Khamsin Virtual Racecar Challenge 2015

[variante]

Beautiful article, Mantium. Thank you for sharing.

About your diffuser: I think the reason why it works could be found in the compression/decompression process and its ground effect interactions. As the incoming air approaches the diffuser's "protuberance" it slows down in its compression stage, far away from the ground; energy "recovery" is fullfilled during the decompression stage, near to the ground. It should be also noticed that the pressure differential generally reaches higher values during the decompression phase. As a third observation, in a well designed geometry potential energy of the compression phase can be stored mostly in form of drag (instead of lift) and released in form of downforce.
I think that your current design lacks a little bit of the optimization to fully take advantage of this last point. Still it looks very promising.

A similar device is the "pelican nose" pioneered by Lotus. I believe it works in the very same way.

[julien.decharentenay]

Thanks a lot to all participants. Congratulation to Variance for the championship and Mantium for the race. Big thanks to Richard and well done for the video.

@mantium It looks like you will be a serious competitor if you are going down the automated optimisation path. Any experience with adjoint solver? I am keen to get my hands on something there that I can play with - I have to admit that I am struggling with the derivation of the equation source term... It was one my plan for this year - but did not get the time.

Feel free to drop Chris or myself an email any time. We will be in touch via the KVRC 2016 discussion. Have a good break and see you next year.

[machin]

Beautiful article, Mantium. Thank you for sharing.

About your diffuser: I think the reason why it works could be found in the compression/decompression process and its ground effect interactions. As the incoming air approaches the diffuser's "protuberance" it slows down in its compression stage, far away from the ground; energy "recovery" is fullfilled during the decompression stage, near to the ground. It should be also noticed that the pressure differential generally reaches higher values during the decompression phase. As a third observation, in a well designed geometry potential energy of the compression phase can be stored mostly in form of drag (instead of lift) and released in form of downforce.
I think that your current design lacks a little bit of the optimization to fully take advantage of this last point. Still it looks very promising.

A similar device is the "pelican nose" pioneered by Lotus. I believe it works in the very same way.

Yeah, I've been thinking of these "double inflection diffusers" (of which this is just an extreme example) in terms of velocity and momentum: where the velocity is highest (where the floor is closer to the ground) accelerating the air away from the ground (creating downforce) creates a larger momentum change than where the velocity is lowest and the air accelerates back towards the ground (the "arch" part), which must produce some lift. Hence net vertical force in the diffuser region is in the downward direction.

If you had freedom in the rules, I was wondering if you could design a complete floor with a "wavy bottom" (!!!!), so you'd end up with a series of low pressure peaks along the whole floor and not just at the inlet and exit? Presumably at some point there would be too much pressure loss and air would start to avoid going under the floor and take some easier route instead... But that would be down to selection of the optimum "waviness" (in terms of number of waves and their radius of curvature in both the "throat" and "arch" areas).

[LVDH]

@mantium It looks like you will be a serious competitor if you are going down the automated optimisation path.

That was the plan for this year.

Any experience with adjoint solver?

Yes, I do. Adjoint is very tricky, though. I used to work with a quite good adjoint solver. Besides the big challenge of obtaining good results you have to be able to use them. Normally you run into the big problem of manufacturing. Here we are mainly constrained by the rules. CAESES is actually able to read adjoint sensitivities and use them for the geometry modification while also obeying constraints. So maybe that would be the way to go for this competition.

It was one my plan for this year - but did not get the time.

Same here. Earlier this year I was working on a process using SU². I hope I will have some time later this year. Next year I might be able to show some results on the car.

Yeah, I've been thinking of these "double inflection diffusers" (of which this is just an extreme example)...

OK, sounds like I did not invent anything new (as always). Could you point me to some examples or even papers?

If you had freedom in the rules, I was wondering if you could design a complete floor with a "wavy bottom" (!!!!), so you'd end up with a series of low pressure peaks along the whole floor and not just at the inlet and exit? Presumably at some point there would be too much pressure loss and air would start to avoid going under the floor and take some easier route instead... But that would be down to selection of the optimum "waviness" (in terms of number of waves and their radius of curvature in both the "throat" and "arch" areas).

I am more than happy to investigate how this thing works. Maybe we could do that as a community in the forum? That way I would not have to do all the work alone.
First we could create a simplified geometry. Who else would be interested in this?

[machin]

Don't get me wrong; I've never seen it done like you have soley in the "diffuser region" before, but essentially you've applied the "double inflection diffuser principle"* to its extreme.

* This is basically: turn the air sharply away from the ground when the velocity is highest (zone 1 in picture A) to create large downforce, and turn the air back towards the ground when the velocity is lowest (zone 2 in pic A) to minimise the lift generated in that portion. The result is more downforce than you would achieve by only turning the air gently away from the ground across the whole "diffuser region" (pic B)

If you look at a longitudinal section of pretty much all the KVRC cars (pic C) you can see this principle is already applied at 3 points; under the front wing, at the "entry" to the floor, and at the exit of the floor ("the diffuser"). Your innovation (I'm not taking that away!) is to add a fourth by squeezing two low pressure peaks into the "diffuser region" (pic D) .... My suggestion is whether you could add more in the floor area (pic E)?

[machin]

@mantium It looks like you will be a serious competitor if you are going down the automated optimisation path.

That was the plan for this year.

It does sound very interesting... Can you give us a brief run-down on how the automatic optimisation works?

I presume its not as simple as giving the optimisation program your current model and it just goes off and makes the car better?!

I presume you need to draw your car parametrically, i.e. for the rear wing you might assign values to the chord length, camber, and angle of attack at various points along the width. To do the optimisation do you then choose which parameters to vary and by how much each time and the total amount and then the optimisation program tests all those configurations...?

....Or do you need to model the two "extreme" configurations and the computer tests various incremental steps between those two configurations?

Both of those would still need a human operator to work as the "ideas man".... I.e. its a "time-saving" process rather than an "ideas generating" process?

[RicME85]

With regards to the full length wavy bottom idea....it would be only possible in a section to the outer edge of the floor.


Green is the flat floor guide, yellow is what defines the height of the diffuser and then the blue on the edge has a max height of 100m with a length of 2053mm and a width of 300mm (missed that off the image)

I have looked into creating something like this in that section but couldnt get it to fit well between the wheels.

[pierlu99]

Hi, I'm from Italy and i love f1 and more particularly the aerodynamics!I often read the forum but today, when I found this thread, I decided tu sumbit immediately! . I use sketchup from 3/4 Years and and I'd like to know more information about KVRC.
I would also like to participate if possibile..

Sorry for My English..

[MadMatt]

Congrats to everyone, well done! Also congrats to the KVRC team, I guess it was quite hard work to look at all these nice models and solve issues and problems!

As for myself, I doubt I will have time to do some CAD anytime soon unfortunately. Freetime is not something I have a lot and I don't want to use it for this I'm afraid But I enjoyed a lot looking at what you guys did and I doubt I would have done any better!

[CAEdevice]

I was wondering if the double diffuser would be safe in a real track. Considering that (it is so in my design, maybe it is the same for Variante, JJR, Mantium) the second stage uses some ground effect in the lower section, what would happen when breaking? The ride height in that section can increase of 10mm or more, and my simulations show that a difference of 10mm influences very much the local ground effect (up to 500N, comparing different heights for the second stage, enough to cause balancing issues).

[machin]

Good question.... It could result in "porpoising" if it is too sensitive to ride height changes...

(I think LVDH (as the inventor) should come up with a name for the arrangement as "double diffuser" has been used before!)

[machin]

With regards to the full length wavy bottom idea....it would be only possible in a section to the outer edge of the floor.
http://i.imgur.com/e9LldVS.jpg

Yeah, I thought it probably wouldn't be possible within the rules (I really should make myself more familiar with them!)

[LVDH]

It does sound very interesting... Can you give us a brief run-down on how the automatic optimisation works?

There are many ways of parametrizing geometries. And this is by far the most difficult part using design space exploration or optimization techniques.
The two main methods are fully parametric and morphing.

The fully parametric approach was used for the wing profiles and in the last race for the full rear wing.
The profile is described as a function dependent of a few parameters. For the last race a wing span was added to the profile with some parameters of the profile changing as a function of the wing span.

Morphing (in CAESES called "Free Form Deformation") can be seen here:
https://www.caeses.com/wp-content/uploa ... al.mp4?_=6

While not entirely morphing this might also help to understand:


The trick for this type of geometry modification is that the geometry has to be made simple enough to ensure that these type of modifications work and do not break anything and fit to the rest of the car.
I did my best work on this when I was still working for an automotive OEM but everything is behind closed doors and not public at all unfortunately.


What I like the most about optimization is that you often see solutions you would not have come up with on your own.
And it was always a dream of mine to invent an aerodynamic device that could be named after me. Maybe the time has come.

(I think LVDH (as the inventor) should come up with a name for the arrangement as "double diffuser" has been used before!)

While I always wanted a device with my name I propose "Mantium Diffuser".
Now it is time to do further investigations and publish a paper. I hope I can get started on that within the next two weeks.

[CAEdevice]

I'm going to make some simulations about the behaviour of the double stage diffuser during breaking with a new interesting tool... See you in a few days

[CAEdevice]

First KVRC2016 car ?