To not leave the "Limited mesh resolution" just hanging as is, grid independence analysis is often used. Essentially it's running the case with a coarser grid, and comparing the results. From that difference we estimate what would happen if the mesh was more refined than in our main case. If the differences are not that dramatic, the mesh is considered good enough. All in all - nice vid!
First of all, I appreciate your use of the Virdis color map :) One comment about the LES, it is strongly dependant on your mesh, such that the mesh be fine enough be to resolve 80% of your turbulent eddies/turbulent knetic energy. Therefore, at the back of the trucks, where probably a recirculation zone exists, you could increase mesh refinement by adding a refinement box in SHM to better calculate this region, this would effect could effect flow detachment and therefore your drag results. But doing this on a home computer is not easy, I understand, I suffer the same when running my some projects at home. Also I think it would be very interesting to do a 2D analysis with RANS since the truck shapes are more or less consistent in you x-direction. The downside is that you cannot have nice animations of the flow with voriticies with LES (since it only works for 3D meshes). And I think usually you need a really low y+ (
I realize this issue with LES that is why I used the RANS for drag prediction. The comment is much appreciated and I have noted the suggestions, Thank you
@@mayurmahale3049 No, it depends on the mesh. Depending on your application, LES would require at least 2-3 orders of magnitude more cells to get similar results to a RANS sim with a good mesh. Really, I would expect
@@killjbieberplz I understand the mesh requirement. It needs to be small enough to capture atleast 80% of the energy cascade using the grid and depend on the sub grid modelling to handle the rest. I also have heard in cases like external flows SA model gives you close enough values without being unduly expensive.
For the F150, I'd like to see the drag with a toneau cover over the bed and another without a tailgate. I'm assuming the tailgate is increasing the drag
I tried to reproduce the animation from the Github source code and couldnt't. The solvers complete but I can't get the volumetric rendering of the air velocity just like in the video. Do you have an additional paraview plugin setup?
Hey very interesting video ! My only complain is about your rythm when talking, its a bit monotonous and lacked energy, otherwise presentation is clear !
Did I hear that correctly? No inflation layers? My understanding is that aero sims without inflation layers are a total waste of time. You are missing the boundary layer effects.
So I have some thoughts. I'd be more interested in the aerodynamic coefficients than the total force. Why not use a symmetry plane to cut down on the number of nodes/elements? To really resolve the boundary layer and get accurate data you'll need a pretty fine mesh and multiple inflation layers on the models. I noticed your windspeed is around Mach 0.38. Compressibility becomes a factor at about Mach 0.3 so does your model actually have kinematic similarity to the full scale? Does the force and drag coefficient on the brick match what you should get for a real one?
Great comment! - by normalizing the shapes the ratio of Cds is proportional to the ratio of total drag. - the remarks on the mesh and domain have been noted - I am using an incompressible solver, so I should be fine in terms of mach number and kinematic similarity - I haven't checked but in the future I will do further experiments trying to match wind tunnel data
@@interfluo6420 So if that simulation was perfect it should exactly match the same results in a windtunnel. But if you ran the test the real air would start to become compressible (slightly) and the simulation won't reflect that. Like you said it assumes an incompressible fluid. I would guess it would introduce some level of inaccuracy. I've never run any simulations with that high of windspeed but that's just my thought.
I loves your video. I'm busy studying Numerical thermoflow and it's nice to see an example of how the theory I'm learning can be applied. What pc species do you recommend for simulations like theses, because currently I have a 7 year i7 laptop and I was wondering if I should invest in a beefier laptop
Thank you for the tutorial :) When i run your cybetruck simple foam project i get following error: --> foam fatal error: invalid wall function specification. Patch type for patch topX must be wall. Current patch type is patch. Do you know the cause of this ? Thank you so much for the reply
edit: i found issue. Wrong boundary conditions, all types were selected as "patch". After correcting them to patch type "wall" and leaving only inlet and outlet as "patch" now simulation is running fine.
moral of the story, cover the cargo area of the truck and avoid eddys. its not super genius or anything. just common sense. more streamlined. f150 can do that in a day. the fact they dont is upto their choice. u can remove the cover in tesla too to fit in longer objects and the drag for that would be shitty too. i am not pro tesla or pro ford. just my opinion that tesla is not that cool as they are marketing it to be.
To not leave the "Limited mesh resolution" just hanging as is, grid independence analysis is often used. Essentially it's running the case with a coarser grid, and comparing the results. From that difference we estimate what would happen if the mesh was more refined than in our main case. If the differences are not that dramatic, the mesh is considered good enough. All in all - nice vid!
mesh refinement studies will be the topic of an upcoming video. Glad you liked it!
Thank you for the clear and honest explanation. Very informative, and I think you are a great communicator.
Great video. I found the explainations just right in depth and speed. The result is kind of hilarious
Awesome video and thanks for sharing the case files! Helped me solve an error :). I now finally have my first working LES simulation! :D
Glad it helped!
Wow this really shows the low comparitive drag of the cybertruck
I also love the professionalism in your videos. New subscriber here!!
Thanks for the support! It is truly appreciated
I find you videos helpful and inspiring. Subscribed 😇
Welcome aboard!
Very nice work!
Thank you!
First of all, I appreciate your use of the Virdis color map :) One comment about the LES, it is strongly dependant on your mesh, such that the mesh be fine enough be to resolve 80% of your turbulent eddies/turbulent knetic energy. Therefore, at the back of the trucks, where probably a recirculation zone exists, you could increase mesh refinement by adding a refinement box in SHM to better calculate this region, this would effect could effect flow detachment and therefore your drag results. But doing this on a home computer is not easy, I understand, I suffer the same when running my some projects at home.
Also I think it would be very interesting to do a 2D analysis with RANS since the truck shapes are more or less consistent in you x-direction. The downside is that you cannot have nice animations of the flow with voriticies with LES (since it only works for 3D meshes). And I think usually you need a really low y+ (
I realize this issue with LES that is why I used the RANS for drag prediction. The comment is much appreciated and I have noted the suggestions, Thank you
@@interfluo6420 but isn't rans even worse at drag prediction? As a friend puts it RANS is a many to one mapping
@@mayurmahale3049 No, it depends on the mesh. Depending on your application, LES would require at least 2-3 orders of magnitude more cells to get similar results to a RANS sim with a good mesh. Really, I would expect
@@killjbieberplz I understand the mesh requirement. It needs to be small enough to capture atleast 80% of the energy cascade using the grid and depend on the sub grid modelling to handle the rest. I also have heard in cases like external flows SA model gives you close enough values without being unduly expensive.
Put the analysis first and caveat it with the instructions are at the end please. Cool video.
For the F150, I'd like to see the drag with a toneau cover over the bed and another without a tailgate. I'm assuming the tailgate is increasing the drag
Nice video, which filter is applied on paraview for that cool postProcessing? Is it a clip by scalar or something like that?
I tried to reproduce the animation from the Github source code and couldnt't. The solvers complete but I can't get the volumetric rendering of the air velocity just like in the video. Do you have an additional paraview plugin setup?
Hey very interesting video ! My only complain is about your rythm when talking, its a bit monotonous and lacked energy, otherwise presentation is clear !
Noted!
Good stuff
Thanks for the visit
Did I hear that correctly? No inflation layers? My understanding is that aero sims without inflation layers are a total waste of time. You are missing the boundary layer effects.
So I have some thoughts. I'd be more interested in the aerodynamic coefficients than the total force. Why not use a symmetry plane to cut down on the number of nodes/elements? To really resolve the boundary layer and get accurate data you'll need a pretty fine mesh and multiple inflation layers on the models. I noticed your windspeed is around Mach 0.38. Compressibility becomes a factor at about Mach 0.3 so does your model actually have kinematic similarity to the full scale? Does the force and drag coefficient on the brick match what you should get for a real one?
Great comment!
- by normalizing the shapes the ratio of Cds is proportional to the ratio of total drag.
- the remarks on the mesh and domain have been noted
- I am using an incompressible solver, so I should be fine in terms of mach number and kinematic similarity
- I haven't checked but in the future I will do further experiments trying to match wind tunnel data
@@interfluo6420 So if that simulation was perfect it should exactly match the same results in a windtunnel. But if you ran the test the real air would start to become compressible (slightly) and the simulation won't reflect that. Like you said it assumes an incompressible fluid. I would guess it would introduce some level of inaccuracy. I've never run any simulations with that high of windspeed but that's just my thought.
I loves your video. I'm busy studying Numerical thermoflow and it's nice to see an example of how the theory I'm learning can be applied.
What pc species do you recommend for simulations like theses, because currently I have a 7 year i7 laptop and I was wondering if I should invest in a beefier laptop
I use a desktop I built myself, it is a fairly standard setup (3600x, 2060 super, etc..), computer configuration is highly application dependent
Hi, great video. I do have one question though, why do you multiply the z force by the pressure when calculating drag?
I don't, I multiply p by norm_z (component of the surface normal in the z-direction) to get the force in the z
What was the characteristic length for Re
Thank you for the tutorial :) When i run your cybetruck simple foam project i get following error: --> foam fatal error: invalid wall function specification. Patch type for patch topX must be wall. Current patch type is patch. Do you know the cause of this ? Thank you so much for the reply
edit: i found issue. Wrong boundary conditions, all types were selected as "patch". After correcting them to patch type "wall" and leaving only inlet and outlet as "patch" now simulation is running fine.
This is fantastic, for giggles can you run this at 100mph?
glad you liked it, I will be doing some higher velocity simulation in upcoming videos
What about the smooth underbody of an electric vehicle vs the obviously not smooth underbody of the conventional vehicle?
I don't think he'd have the time or algorithms to map that out
How many elements did these mesh have?
right around 1.5 million
Would it be okay if I posted your work on Twitter?
yes, feel free to share however you like
moral of the story, cover the cargo area of the truck and avoid eddys. its not super genius or anything. just common sense. more streamlined. f150 can do that in a day. the fact they dont is upto their choice. u can remove the cover in tesla too to fit in longer objects and the drag for that would be shitty too. i am not pro tesla or pro ford. just my opinion that tesla is not that cool as they are marketing it to be.
6 minutes before the damn video starts.
tru. i skipped to the end lol
Honestly. Just Talk like your normal self. Instead of poorly reading an overly done up script.
Bro it's super hard to even hear your voice......
noted, I am actively working on my audio