Background-oriented schlieren improved by machine learning for studying convective flows

Optical methods are vital for studying fluid flow and heat transfer, because of their precision and non-intrusive nature. Recent progress simplifies measurements and boosts data processing. We focus on improving fringe pattern analysis, crucial in techniques like background-oriented schlieren (BOS), especially for visualizing convective heat flows. Our solution, based on conditional generative adversarial networks (cGANS), effectively enhances fringe quality by training on simulated data with added noise. This approach proves valuable for refining phase estimation accuracy in experimental setups, offering promising insights into heat transfer dynamics.