length of the car for the reference. that is what i use for the reference length in the coefficient of friction result control. Also i feel that its best to put more cells where you are trying to measure. If you want to look at the impact of ground effect and diffuser function → then add more boundary layers to the floor to make it more accurate. If you are looking at the rear wing, then you can reduce the floor BL cells because it wont have a big impact on rear wing results.
As Retsam said:
Also for your other question
I think Retsam means that Boundary layer cells for the moving floor in unimportant areas can be thicker or ‘extended’ to reduce the overall amount of cells. For example: The areas further in front or behind the car can be bigger and less accurate because these areas are not important to measure, then approaching the car you want more cells to increase the accuracy.
Just remember that everything in simulation is a trade off between accuracy and speed .
I realized the importance to pay attention to meshing correctly a geometry (and applying BL with care), when we tried (@DaleKramer, @Ricardopg and me) to get 100% matching results on NACA0012 airfoil profile, expecting that correct simulation have to be as good as tunnel tests executed by NASA in the past. We spent many weeks tweaking mesh parameters, BLs (1 - 64 layes), Numerics, practically everything. At that moment TET mesh was proposed to become a ‘standard mesh’ and I put a ‘pressure’ on it, running ‘quasi 2D’ simulations dozens of times, for all angles tested i tunnel.
In a kind of intuition I applied a ‘patch’ on the profile surface (adding few small faces) and making BL more dense there while compared to default automatic TET mesh. And that was the ‘trick’ which allowed me to get results much more acceptable (92-96%, for lift and drag [drag being very difficult to match tunnel results]). Here is the example of NACA0012 TET mesh: you can see strange 2-layers BL near the LE, but not on the LE.
At that moment, pandemia came and we all abandoned that study. I regret I left it without descriptions and necessary comments, as I believe it would help others to understand the importance to observe pressure gradients and try better BLs and mesh in those zones.
I admit that adding special faces to the geometry is kind of overkill (in that case justified by whish to be in the vicinity of real results). Automatic TET mesh is trying to do that already. As Don states, you should realize a trade-off: keep it simple as possible, run enough simulations to grasp the flow behavior and you can survive with your core-hours budget for a year or more.
Thank you guys for opportunity to share some knowledge!
Thank you for the insight on this doubt of mine. I have understood surely why we use boundary layer on the ground. Also, the post you have made will take a little time for me to read and understand as it’s a lot to take in actually!
And Thanks @Retsam for sharing your knowledge on BLs. I am really new to this topic and perhaps I have still a lot more to read and learn on BLs and their proper usage. Also, can you post the link to this NACA sim, just wanted to check it out of curiosity ? Thanks in advance
OK @TheDound26: I added you to ‘public’ which can view and copy that NACA0012 (unfinished) project. Here it is: [NACA0012 airfoil study]. Being new (and exited) brings benefits to the community. You are welcome!
Thanks in advance