SFMC23

Machine learning based Viscous-Inviscid coupling for high order solvers

  • Otmani, Kheir-eddine (Universidad Politecnica de Madrid)
  • Ntoukas, Gerasimos (Universidad Politecnica de Madrid)
  • Ferrer, Esteban (Universidad Politecnica de Madrid)

Please login to view abstract download link

An unsupevised learning clustering technique has been used to decompose the computational domain into a viscous, rotational region and an inviscid, irrotational region, different sets of equations have been solved in each region namely: the Navier-Stokes in the viscous dominated region and the Euler equation in the outer inviscid region. The methodology has been validated with a LES test case for a flow past a cylinder at $Re=3900$ showing that the methodology could reduce the computational time up to $12\%$ while maintaining the same level of accuracy.