Riassunto analitico
Understanding the complex behavior of high Reynolds flows in Wind Tunnels can be very challenging. Over the years, state-of-the-art facilities have developed specific techniques that allow a better visualization and a deeper understanding of the flux evolution, as well as the turbulent structures present in the field. One of the most recognized approaches is called Particle Image Velocimetry (PIV). Thanks to fine reflective particles released in the flow and a tailored laser system, it permits the acquisition of successive instantaneous velocity fields (called snapshots) at a given frequency, usually in the order of 10^(1-:-2) Hz.
Given the high Reynolds tests conducted in the Wind Tunnel to analyse, for example, the aerodynamic performance of a racecar, small turbulent structures emerge in the fluxes, that can oscillate up to the Kolmogorov frequency (generally much greater than the acquisition frequency). This phenomena results in an aliasing error that affects the reliability and precision of the collected data.
In this context, the objective of the present work is to address the aliasing error present in PIV experimental data using the combination of two different techniques: Proper Orthogonal Decomposition (POD) and Compressed Sensing (CS). POD is a mathematical tool that enables a dynamic decomposition of the flow filed in spatial modes (velocity fields comparable to snapshots, but with more general structures) that are then scaled using a time coefficient (a scalar parameter evolving in time, one for each spatial mode) and ordered based on their respective energy. Compressed sensing instead is a technique used to reconstruct non-equidistant sub-Nyquist acquisitions, and in this work is used to retrieve the full evolution of time coefficients, in between sampled values.
|
