Use of Computer Simulation

[xxChrisxx]

Thread to discuss computer simulation in design.

(hopefully mods should be moving posts that were off topic in another thread to here)

Until then, you are viewing a useless thread

[Fil]

Like the first driver that jumps in for Wet tyres when it starts to rain, Manor GP have done the same with CFD.

It'll be fascinating to see whether the conditions are right for F1 design solely based on CFD, or whether an Intermediate, of Wind Tunnel & CFD, is still the right choice in design.

I'm sure the field will be watching Manor's progress with keen interest.

[Jersey Tom]

It'll be fascinating to see whether the conditions are right for F1 design solely based on CFD

Nope. No way.

[Ciro Pabón]

The mods will move every useless post on simulation here. We do simulations for a living, so we understand how useless it is...

I would like to add that the crux of simulation is the kind you employ: if macrosimulation, based on general equations, or microsimulation, based in the behaviour of the individual units of the medium you are simulating.

Of course, CFD is a macro-simulation, as I understand it: you do NOT simulate the behaviour of individual molecules. That's why there is a perpetual race in CFD to minimize the side of the grid of cubes that represent the points where the "currents" of air interact with each other, through the equations.

As I simulate traffic, the behavior of individual cars can be faithfully reproduced, throwing an random number to represent the aggressiveness of the driver: the behaviour of a driver when he follows another car and the opportunities or gaps he accepts to overtake are well known.

We also use macro-models, where the entire flow is somehow "summarized" in an equation of flow, similar to the ones used in CFD: the cars behave like a gas, where volume and speed vary, depending of the density of "molecules".

I wonder if CFD some day will become micro in that sense. That'll be the day we'll have very fast computers...

[xxChrisxx]

Computers are becomign fast enough now to acutally simulate the entire flow, even the smallest eddies. Unfortunately even the most poweful supercomputer would take weeks/months of computation time.


This arose from the Manor GP CFD thing. Auto made the argument that windtunnels should be banned.

But this would mean no simulation could be validated, and thereforethe results caouldnt be treated with confidence.

Manor's hopes rest on getting a good flow model straight away that correlates with reality. If they screw up and didnt validate they could end up with that 'wtf' moment when the on track performance down't match the simulations.


The posts that need moving were in this thread
viewtopic.php?f=4&t=7645

[dumrick]

Like the first driver that jumps in for Wet tyres when it starts to rain, Manor GP have done the same with CFD.
(...)
I'm sure the field will be watching Manor's progress with keen interest.

Actually, I don't think it is quite so dramatic. A lot of competition vehicles are nowadays made without ever going to the windtunnel, mainly those whose budgetary or time constraints on the project make it impossible to afford a windtunnel.
But even top of the range series vehicles are increasingly being done using only CFD. Wirth Research projected last year's LMP1 Acura ARX-02 entirely relying on CFD data and, apparently, have evaluated the results and concluded they were good enough to embrace a F1 program in the same way.

But this would mean no simulation could be validated, and thereforethe results caouldnt be treated with confidence.

Manor's hopes rest on getting a good flow model straight away that correlates with reality. If they screw up and didnt validate they could end up with that 'wtf' moment when the on track performance down't match the simulations.

That could be also a two-edged sword. Screw-ups on wintunnel calibration just in recent years has affected organisations like Toleman-Benetton-Renault and Tyrrell-BAR-Honda-Brawn-MercedesGP. Just go ask the Honda guys if, then, they wouldn't have liked to rely only on CFD data... The fact is that the tunnel can give you a lot of detail and fine-simulation, mainly when we are dealing with some chaotic flow structures. But, it only takes the calibration to get a tad wrong for the correlation (or lack of) you speak about to start steering the aero teams on the wrong direction.

If we are talking about "computer simulations" in F1, I can't resist to mention that only last week, I saw the news that Force India has purchased a version of rFactor PRO to build a simulator, making them the 5th F1 team to have a rFactor-based simulator (the others being Williams, Red Bull, Ferrari and ...?). Being rumoured that McLaren's simulator is based on proprietary technology, that would make at least 6 F1 teams relying on interactive simulators.
I kind of can have a notion how it is important to go through things like ballast placements with the actual drivers and their driving style reflecting the potential benefits or not of such change. And, the other way, sure it is interesting for the drivers to be able to feel changes like a new diffuser, and test it freely to its simulated limits before they actually drive the very expensive real counterpart. Of course, a team has all the data to model the cars in the software and it seems it just takes make to run that model through the gMotor2 graphic and physics engine to come out something very believeble indeed...

[xxChrisxx]

That could be also a two-edged sword. Screw-ups on wintunnel calibration just in recent years has affected organisations like Toleman-Benetton-Renault and Tyrrell-BAR-Honda-Brawn-MercedesGP. Just go ask the Honda guys if, then, they wouldn't have liked to rely only on CFD data... The fact is that the tunnel can give you a lot of detail and fine-simulation, mainly when we are dealing with some chaotic flow structures. But, it only takes the calibration to get a tad wrong for the correlation (or lack of) you speak about to start steering the aero teams on the wrong direction.

Which is precisely the point. In terms of fidelity to reality it goes

simulation < windtunnel < track data

Windtunnel testing is more reliable than unvalidated CFD data. Of course if you bugger up the windtunnel, then you've automatically buggered up the simulation variables.

This is what happened to Honda in 07.

If we are talking about "computer simulations" in F1, I can't resist to mention that only last week, I saw the news that Force India has purchased a version of rFactor PRO to build a simulator, making them the 5th F1 team to have a rFactor-based simulator (the others being Williams, Red Bull, Ferrari and ...?). Being rumoured that McLaren's simulator is based on proprietary technology, that would make at least 6 F1 teams relying on interactive simulators.
I kind of can have a notion how it is important to go through things like ballast placements with the actual drivers and their driving style reflecting the potential benefits or not of such change. And, the other way, sure it is interesting for the drivers to be able to feel changes like a new diffuser, and test it freely to its simulated limits before they actually drive the very expensive real counterpart. Of course, a team has all the data to model the cars in the software and it seems it just takes make to run that model through the gMotor2 graphic and physics engine to come out something very believeble indeed...

It's impressive just how realistic simulators are, infact its becoming impressively realistic in games these days. GT5 with a Logitech G25, is pretty damn close to the real thing in terms of handling and reaction. The only thing i've found is that the lack of sensation makes it difficult to judge speeds and braking distances. theres always that tendancy to push it too far, because obv you can't get hurt sat staring at a screen.

[Pandamasque]

It'll be fascinating to see whether the conditions are right for F1 design solely based on CFD

Nope. No way.

Let's wait and see what it's Wirth.

[autogyro]

It'll be fascinating to see whether the conditions are right for F1 design solely based on CFD

Nope. No way.

Let's wait and see what it's Wirth.

If Nick does design a brilliant car without the use of a wind tunnel, then others are going to have to up their game by a huge margin or be left behind.

[SZ]

It'll be fascinating to see whether the conditions are right for F1 design solely based on CFD

It'll be fascinating to see when Wirth's budget extends to a decent tunnel testing program.

[autogyro]

Hahaha

We will see.

[tc9604]

Perhaps Manor have a point in that there is so much they *can* do pre-season without tunnel testing, and potentially get a driveable car ready for the season start, although they have already been sensible enough to admit that they won't be challenging the established teams just yet.

I'm sure in future if they prove successful with what they've got they will be keen to get some wind tunnel data. As stubborn/brave/economical as they have been so far, I'm sure they will reach a point where their calculations would need to be verified.

[Jersey Tom]

I'm sure if they had the budget they'd be all over a windtunnel rather than 100% CFD.

Simulation tools, FEA, and CFD are truly great things. But they're also EXTREMELY sensitive to how they're set up and run.

Even with some of the most advanced finite element tools on the planet, there are big limitations and you HAVE to work in conjunction with real world testing. That's just how it is. There are certain things you can't pull out of a computer simulation.

So while I'm sure they might have a decent starting aero platform, I am 99% certain it will not be anything dominant.

[autogyro]

It sounds like there are a lot of not to good people working in F1 aero.
First you say that it is rarely possible to get accurate wind tunnel results, then you say it is also difficult to get accurate computer results. Then back to slagging off wind tunnels.
Seems there is a little bit of prima donna about on this subject.

[SZ]

It sounds like there are a lot of not to good people working in F1 aero.
First you say that it is rarely possible to get accurate wind tunnel results, then you say it is also difficult to get accurate computer results. Then back to slagging off wind tunnels.
Seems there is a little bit of prima donna about on this subject.

There's a common belief that wind tunnels or computational methods are able to simulate the real world with exacting accuracy.

That's horseshit.

Both are environments that seek to simulate various salient aspects of practical flows with varying degrees of accuracy. Neither represents an exacting replication of the practical environment. Of the two, a wind tunnel may be more 'practically' accurate and offer greater test throughput, though there are inherent compromises in what kind of flow interrogation may be obtained. Computational methods are inherently compromised for many flows at feasible levels of computational power, however the opportunities to interrogate and explore flow properties are nearly unlimited.

It is rarely possible to achieve exacting replication of practical flows in either; wind tunnels are in fact setup for freestream flows rarely encountered in the real world, but that are at least exacting in their repeatability. There's a very decent science - bordering on an art form - in quantifying the transferability of tunnel tests to road/track results. Let alone that wind tunnel test methods are constantly evolving.

There are severe limitations to computational methods that require experimental validation to enable best approximation.

At best, experimental or computational methods simulate the practical environment with incredible repeatability. Accuracy is another matter, and there's genius required in interpretation of as much.

No prima donna attitude whatsoever... Jersey Tom and others have some valid points.

Manor can say what they want. There is some development work you simply cannot do feasibly in CFD.

Does anyone have numbers on their computational capacity? Would give some idea of their relevant development scope.