Advanced Flow Measurement and Control Based on Data-Driven Mode Decomposition
Tako Nonomura
Abstract: The author’s group has proposed the use of data-driven mode decomposition techniques, such as proper orthogonal decomposition and dynamic mode decomposition, for image-based fluid experiments including PIV, BOS, and PSP. Detailed investigations have demonstrated that these techniques can accurately capture and represent complex unsteady flows. Building on this foundation, a spatio-temporal superresolution framework was developed. This framework combines complementary measurements of different resolutions to reconstruct flow fields at extremely high temporal rates, enabling the resolution of supersonic flow phenomena at up to 200 kHz—well beyond conventional experimental limits. With this approach, unsteady flow structures that were previously inaccessible have been successfully visualized. In parallel, the group developed sparse-sensor optimization methods to reconstruct entire flow or pressure fields from limited measurements. Building on this concept, they proposed sparse-processing PIV, which achieves real-time velocity-field measurements at 2000 Hz, currently the fastest in the world. Integrating these advances, the group conducted wind tunnel experiments in which sparse real-time PIV served as the observer, a plasma actuator as the actuator, and modal coefficients obtained from decomposition as state variables for feedback. This setup enabled the world’s first successful visual feedback control experiment that suppressed flow asymmetry caused by Kármán vortex shedding, thereby demonstrating the practical potential of advanced, data-driven fluid control. In the presentation, those results are presented.
|
