Marussia F1 is using CFD technology to help optimise the performance of its race car. Stephen Harris reports
Formula One (F1) isn’t a sport for the little guy. You need access to world-class engineering expertise and millions of pounds’ worth of financial backing to compete against the likes of Ferrari and McLaren. But in relative terms, the Marussia F1 team — formerly known as Virgin Racing — operates on a much smaller budget than the big players. So when Pat Symonds, former executive engineering director of Renault F1 and now technical consultant for Marussia, was told the team wanted to achieve a podium place at the inaugural Russian Grand Prix in 2014, he knew he had his work cut out for him. ‘At the time we were setting this mission statement, McLaren, Red Bull and Ferrari were the only teams that had a realistic chance of a podium finish,’ he said. ‘The only way we could do it was to really be novel. We weren’t going to do it in the period required just using traditional methods.’
The solution lay not in engine mechanics or a radical new car design but in the field of IT. Today, it’s virtually impossible to start an F1 car if its computer systems are not working. IT is vital not just for designing and testing the cars but also for controlling their components, monitoring their on-track performance and changing their settings in the pits.
For Symonds, computing power offered the potential advantage that Marussia needed if it was going to compete at the highest level on a limited budget. In particular, it offered an alternative to highly expensive wind-tunnel trials. ‘Generally speaking, nothing contributes more [to a car’s performance] than aerodynamics,’ he said. ‘The research money that we put into making our car go faster is nearly all spent on it. And so we needed to think smarter.’ The smarter answer was to invest in the company’s computational fluid dynamics (CFD) capability. In the case of F1, CFD means using digital software to model how air passes over the car to increase down-force and reduce drag, and, importantly, understand how this correlates with the car’s real performance on the track. ‘CFD is in effect a digital wind tunnel – a digital space that you design your car in and try to understand the aerodynamics,’ said Symonds.
“The research money we put into making our car go faster is nearly all spent on aerodynamics”
Pat Symonds, technical consultant, Marussia F1
F1 regulations limit the amount of aerodynamic testing a team can do. Reducing the hours spent in a wind tunnel means they can run more CFD simulations – and at a sixth of the cost. In fact, for Marussia’s first two seasons, the team did no wind-tunnel testing at all, providing the perfect base to build possibly the biggest CFD capability in the sport.
As you enter Marussia’s testing centre at its base in Banbury, Oxfordshire, you’re greeted by a wall of server-like boxes, their many blinking lights hinting at the incredible computing power they contain. With a performance rating of 72 teraflops – or 72 trillion calculations per second – Marussia’s supercomputer is one of the most powerful in the UK. F1 rules only allow an average performance of 40 teraflops but having this high peak capability leaves room for intense use when needed and space to run experiments on last year’s design without breaking the rules.
Each model is built using a 3D car design broken down into a mesh of tiny triangular cells that extends across the whole surface of the car – and into the space around it. The supercomputer then solves a series of equations that model how air moves through each cell and interacts with the car, a process that can take 24 hours and outputs a huge amount of data. ‘We can know the velocity, we can know the pressure, and all the forces and all the physical characteristics of any part of the car,’ said Daniel Jean, Marussia’s head of CFD. ‘We can know what the flow structures look like and the exact position of vortices. In a wind tunnel it is difficult to get this.’ Marussia has even used CFD to model visible changes to other teams’ cars to determine what advantage they might have.
Automating around 95 per cent of the model building has helped Marussia and its IT partner CSC optimise the process. Once the simulations are ready to run, the difficulty is using the computer in the most efficient way to manage the huge amount of data produced, said Ian McKay, high-performance computing service manager at CSC. ‘By putting the mesh over the design you break it down into many chunks and send each one to a different computer node, and you’ve got to co-ordinate that and what happens next when each calculation is finished.’ Of course, computer simulations are not perfect, hence why it is so important for the engineers to learn how the models correlate with real-world performance, and why Marussia does now use a wind tunnel. It also helps that it doesn’t take hours to set up a new wind-tunnel experiment if testers want to move something on the car. But by having a strong CFD capability as its base, the team is able to design the most efficient trial process.
Marussia’s attempts to improve the team’s performance using IT don’t stop at the design and testing stages. Once the car is out on the track, every aspect of its performance is monitored in real time via a network of sensors connected wirelessly to the garage. An engine-control unit manages the car’s systems, allowing engineers to alter the clutch torque or rev and kill the engine via a laptop. Even the pit itself utilises computer technology to optimise the position of the team.
When F1 engineers prepare to fly out to each of the season’s 20 race locations around the world, they don’t just have to worry about the car itself but a substantial amount of computer hardware as well, which has to survive wildly varying temperatures, desert sands and city pollution. Then there’s the issue of maintaining radio communication with the car, especially on urban tracks where the car can quickly disappear behind tall buildings. ‘The cost of the engine is such that we need to know that everything is in its normal operating range,’ said Ian Jackson, head of trackside IT for CSC. ‘They won’t start the car unless they have all the data they need, so we’ve got to make sure the network is reliable.’
CSC has also reduced some of the chaos of the pit by developing cloud-based time-management software that displays the team’s schedule and approaching deadlines – which are so strictly enforced that the team can even be penalised if engineers arrive at the garage too early or leave too late. CSC’s software is even helping the pit crew to co-ordinate its exact movements when the car comes in, via a program that records digital video of the pit stop and displays an analysis of each member’s timing, enabling them to fine-tune their performance.
‘It’s small things such as how they hold the tyres and how they stand,’ said Jackson. ‘If we can make a tenth of a second’s difference to the car’s stationary time, the cost of engineering that into the car would far outweigh the cost of doing something more efficiently in the pit stop.’
Looking to the future, Symonds expects IT only to become more important to all F1 teams, particularly in the field of aerodynamics. ‘There’s a parallel in vehicle dynamics,’ he said, citing the high costs, variability and strict regulation of testing vehicle handling on the track. ‘There we’ve moved almost totally over to simulation. And I think that’s the trend we’ll see in aerodynamics. With CFD we’re not quite there yet. There are 300 million equations to solve. But we will get there.’
Genetic algorithms could help specialists deal with the masses of data generated by sensors
One of the key challenges for the future of IT in Formula One (F1) is similar to that now found in many sectors: how to best make use of the ‘big data’ captured through improvements in sensor and communication technology. An F1 car is equipped with hundreds of sensors monitoring everything from water temperature to lateral acceleration, creating real-time data feeds for the race team and a wealth of information to be recorded and studied in detail once the car returns.’The problem now is that we have gigabytes of data,’ explained Marussia’s technical consultant, Pat Symonds. ‘We tend to have specialists looking at each area of the data. I think there is a huge future for genetic algorithms – algorithms that teach themselves.
‘A bit of a holy grail is to have all this data, decide when a car is handling well and work out what has led to it so that the computer starts to teach itself what it is looking for.’