2024 Scuderia Ferrari F1 Team

[ferrarifire]

2026 power units are lot simpler than what it is today , i believe all the competitors will meet the benchmark and the differences will be very small . Post 2010 , the way f1 operates have completely changed . It is becoming a tech/engineering war rather than people caliber .

Dont want to go in details , In short i have few connections in mercedes hpp , obviously they know about their competitors SWOT more than what is available in public forum . Now more or less ferrari have the required ingredients to success and lets see how it is getting unfolded in the future . This also indirectly answers the question about Lewis switch . f1 racing is a very small closed group and everyone knows where the wind is moving .

Lots of interesting points there; out of curiosity, how do you know that regarding Ferrari?
I don’t remember any article writing so specifically about that.

Thanks.
That can point also to the fact that Ferrari’s 2026 power unit is going to be at the top of the stakes (as premature as it can be now)?

[ing.]

Note, Mercedes is struggling for the same reason. Their HPP division uses outdated software and simulations (I can vouch for this as my friend works there), and they won't recover until they update their simulation systems and resolve their on-track correlation issues. Of course, they won't publicly acknowledge this.

How much do you expect Merc will gain from recruiting Ferrari simulation guy and does Ferrari have the depth to easily back-fill his role?

[Seanspeed]

Please tell me when exactly McLaren pressured Ferrari in any of the races this season. I can't remember.

Mclaren were literally a very close competitor in just the last race and absolutely required Ferrari to work hard to overtake them. Norris would have been a much bigger threat for 3rd had they not put him on the strategy they did. Do we really need to hash out obvious information? I've got no interest in this kind of time wasting.

I think you'll agree with me if I tell you that, with all respect, I don't expect them to outdevelop Ferrari.

Ferrari have a long history of entering the season with a great car and then failing to execute well on developments throughout the season. You can expect all you want for whatever reasons, but we simply dont know what's going to happen, and little is guaranteed. If the sport was honestly that predictable, I probably wouldn't even watch it.

[ferrarifire]

It is not going to be that easy..every team has different type of simulation tools and the runs are different...newly recruited guy can help in optimising the process but usually system overhaul typically takes about 2-3 years of cycle..

Note, Mercedes is struggling for the same reason. Their HPP division uses outdated software and simulations (I can vouch for this as my friend works there), and they won't recover until they update their simulation systems and resolve their on-track correlation issues. Of course, they won't publicly acknowledge this.

How much do you expect Merc will gain from recruiting Ferrari simulation guy and does Ferrari have the depth to easily back-fill his role?

[deadhead]

It is not going to be that easy..every team has different type of simulation tools and the runs are different...newly recruited guy can help in optimising the process but usually system overhaul typically takes about 2-3 years of cycle..

How much of the simulation aspect is software based? I can't help but think that machine learning (AI) should be able to help quite a bit in that aspect of the sport. From what we've seen on track, one would assume that RBR have the best simulation tools on the grid. Does silicon valley (Oracle) have anything to do with this? I just wonder.

[aleks_ader]

Ai is not for simulations. This is physics problem not quazi AI mimicks...

[Emag]

Data Science certainly is a powerful tool. If we assume all teams have the same set of data available, what you pull out of it and how your correlate all the different variables can definitely make a difference.

It may lead you to pursue wrong development paths if you misinterpret the data and make the wrong assumptions on possible culprits for weaknesses.

[Venturiation]

Ai is not for simulations. This is physics problem not quazi AI mimicks...

You can teach the ai the physics based on your data and cfd
and ask it to make a part of the car by trying every millions of shapes possible and pick the best one
Then you test that part in the CFD, then wind tunnel

[ferrarifire]

Redbull uses Ansys simulation ( california based company) for the simulation cfd, dyno and mi in conjunction with Oracle oci..

It is not going to be that easy..every team has different type of simulation tools and the runs are different...newly recruited guy can help in optimising the process but usually system overhaul typically takes about 2-3 years of cycle..

How much of the simulation aspect is software based? I can't help but think that machine learning (AI) should be able to help quite a bit in that aspect of the sport. From what we've seen on track, one would assume that RBR have the best simulation tools on the grid. Does silicon valley (Oracle) have anything to do with this? I just wonder.

[deadhead]

Redbull uses Ansys simulation ( california based company) for the simulation cfd, dyno and mi in conjunction with Oracle oci..

It is not going to be that easy..every team has different type of simulation tools and the runs are different...newly recruited guy can help in optimising the process but usually system overhaul typically takes about 2-3 years of cycle..

How much of the simulation aspect is software based? I can't help but think that machine learning (AI) should be able to help quite a bit in that aspect of the sport. From what we've seen on track, one would assume that RBR have the best simulation tools on the grid. Does silicon valley (Oracle) have anything to do with this? I just wonder.

This forum is like a gold mine if you like the world of motorsports.

Thank you!

here a bit of AI stuff from their website

https://www.ansys.com/blog/optimize-des ... tamodeling

https://www.ansys.com/blog/simulation-a ... ent-slide4

I am going to guess that some of that is being applied in F1 or it will eventually

[ferrarifire]

All these tools comes very handy during the race week..post p1,p2 practices the teams work round the clock running multiple simulations over night to arrive at the qualify setup based on the practice data...this is the reason many teams get it wrong during sprint weekend as they don't ve relevant data to fine-tune the setup

Races are won on the track, but championships are won at the factory. The team generates up to 50 TB of CFD data per week prior to race events by running millions of simulations. To accommodate the team’s workload, high-performance computing systems are needed.

Redbull uses Ansys simulation ( california based company) for the simulation cfd, dyno and mi in conjunction with Oracle oci..

How much of the simulation aspect is software based? I can't help but think that machine learning (AI) should be able to help quite a bit in that aspect of the sport. From what we've seen on track, one would assume that RBR have the best simulation tools on the grid. Does silicon valley (Oracle) have anything to do with this? I just wonder.

This forum is like a gold mine if you like the world of motorsports.

Thank you!

here a bit of AI stuff from their website

https://www.ansys.com/blog/optimize-des ... tamodeling

https://www.ansys.com/blog/simulation-a ... ent-slide4

I am going to guess that some of that is being applied in F1 or it will eventually

[ferrarifire]

This is not true ...for instance Ferrari uses AWS to create an AI and machine learning program with the virtual speed sensor using Amazon SageMaker(Machine Learning Service) . This sensor provides valuable data without needing additional physical hardware on the race car and read realtime from the track ..

F1 teams moved its CFD (Computational fluid dynamics ) simulation environment to a high-performance computing (HPC) platform using ML /AI and has reduced CFD simulation time by 80%, from 60 to 12 hours. They can now schedule a simulation to run at night and have the results the next morning. With faster results, They can perform more simulations in general and get to the final car design faster.

Ai is not for simulations. This is physics problem not quazi AI mimicks...

[dialtone]

This is not true ...for instance Ferrari uses AWS to create an AI and machine learning program with the virtual speed sensor using Amazon SageMaker(Machine Learning Service) . This sensor provides valuable data without needing additional physical hardware on the race car and read realtime from the track ..

F1 teams moved its CFD (Computational fluid dynamics ) simulation environment to a high-performance computing (HPC) platform using ML /AI and has reduced CFD simulation time by 80%, from 60 to 12 hours. They can now schedule a simulation to run at night and have the results the next morning. With faster results, They can perform more simulations in general and get to the final car design faster.

Ai is not for simulations. This is physics problem not quazi AI mimicks...

This is a bit too much buzzword bingo in the same sentence.

Ferrari was transmitting telemetry from the car to their data centers a decade ago. With Amazon they simply ship it to Amazon. They probably use SageMaker for some stuff but having worked in this area it’s probably more about strategy than CFD, primarily because I doubt they wrote a python notebook to run in sagemaker, or use the standard models available on it. On top of that, if they really generate 50TB per race, besides it costing a literal fortune, it wouldn’t even scale that well unless their models are very simple.

Lastly for CFD they are probably using their software to run those simulations as CFD is very sensitive to network latency which with SageMaker is kinda unavoidable. I presume they could run CFD on a few big boxes in aws, but they probably run that stuff locally.

Actually lastly: all of CFD time is capped in computing power so whatever drop in time is due to better algorithms or more specific use.

[aleks_ader]

I dont believe in AI and its buzzwords. Sure it can develop maybe new algorithms for ground effect approximations or whatever. I dont doubt maybe they can extract so "assinine" raw data from Friday session and maybe provide some unorthodox solution referring to the setup.

But if your understanding of PHYSISCS IS BAD, NO AI will help u beacuse u dont understand it. Nothing will beat physics first principle laws and accurate data collection... Also nah teams dont have budget for huge dataset training. I would rather pay better programmers and involve anyss (or whoever) and make modification and optimisations there. Also nothing will beat that skillfull CFD engineer. how to make multiple snapshots of simulations and nice planning how to make multiple sims on "partial" data.

[Xyz22]

https://www.the-race.com/formula-1/mcla ... explained/

Thinking back 12 months ago, when Ferrari was quite openly admitting it wasn’t that fussed that Sanchez was leaving, the summary from various conversations The Race had was: ‘he’s a great technical mind, but he doesn’t play well with others’.

Interesting.