Development of a rapid optical flow-based particle image velocimetry framework


Creative Commons License

Karaca E., PERÇİN M.

Ankara International Aerospace Conference, AIAC 2025, Ankara, Türkiye, 17 - 19 Eylül 2026, cilt.3271, (Tam Metin Bildiri)

Özet

This study aims to develop a particle image velocimetry (PIV) framework, intended for in-house use as a live telemetry tool during wind tunnel experiments, by implementing recently developed optical flow techniques to enable rapid flow field assessments of low-resolution PIV images without post-processing. The objective is to implement a data-driven optical flow algorithm that is capable of estimating velocity fields from image sequences in near real-time. The implementation consists of two parts: pre-processing and computation. Pre-processing includes data acquisition and image filtering, while the computation part involves applying a convolutional neural network-based PIV algorithm optimized through parallelization. Validation was performed on various flow cases, which includes Rankine vortex, channel flow, backstep flow, and turbulent mixing. Results show endpoint errors between 2% and 12%, with a processing speed of around 10Hz, which demonstrates its potential for live telemetry of flow structures during experiments. Future work includes object masking for moving boundaries, higher-resolution inputs, and integration with the experimental systems of METU Center for Wind Energy Research (RUZGEM) for on-site deployment.