Our research highlights serve as a collection of feature articles detailing recent scientific achievements on GCS HPC resources.
With the advent of air taxis and drone-based delivery services, demand is growing for a new generation of helicopters that are as safe, quiet, and energy-efficient as possible. A research group led by Prof. Manuel Keßler at the University of Stuttgart’s Institute of Aerodynamics and Gas Dynamics (IAG) has been busy answering this need. Keßler’s team has long focused on improving helicopters by using supercomputers at the High-Performance Computing Center Stuttgart (HLRS) to understand the complex aerodynamics surrounding helicopter rotors.
“My group has been focused on this topic for two and a half decades, and throughout that period, we have had a close cooperation with industry,” Keßler said. “They come to us with questions they need support with, and we provide them with knowledge and insights, as well as simulation tools for answering these questions. We’ve had a long-running collaboration with Airbus Helicopters, and they are using many of the same tools in their research as we are. So, the outcome of our research finds its way directly into industrial products.”
Recently, Keßler and his collaborators worked closely with Airbus on modelling and simulation efforts to reduce noise and better understand flight dynamics in the company’s new RACER prototype. RACER, which took its first test flight in April 2024, is unusual in that it has both helicopter rotor blades and airplane propellers on wings. These features will enable the new aircraft to respond more rapidly in emergency situations. The collaboration to improve RACER with modelling and simulation is an example of the ways in which public HPC centers contribute to European economic development and competitiveness.
Modelling and simulation expedite time to market, complement experiment
While helicopters are not a recent invention, making them more efficient, safer, or quieter requires understanding a complex mix of physical phenomena at a fundamental level. To develop models of these phenomena, the researchers must break down the helicopter’s complex geometry into a very fine computational grid of roughly 200 million grid cells. In a standard simulation, the researchers include 10 rotations of the helicopter rotor. To capture the small-scale changes that can affect flight dynamics or acoustics, they typically divide each rotation into 720 small time steps. This means completing, tracking, and understanding relationships among more than a quadrillion individual calculations. “These simulations are too large for us to attempt on a workstation or smaller cluster; we need to have access to supercomputers to perform these calculations in an efficient manner,” Keßler said.
In their collaboration with Airbus and as part of the project CA³TCH within the framework of the European Union’s Clean Sky 2 programme, Keßler’s team modelled the new RACER design both to reduce the noise that it generates and to gain a more complete understanding of flight dynamics, lowering risk before the craft’s first flight. The team ran a suite of simulations under a variety of circumstances, such as hovering during a crosswind and determining the so-called “roll angle,” or the maximum sideways angle at which the craft can tilt before the propellers touch the ground. Because RACER has a traditional rotor as well as wings, there is very little prior experimental data to rely on, making the need for modelling and simulation even greater. “Based on our well-established workflow and the rapport that we have built with our industry partners, they take our simulations seriously if we indicate that a model could have a stability problem,” Keßler said. “I’ve always said that our duty is to ensure that after an aircraft’s first test flight, the pilots exit with a grin on their faces and were not sweating while they were in flight.”
Accelerated computing comes with challenges and opportunities for designing next-generation aircraft
As HLRS pivots from Hawk, its current flagship supercomputer, to Hunter, a next-generation system that will become available in 2025, research teams like Keßler’s see both opportunities and challenges. Hunter will be the first system at HLRS that includes a large number of GPU accelerators. While GPUs have become a common solution for accelerating performance and improving energy efficiency in the world’s fastest computers, they are also a new technology for some scientific users of HPC. For researchers who have not yet used GPU-centric supercomputers, this transition will mean modifying — sometimes heavily — their code packages to take full advantage of the system architecture. With a strong command of Hunter’s system, the team should be able to run more high-resolution simulations, further shortening production timelines and improving safety in the process.
Keßler pointed out that HLRS’s user support staff has proactively helped make sure his team is ready to use the center’s next-generation system as soon as it comes online. “The center has already been very supportive during this transition,” Keßler said. “We were recently invited to a meeting where we discussed the different options with staff from different sides of this installation, including from the hardware manufacturers. HLRS came to us actively and said, ‘We will be making this transition in the future, so how can we start helping you prepare for it now?’”
Keßler indicated that while HLRS’s staff has neither detailed knowledge of his team’s code nor the capacity to rewrite codes for the center’s dozens of research projects, he considers this early collaboration essential. It will help to ensure that his team can continue to deliver simulations to industry partners, providing insight for next-generation helicopter designs and ultimately ensuring that safe, reliable, and efficient European helicopters continue to come off the production line.
-Eric Gedenk
Funding for Hawk was provided by Baden-Württemberg Ministry for Science, Research, and the Arts and the German Federal Ministry of Education and Research through the Gauss Centre for Supercomputing (GCS).