PIV-in-the-Loop Optimization of Active Turbulence Grids for Tailored Inflow Generation

Active turbulence grids can generate tailored turbulent inflow conditions, but identifying suitable actuation patterns is difficult when many grid axes must be tuned simultaneously. We present a PIV-in-the-loop framework that automates this process by combining spatially resolved velocity-field measurements, a reduced-order parameterization of grid motion, and Gaussian-process surrogate models. The method first explores the accessible actuation space through uncertainty-guided active learning, then uses the learned models to identify grid motions for prescribed turbulent inflow objectives. We demonstrate the approach for homogeneous turbulence with prescribed turbulence intensity, a mean-velocity gradient at nearly constant turbulence intensity, and a turbulence-intensity gradient at nearly uniform mean velocity. The results show that closed-loop PIV feedback can reproducibly tune active-grid motions and provide a scalable route toward on-demand turbulent inflow generation for wind-tunnel experiments.