Abstract
The ability to understand unsteady fluid flows is foundational to advancing technologies in energy, health, transportation, and defense. This work uses data-driven methods (i.e., machine learning) to interpret and control unsteady fluid flows through experiments. Specifically, these methods are used to control, optimize, and model cross-flow turbines. Cross-flow turbines (i.e. vertical axis turbines), are devices that can be used to convert the kinetic energy in wind to electricity. A key advantage of cross-flow turbines over axial-flow turbines is that they can efficiently operate in close-proximity in arrays. We demonstrate how data driven methods can be used to efficiently explore, model, and interpret the high-dimensional space cross-flow turbine dynamics occupy through the following three projects. First, robust principal component analysis (RPCA), a method borrowed from robust statistics, is used to improve flow-field data by leveraging global coherent structures to identify and replace spurious data points. We apply RPCA filtering to a range of fluid simulations and experiments of varying complexities and assess the accuracy of low-rank structure recovery. First, we analyze direct numerical simulations of flow past a circular cylinder at Reynolds number 100 with artificial outliers, alongside similar particle image velocimetry (PIV) measurements at Reynolds number 413. Next, we apply RPCA filtering to a turbulent channel flow simulation from the Johns Hopkins Turbulence database, demonstrating that dominant coherent structures are preserved in the low-rank matrix. Finally, we investigate PIV measurements behind a two-bladed cross-flow turbine that exhibits both broadband and coherent phenomena. We demonstrate that more persistent dynamics can be identified when RPCA is utilized in lieu of traditional processing methods. In all cases, both simulated and experimental, we find that RPCA filtering extracts dominant coherent structures and identifies and fills in incorrect or missing measurements. Second, the performance of a two-turbine array in a recirculating water channel was experimentally optimized across 64 unique array configurations using a hardware-in-the-loop approach. For each configuration, turbine performance was optimized using tip-speed ratio control, where the rotation rate for each turbine is optimized individually, and using coordinated control, where the turbines are optimized to operate at synchronous rotation rates but with a phase difference. For each configuration and control strategy, the consequences of co- and counter-rotation were also evaluated. Arrays with well-considered geometries and control strategies are found to outperform isolated turbines by up to 30%.Third, the performance and wake of a two-turbine array in a fence configuration (side-by-side) are characterized. The turbines are operated under coordinated control. Measurements were made with turbines co-rotating, counter-rotating with the blades advancing upstream at the array midline, and counter-rotating with the blades retreating downstream at the array midline. From the performance and wake data, we found individual turbine and array efficiency to depend significantly on rotation direction and phase difference. Persistent dynamics that exist across all flow fields, as well as differences between cases are identified. Each of these projects demonstrate how data-driven methods can be used to explore, model, and interpret cross-flow turbine dynamics and other fluid systems.