Towards an entirely virtual design process : The challenge for turbulence modeling
Séminaire Dr Neil Ashton - University of Oxford e-Research Centre
Mercredi 17 février à 14 h 30
salle Castex rez de chaussée
Turbulence is one of the greatest challenges for fluid mechanics and physics. Indeed, turbulent flows are all around us, from the flow of air over an aircraft or car, to the ocean’s currents. Over the past 30 years the use of Computational Fluid Dynamics (CFD) in the aerospace and automotive sectors have grown to become a major design tool, used from conceptual analysis all the way through to the detailed design phase. With computational resources becoming ever cheaper and more powerful, for many problems grid convergence is now becoming possible with Reynolds-Averaged Navier-Stokes (RANS) approaches. Yet there is a general consensus that RANS models (even at second moment closure level) struggle to capture certain classes of flows where large scale unsteadiness and 3D separation exist. In such flows there is a tendency for these models to under-predict the turbulent shear-stress in the initial separated shear layer which leads to an over-prediction of the recirculation region.
Whilst computational resources have grown exponentially, they are not yet powerful and readily available enough to use methods such Large Eddy Simulation (LES) on complex bodies. Hybrid RANS-LES methods such as Detached-Eddy Simulation (DES) have helped to bridge the gap between RANS and LES, however generating suitable meshes and addressing the transition between RANS and LES regions has limited their use within industrial design processes.
This talk will focus on the suitability of these turbulence models for some of the grand challenges of modern day CFD, illustrating the importance of suitable meshes, numerical schemes and boundary conditions. The performance of a range of turbulence models are discussed on a number of fundamental and real-life applications. Finally, the case for an automated embedded LES approach is made and in particular how High-Performance Computing can make these methods suitable for industrial design processes.
Bio : Dr Neil Ashton is a visiting research fellow at the University of Oxford e-Research Centre. He completed his Masters and PhD at the University of Manchester where his thesis focused on the development of new hybrid RANS-LES turbulence models. During his PhD he was also involved with various EU-funded turbulence modelling projects which looked at developing hybrid RANS-LES approaches suitable for the aerospace industry. He is now involved with the follow up EU FP7 Go4hybrid project and has also participated in a number of international AIAA workshops. After his PhD he worked at the Lotus Formula 1 team, where he was responsible for developing the CFD methodology and optimising the use of High Performance Computing (HPC). His current research is focused on developing new RANS and hybrid RANS-LES models suitable for complex industrial flows. He is a member of the AIAA Turbulence Modelling working group and has active collaborations with a number of institutions including Audi, Barcelona Supercomputing Centre and NASA Ames Research Centre.