Corrado physics question.

[veedub 0]

No, none here either!

 

I still think Purple is faster!!

[Yandards 0]

Yes on the mountains near Inverness yesterday. (Snow that is)

 

If you look at the cross-section of the spoiler then you will see that it is not wing shaped at all, big curve at the back, as such it was never intended or designed to function like an inverted wing to produce downforce. It will produce a tiny amount of downforce due to the fact that it is a flat surface protruding into the airflow but the greatest effect will be the disruption of the airflow over the rear of the car which will only decrease lift by creating turbulence and thus drag.

 

The Corrados spoiler produces a swirl effect at the back of the car hence the reason they get covered in crud below the spoiler.

 

Dec interesting sounding project, wonder if it would be worth producing a curved insert to fit on the underside of the spoiler.. ;)

[dr_mat 0]

Yes on the mountains near Inverness yesterday. (Snow that is)

 

Really? Cool. Someone gets a white christmas. If you happen to live up a mountain near inverness, anyway..

 

Dec interesting sounding project, wonder if it would be worth producing a curved insert to fit on the underside of the spoiler.. ;)

 

Doubt it, since the angle of attack of the spoiler compared to the airflow is quite high, I suspect that you'll find it's past the stall angle, so could never be used as a wing..

[Dec 1]

You cant just say, that because something produces turbulence it will decrease lift!....it's far more complicated than that!

 

And the curve at the back of the spolier wont have a huge effect on the spoliers characteristics. The flow will seperate fropm the lower surface of the spolier at the front edge creating a sort of virtual flow surface below it.

 

All aerodynamic forces can only act on a surface. And every surface in a flow will produce some sort of force.

 

Turbulence behind the car cant affect either lift , drag or downforce as there is no surface of the car there for it to act on!

[Dec 1]

Dec interesting sounding project, wonder if it would be worth producing a curved insert to fit on the underside of the spoiler.. ;)

 

Ok....I've been trawling through my CAD files and the closest model I can find to the Corrado is a Laguna. Both hatchback, so flow over the rear should be similar!

 

I can setup any sort of spolier desings you can think of on the back of it (or anywhere else you may like) and see what the outcome is!

 

Just post up a rough pic of it and I can draw it up!.....Results will take about a day to get back depending on computer load! :wink: :wink:

[davidwort 0]

whale tail :)

[dr_mat 0]

You cant just say, that because something produces turbulence it will decrease lift!....it's far more complicated than that!

 

Turbulence behind the car cant affect either lift , drag or downforce as there is no surface of the car there for it to act on!

 

Indeed, it's not that simple, but the net result is slower airflow over the top of the car, which results in the reduction in lift. (Due to the differing air pressure above/below the car etc etc ..)

 

Apparently...!

[Dec 1]

You'd think I'd have something productive to do in work...but it's pretty slow coming up to Christmas...so!

 

Pic is of a basic 2d CAD sketch of the profile of my rado (In purple for the true believers)...I basically have converted it to 3d, stuck a few wheels underneath it, and processed it through various aerodynamic software packages (50K per years worth of someone elses software being misused here!!! :lol: :oops: :oops: )

 

Anyways, the calculations are up and running pretty well.....should get some results later on tonight, or tomorrow morning.

 

It's only a pretty basic model, but should show the main flow features and produce some nice streamline pics! 8)

[HiAsAKite 0]

Cool-! did my 4th year project on cfd, albeit turbine blade modelling with a bit of laminar to turbulent flow transition point modelling for good measure...

 

mind you I didn't have 50Ks worth of software to model it on- I had to write my own using Fortran77..

 

those were the days eh?-unfortunately now do nothing connected with that (butit does mean I can afford my 'rado :D )

[Dec 1]

We use Fortran for the most accurate simulations that the expensive software cant deal with!

 

Much more efficient and generally better, but takes an age to set-up properly..and our code is useless for modeling large things like a whole car!

[HiAsAKite 0]

that would follow- ... and to be honest the point of my project was more to prove it could be modelled using a given set of techniques, rather than designing the perfect turbine blade...

 

..anyway- hope your run converges tonight... unlike when I'd come in on a Monday morning and find mine had crashed at 5.45 on the Friday....

[Dec 1]

Well here's a partially converged solution on a pretty rough grid...so not exactly super accurate. But good enough to see what the flow is doing.

 

Shows pressure coefficient (red = high pressure, green/blue = low pressure), looking from above the front of the 'car' :lol: at just under 70mph

 

You can see the high pressure at the front of the car and the lower side of the windscreen (drag), and the lower pressure green/blue regions as the flow accelerates over the bonnet and roof of the car (lift). And if you look at the back, positive red region of pressure acting on the upper surface at the rear windscreen/spoiler area. Which means there is a negative lift/downforce/pressure pushing downwards on that area!

[Dec 1]

And a side profile basically showing the speed of the airflow over the car.

 

Apologies for the dodgy Corrado model :oops:

 

That big red region above the roof is accelerated, low pressure flow which causes your lift. The flow under the car is also accelerated, lowering the pressure beneath the car, but not to the same extent as the upper side.

 

Once I let the sims finish running, I can make a quick change and put the spoiler in more of an 'up' position and see what happens! Or maybe even get back to the orig question and lower it a bit!....I'd need a far better model and grid to really determine any drag reductions tho. (assuming I don't have any real work to be doing) :roll: :roll:

[HiAsAKite 0]

Cool!- in your boundary conditions did you specify that the road surface and freestream velocities were the same wrt to the car or just that the flow immediately adjacent must be parallel to it (ie no road drag)?

 

Then again given the turbulent flow effects between the car and road at speed its not the easiest thing to model accurately - how good is your pacakge in this respect?

 

Going back to the cd debate I seem to recall seeing somewhere that the corrado CD was 0.32- in practice however you need to include the frontal area in any 'which car is more aerodynamic' debate as this is just as significant.

 

 

nice to see some CFD outputs anyway..:)

[Dec 1]

yeah, the road surface is set to be moving at 30m/s immediately below the car.

 

The cfd package is pretty good at pretty basic models like this. Unfortunately it would take a long time to do a decent 3d cfd model on a shape even as basic as this car model. The grid I've used is too coarse (you can see how coarse it it by looking at the boundary layer around the car) and to reduce computational time, the viscous model I used is also pretty simple.

 

It's not accurate enough to correctly predict the lift/or drag on a Corrado, but it's good enough to basically show what the airflow is doing as it passes over a car shaped body!

 

As I said, if anyone happens to have a proper CAD model of a C, I could do a full test on it!.....it would probably take about a month to get decent results though! 8)

[dr_mat 0]

We use Fortran for the most accurate simulations that the expensive software cant deal with!

 

Much more efficient and generally better, but takes an age to set-up properly..and our code is useless for modeling large things like a whole car!

 

Fortran is nice and efficient because it's very limited. It's great for maths, cos that's all it does. And the net result is if you compile -O3 or -O4 the compiler can turn your dodgy algorithms into streamlined machine code.

 

The C compiler, for example, can't make anywhere near as many optimisations simply because C is so much closer to the machine code anyway.

 

I used to write fortran at Uni ..

 

Of course, in the real world, people like McLaren use StarP to do their math modelling (on a bunch of Altix machines). I don't think they do much algorithm coding in fortran any more.

[dr_mat 0]

..it would probably take about a month to get decent results though!

 

What are you running it on?

[Dec 1]

The machine its currently running on is a 56 processor (xenon) cluster. I don't exactly want to take the P*ss and take up the entire machine for something like this though...so it's only currently running on 4 of those processors. Problem isn't how long it takes to run each simulation, it's how many times you have to tweak each of those simulations and re-run them to get a good result!

 

We use Fortran as all the commercial packages essentially only model the flow features, they don't truly calculate them. We have a 6th order accurate DNS Fortran based code (pretty much as accurate as available anywhere) for measuring nanoscopic flow features that the likes of StarCD, Fluent etc simply wont even know exist! It can only be used for tiny sections of a model though due to its high accuracy!

[Jim 2]

Awesome! There is always someone on this place that knows something about ANY given subject! Look forward to seeing further models / results (spoiler, lowered, etc!)

[dr_mat 0]

The machine its currently running on is a 56 processor (xenon) cluster. I don't exactly want to take the P*ss and take up the entire machine for something like this though...so it's only currently running on 4 of those processors.

 

Ah, ok. Politically awkward to use all the CPUs.. :)

McLaren use IA64 processors (and lots of them). In fact, a lot of the F1 teams seem to use IA64. Must be something to do with that 8GFlops/core of 128-bit floating point goodness..

[veedub 0]

Good work fellas! Im well impressed!

[Dec 1]

 

Ah, ok. Politically awkward to use all the CPUs.. :)

McLaren use IA64 processors (and lots of them). In fact, a lot of the F1 teams seem to use IA64. Must be something to do with that 8GFlops/core of 128-bit floating point goodness..

 

Yup, might get caught too!....nobody will notice 4 little processors doing something random!

 

A few friends of mine that I did my masters with, work for Mclaren and Red Bull F1 now...not sure exactly what sort of computers they use for their cfd work though.

 

The in house computers we use are only really used for 2D work...for the big 3D jobs we have remote access to the computers at the national supercomputer centre!....no idea what processors are in there either..but they are pretty d&rn quick!! :shock: :shock: :shock:

[dr_mat 0]

The national supercomputing resource at manchester uni? That's a pissy cluster now, it used to be proper machines, proper scalable machines..

The PC cluster rules the world now. Why make it well when you can make it cheap and do the rest in software?

[Jim 2]

Everything comes down to cost these days. And for sheer $ / MHz, off the shelf stuff annhilates SGI kit (which is what i'm assuming you're referring to as 'proper' stuff) :)

[Dec 1]

Nah, the current one being used is the High Performance Computing Facility owned by Cambridge/Cranfield Universities! :wink:

 

The clusters we use in here are basically a bunch of stripped down high speed PC's designed to run jobs in parallel......they are actually far quicker than the 'proper' big machines at small cfd jobs...as the big machines tend to almost choke themselves and never get up to their full speed on small simulations! :roll: