Reduced Order Methods for Automotive and Nautical Applications


Wednesday, 31 August, 2016 - 11:00

Dott. Filippo Salmoiraghi, SISSA mathLab, Wednesday August 31, 2016 at 11 am, Room A-133, SISSA main campus

Title: “Reduced Order Methods for Automotive and Nautical Applications


Abstract: We present a new pipeline for the Model Order Reduction, from the geometrical parametriza-

tion to the construction of the reduced basis, for complex engineering problems. The pro-

posed approach relies on Free Form Deformation (FFD) for the geometrical parametrization,

a "leave one out" strategy for the selection of the parameters value and Proper Orthogonal

Decomposition (POD) with interpolation of the POD coefficients for the online evaluation.

Since most of the time engineers are interested in the outputs of a problem instead of the

whole solution, both the parameters selection and reduced basis construction are performed

directly on the output of interest. After a general presentation of the pipeline, we show the

application to the solution of steady RANS equations around drivAer model with paramet-

ric bumper shape. The outputs of interest are the pressure and wall stress fields on the car,

which are fundamental for the computation of drag and lift coefficients. The results show

a speed-up of order of millions with an error of about 4%. Alongside we briefly present two

new Python libraries developed at SISSA mathLab for model order reduction, namely

PyGeM and EZyRB.