Verne Global

HPC | Manufacturing |

25 January 2018

Within the world of F1, computational fluid dynamics is in pole position

Written by Nick Dale

Nick is Senior Director at Verne Global and leads our work across HPC and specifically its implementation within discrete and process manufacturing. He is based in our London headquarters.

Ever since Lotus discovered the importance of downforce on F1 car performance and built the type 79 car - the most dominant race car of its day - understanding the flow of air around a car has been at the absolute core of race car design and construction. Today, computational fluid dynamics (CFD) is quickly augmenting the use of wind tunnels (which have a number of important shortcomings) to gather data about the aerodynamic efficiency of car designs - a trend that’s become drastically more pronounced since the explosive growth of computational power in the late 1990s.

Robust CFD models can not only better simulate the effects of the tremendous speeds that F1 cars will encounter on the track, but they are also better at accounting for subtle and micro variables that can have an impact on a car’s efficiency, such as how the air around a car and its tyres warms up at high speed. This has created something of an arms race among F1 teams for better CFD systems.

As a result, across the last few years virtually all F1 teams have either built their own dedicated HPC clusters to help produce and design their cars, or else have partnered with a technology services company to help them process their CFD workloads. Notable initiatives include Renault Sport Formula One (formerly Lotus F1), who have at different times worked with both traditional technology vendors like Dell, with whom they constructed a specialised system that included an HPC cluster and rugged solution for trackside testing, and Boeing, with whom they cooperated to optimise both company’s overall CFD capability.

Red Bull Racing — one of the dominant racing teams of the modern era — has also been ambitious about its HPC capacity, recently signing a deal with IBM in order to further increase the scheduling of their CFD design-related workloads. Not to be outdone, current World Champions Mercedes racing signed a partnership deal with analytics and integration specialist Tibco Software, while a number of other teams have worked with ANSYS to develop greater CFD processing capability.

Surprisingly, one of the highest-profile embraces of CFD by an F1 team was Marussia (their car shown below), who appeared on the scene in 2007 with a very modest $30 million in the bank. In a sport where leaders like Ferrari, McLaren and Red Bull annually spend $9 million on just one car and over $200 million to field a team, this was an almost impossibly small budget. However Marussia planned to compensate for this lack of economic clout by going all-in on HPC and CFD technology, and eschewing wind tunnels completely. This was a strategy that no team in F1 had ever tried before.

Marussia F1. (Image courtesy of Marussia)

Employing what was at the time largest CFD array in Europe, the 'Kabati' supercomputer was composed of four clusters with over 600 servers that could provide 72 teraflops of power. Although F1 regulations prevent the team from using more than 40 teraflops of computer power at any one time, the Kabati was still far and away the most powerful in use by any F1 team, and the tenth largest supercomputer in the UK at that time. Using this robust HPC capability, Marussia could crunch numbers sufficient to calculate the flow of air over their car’s entire chassis, saving the team the considerable cost involved in running numerous wind tunnel tests. Although, Marussia’s experiment in F1 was unfortunately short-lived, succumbing to financial pressures just a few years later, the team’s confidence in HPC broke some important precedents and opened new doors for future teams and helped shape the technical deployments of many of today's F1 starting grid.

Despite all the enthusiasm for CFD applications within F1 racing, there do remain skeptics who point out that it might not always provide the best solution. While wind tunnels are not perfect simulations of the real world, taking models out of the digital world where geometry is perfect, and putting them in a physical one, can be a very important means to verify and test the conclusions that CFD provides.

Still, the march of greater CFD application seems inexorable, as it now fully permeates every aspect of F1 car design, including tyre and engine development. This will be increasingly true as F1 cars — which are built from the ground up each year — embrace additive manufacturing and other technologies that dovetail with the insights provided by CFD, and bring new efficiency to the very tight development cycles common throughout the sport.

As Verne Global's resident F1 enthusiast I'm already looking forward to the Australian GP on March 26th to see what new cars will be competing and to see the latest technologies and designs that are deployed, many of which will have CFDs influence behind them.


Sign up for the Verne Global newsletter

Opinion, thought leadership and news delivered directly to your inbox once a month.