Mercedes AMG F1 W08 EQ Power +

[F1NAC]

how much advantage is this oil burning? is it only useful in qualy or more useful in the race?

Quali mode

Getting bigger Baaaang for shorter time

[pimpwerx]

Some nice shots of the details on W08:

https://imgr3.auto-motor-und-sport.de/M ... 113157.jpg

Via AMuS

I shared this particular image with my FB today, because while I got a computer engineering degree thanks to a scholarship, I was and always be a mech/aero engineer at heart. I was on the chassis/suspension team for our uni's Formula SAE team. The manhours required to sculpt these elements is completely lost on the layman, but my engineer friends understand the CFD investment.

I had a crazy nostalgic moment on FB following Lewis' pole record, and had to share this. I remember sitting in a CFD lecture from a Ford engineer, explaining the number of hours needed to design a consumer car intake plenum. This image puts all that in perspective. This is truly the pinnacle of motorsports, folks. Like if Picasso or Bach were race engineers.

[Zynerji]

How much of this is currently done by AI iterations on their super clusters now?

As far as I understand, only CFD is capped by regs, not AI CAD design.

[Sevach]

[henry]

How much of this is currently done by AI iterations on their super clusters now?

As far as I understand, only CFD is capped by regs, not AI CAD design.

What do you think the AI would be iterating towards? If the end goal is aerodynamic performance, what measure might it use if it can't use CFD? Or is the AI given the surfaces but iterates to some other goal such as weight or stiffness?

I'm not trying to trip you up here. I too have heard that Mercedes used some form of AI on their W floor and I'm genuinely interested in how it might be employed.

[Zynerji]

Honestly, I would use a bi-directional ray tracer, and change the "bounce" math with Navier-Stokes math.

Then you start with the basic box of the car, tell it what ray densities you would like in certain areas (pressures), and let it iterate until it converges.

I spoke at length with Dade from LuxRender, and he said this is not simple, but absolutely possible. It was beyond his scope for that project, but he did give me his blessing to modify it to this end.

The bi-directional nature of a ray tracer would be the key to this, as CFD is not bi-directional (when I was looking at this in 2010).

[J.A.W.]

how much advantage is this oil burning? is it only useful in qualy or more useful in the race?

Both.. see Hamilton's post safety car run.. & the advantages are manifold.. with various potentials such as..
..fuel burn catalyst carrier/additional energy as defacto fuel/mech-pressure sealing/friction reduction/& etc..

[siskue2005]

how much advantage is this oil burning? is it only useful in qualy or more useful in the race?

Both.. see Hamilton's post safety car run.. & the advantages are manifold.. with various potentials such as..
..fuel burn catalyst carrier/additional energy as defacto fuel/mech-pressure sealing/friction reduction/& etc..

but the rule only take effect from next week onwards, which means that Vettel also had engine which can burn that much oil. So ur point is not valid.

[basti313]

how much advantage is this oil burning? is it only useful in qualy or more useful in the race?

Both.. see Hamilton's post safety car run.. & the advantages are manifold.. with various potentials such as..
..fuel burn catalyst carrier/additional energy as defacto fuel/mech-pressure sealing/friction reduction/& etc..

but the rule only take effect from next week onwards, which means that Vettel also had engine which can burn that much oil. So ur point is not valid.

With being on the side of off-topic, but maybe ok here as customers do not have the Q-mode: Ferrari had a less well applied technique with extra oil in an extra oil tank. This is all banned. For Merc only the amount of oil changes. So there should be no change on the Q-mode. But the ~2-4 laps they can go in this mode during the race will be reduced.

[J.A.W.]

how much advantage is this oil burning? is it only useful in qualy or more useful in the race?

Both.. see Hamilton's post safety car run.. & the advantages are manifold.. with various potentials such as..
..fuel burn catalyst carrier/additional energy as defacto fuel/mech-pressure sealing/friction reduction/& etc..

but the rule only take effect from next week onwards, which means that Vettel also had engine which can burn that much oil. So ur point is not valid.

"...ur point is not valid."?

Ziggurat you cannot be serious.. oleaginous usage protocols are the "...point..." - if only lately applied as.. rules.

[Big Mangalhit]

Both.. see Hamilton's post safety car run.. & the advantages are manifold.. with various potentials such as..
..fuel burn catalyst carrier/additional energy as defacto fuel/mech-pressure sealing/friction reduction/& etc..

but the rule only take effect from next week onwards, which means that Vettel also had engine which can burn that much oil. So ur point is not valid.

With being on the side of off-topic, but maybe ok here as customers do not have the Q-mode: Ferrari had a less well applied technique with extra oil in an extra oil tank. This is all banned. For Merc only the amount of oil changes. So there should be no change on the Q-mode. But the ~2-4 laps they can go in this mode during the race will be reduced.

The rule only apllies for engines introduced in Monza and afterwards. Meaning Mercedes will not be affected by this new directive, and probably promptly introduced the engine to circumvent it

[itwasntme]

but the rule only take effect from next week onwards, which means that Vettel also had engine which can burn that much oil. So ur point is not valid.

With being on the side of off-topic, but maybe ok here as customers do not have the Q-mode: Ferrari had a less well applied technique with extra oil in an extra oil tank. This is all banned. For Merc only the amount of oil changes. So there should be no change on the Q-mode. But the ~2-4 laps they can go in this mode during the race will be reduced.

The rule only apllies for engines introduced in Monza and afterwards. Meaning Mercedes will not be affected by this new directive, and probably promptly introduced the engine to circumvent it

https://it.motorsport.com/f1/news/consu ... 45275/?s=1

[basti313]

With being on the side of off-topic, but maybe ok here as customers do not have the Q-mode: Ferrari had a less well applied technique with extra oil in an extra oil tank. This is all banned. For Merc only the amount of oil changes. So there should be no change on the Q-mode. But the ~2-4 laps they can go in this mode during the race will be reduced.

The rule only apllies for engines introduced in Monza and afterwards. Meaning Mercedes will not be affected by this new directive, and probably promptly introduced the engine to circumvent it

https://it.motorsport.com/f1/news/consu ... 45275/?s=1

I was fooled by this one too. But this whole page of words is based only on one sentence of Charlie Whiting, which unfortunately tells nothing about engine Nr.4 in the Merc.

[wesley123]

How much of this is currently done by AI iterations on their super clusters now?

As far as I understand, only CFD is capped by regs, not AI CAD design.

Last year I believe only their barge board package visibly benefited from machine learning(I hate the term AI as it is not true AI), this excluded the elements under the tub, though.

For this year, I think everything regarding bargeboards from the nose tip onwards benefits from machine learning. The base parts with the large bargeboard are probably designed by a human, but any further detail is machine learning. The thing that for me really gave it away was that on a previous iteration of the floor, the second or third outermost tooth was like 10mm shorter than the rest of the surrounding teeth that were all the same length. There is no way that in a limited formula that an engineer would come up with such a insignificant detail.
The developments also weren't done over a course of the season. It went from a regular package to this full-detailed iteration in a single development.

The Ferrari and Mercedes are really out of their league compared to the rest of the field, the other cars don't even come close to the terms of detail.

What do you think the AI would be iterating towards? If the end goal is aerodynamic performance, what measure might it use if it can't use CFD? Or is the AI given the surfaces but iterates to some other goal such as weight or stiffness?

I'm not trying to trip you up here. I too have heard that Mercedes used some form of AI on their W floor and I'm genuinely interested in how it might be employed.

Dumbed down it is essentially just doing what a human does. The difference is that a computer can do calculations much faster, and thus is capable to come up with more detailed solutions. It is machine learning, so it probably knows the aerodynamic rules and has learned their application on the current car. It can see results to a much more detailed degree and can do so much quicker than a human could.

[henry]

Honestly, I would use a bi-directional ray tracer, and change the "bounce" math with Navier-Stokes math.

Then you start with the basic box of the car, tell it what ray densities you would like in certain areas (pressures), and let it iterate until it converges.

I spoke at length with Dade from LuxRender, and he said this is not simple, but absolutely possible. It was beyond his scope for that project, but he did give me his blessing to modify it to this end.

The bi-directional nature of a ray tracer would be the key to this, as CFD is not bi-directional (when I was looking at this in 2010).

Are you suggesting an alternative, non conventional, CFD method? So that the machine design algorithm can evaluate aerodynamic outcomes as it iterates designs?

Essentially a typical F1 "cheat", it's CFD, but not conventional CFD so it doesn't count to the processing power and time restrictions.

I think I can see how you might model the space and use reflectance and transmission functions using Navier Stokes but I think to get into that would be very off topic for this thread.