Latest news with #KennethJansen

Yahoo

US supercomputer simulates 1 quintillion calculations to design better airplanes

U.S. Department of Energy's(DOE) Argonne National Laboratory is helping researchers explore novel ways to design aeroplanes using its Aurora supercomputer. Known as one of the world's first exascale supercomputers, Aurora can perform over a quintillion calculations per second. A team from the University of Colorado Boulder uses Aurora's exascale supercomputing capabilities with machine learning techniques to dissect the airflow around commercial aircraft. Through this activity, they aim to provide insights to inform the design of next-generation airplanes. 'With Aurora, we're able to run simulations that are larger than ever before and on more complex flows," said Kenneth Jansen, professor of aerospace engineering at the University of Colorado. ​ "These simulations help improve the predictive models that are applied to even more complex cases, such as capturing the flow physics around a full vertical tail and rudder assembly of an aircraft at full flight scale,' he continued. Reason to study the airflow Airplanes often have vertical tails larger than needed for most flights, to tackle worst-case scenarios like taking off in a crosswind with only one engine running. The team of researchers thought that if they could study the physics of the flow better, they could design a smaller tail that is still effective in such scenarios. They are using a tool called HONEE to run detailed airflow simulations. The tool helps model the complex, chaotic behaviour of turbulent air. These high-quality simulations are then used to train AI models called subgrid stress (SGS) models. SGS models are important because they help predict the effects of tiny turbulent air movements that aren't directly visible in lower-resolution simulations, but are crucial for accurate airflow predictions. All about turbulence Traditional turbulence is dependent on two things: extensive stored datasets and slow offline analysis. The team's new method uses machine learning simultaneously with simulation to save time and bypass the need to store massive chunks of data. 'Online machine learning refers to carrying out simulations that produce training data at the same time that the actual training task is carried out,' Rowe said. 'We're ensuring we can examine the simulation fields in real time and extract dynamics as the simulation progresses,' Kenneth Jansen explained. ​ 'On the machine learning side, we're using those same analytics to understand how they impact turbulence models. Machine learning allows us to uncover modelling behaviors that complement and extend our current understanding," he continued. Using this method, the team has built predictive models that can detect the behavior of turbulent air. This approach allows scientists to test new ideas for real-time flow control and evaluate how smaller tail designs can fare in extreme conditions. The tool stack The team uses Aurora's exascale power and advanced tools like HONEE for fluid simulations, SmartSim for real-time data streaming and in-node training, and PETSc for scalable numerical calculations. This integrated approach enables faster, more efficient aircraft design by combining simulation and machine learning without costly physical testing. Solve the daily Crossword