Our latest research explores the use of neuromorphic Event-Based Vision (EBV) cameras to represent low-order model coordinates in turbulent flows, in the quest for more data-efficient and high-fidelity flow measurements.
Unlike conventional imaging systems, EBV cameras operate asynchronously, capturing only changes in temporal contrast at each pixel. This allows for high-frequency output with minimal data bandwidth and enhanced sensitivity (particularly in low-light conditions). Finally, it is a cost-effective alternative to high-speed cameras, providing a continuous stream of data. Our study positions EBV sensors as a promising technology for achieving data-efficient, high-resolution flow measurements—paving the way for advanced turbulence modeling and flow control applications.
For experimental validation, we assessed Pulsed Event-Based Imaging Velocimetry (EBIV) against traditional Particle Image Velocimetry (PIV) in two synchronized experiments, a submerged water jet and an airflow around a square rib in a channel. Through Proper Orthogonal Decomposition (POD) and Low-Order Reconstruction (LOR), we demonstrate that EBIV accurately identifies dominant flow structures, provides reliable reduced-order models for flow control applications, and maintains robust performance despite noise in higher-order modes.
Read the open access publication, co-authored by Luca Franceschelli, Chris Willert (German Aerospace Center), Marco Raiola, and Stefano Discetti here. You can also access the data sets in Zenodo. The authors acknowledge the invaluable help of Michael Schroll of the DLR Institute of Technology in setting up the air flow experiment.
