Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Improvement of RANS turbulence modelling capabilities through Machine Learning
The student will study the efficacy of applying Machine Learning techniques for providing improved initial conditions to the computational simulations and improving the accuracy of low-cost turbulence models.
The study of Formula 1 car aerodynamic performances involves the use of computational methods to approximate the Navier-Stokes equations at high Reynolds numbers, among other tools. These methods can use a variety of different modelling techniques to deal with the turbulent nature of the flow, where an increased accuracy typically corresponds to a higher computational cost.
Requirements:
students from Mathematical Engineering, Mathematics, Data Science and Scientific Computing courses. A solid knowledge of Machine Learning foundations, as well a passion for programming (Python and C++, preferably), are a must. A basic knowledge of fluid dynamics and numerical analysis is a strong asset.
The study of Formula 1 car aerodynamic performances involves the use of computational methods to approximate the Navier-Stokes equations at high Reynolds numbers, among other tools. These methods can use a variety of different modelling techniques to deal with the turbulent nature of the flow, where an increased accuracy typically corresponds to a higher computational cost. Requirements: students from Mathematical Engineering, Mathematics, Data Science and Scientific Computing courses. A solid knowledge of Machine Learning foundations, as well a passion for programming (Python and C++, preferably), are a must. A basic knowledge of fluid dynamics and numerical analysis is a strong asset.
The student will study the efficacy of applying Machine Learning techniques for providing improved initial conditions to the computational simulations and improving the accuracy of low-cost turbulence models. The approach will be based on a hybrid dataset of results obtained from both low fidelity and high-fidelity models. The final purpose will be the design, the efficient implementation and the testing of a ML improved RANS turbulence model and simulation workflow.
The student will study the efficacy of applying Machine Learning techniques for providing improved initial conditions to the computational simulations and improving the accuracy of low-cost turbulence models. The approach will be based on a hybrid dataset of results obtained from both low fidelity and high-fidelity models. The final purpose will be the design, the efficient implementation and the testing of a ML improved RANS turbulence model and simulation workflow.
- Davide Fransos, Davide.Fransos@sauber-group.com-
- Francesco Del Citto, Francesco.DelCitto@sauber-group.com
- Prof. Dr. Filippo Coletti, fcoletti@ethz.ch, Institute of Fluid
Dynamics, ETH Zurich
- Davide Fransos, Davide.Fransos@sauber-group.com- - Francesco Del Citto, Francesco.DelCitto@sauber-group.com - Prof. Dr. Filippo Coletti, fcoletti@ethz.ch, Institute of Fluid Dynamics, ETH Zurich