Turbulence enhancement of a fan array wind generator using geometric texturing and optimization-based control

Fan array wind generators (FAWG) are designed to generate a rich set of turbulent flows reminiscent of those found in natural environments. In this study, we experimentally investigate a square FAWG consisting of 10x10 individually controllable fans with 4 cm width and a maximum velocity of 17 m/s. The goal is to maximize the turbulence intensity in the test region. Two approaches for fan operation are investigated: first, geometric texturing of the duty cycle distribution, and second, maximization of the turbulence intensity at selected hot-wire sensors with particle-swarm optimization. We find that geometric texturing (specifically a checkerboard pattern) yields a robust, uniform turbulence field (Tu ~ 0.14) driven by jet interactions. Conversely, particle swarm optimization achieves higher local turbulence (Tu ~ 0.28) but significantly sacrifices spatial uniformity. This study underscores the trade-off between local maximization and global uniformity in active turbulence generation.

💡 Research Summary

This paper investigates how to maximize turbulence intensity in a fan‑array wind generator (FAWG) and compares two fundamentally different control strategies. The experimental platform consists of a square 10 × 10 matrix of individually addressable axial‑flow fans (each 4 cm wide, maximum jet velocity 17 m/s). The authors aim to create a turbulent environment that mimics the highly unsteady atmospheric conditions encountered by unmanned aerial vehicles (UAVs) and electric vertical take‑off and landing (eVTOL) aircraft, which are increasingly deployed in complex urban airspaces.

The introduction reviews the limitations of traditional passive turbulence generators (grids, roughness elements, spires) and positions FAWGs as a flexible, active alternative. Prior work on FAWGs has focused mainly on reproducing mean wind profiles or standard atmospheric‑boundary‑layer statistics; few studies have attempted to deliberately maximize turbulence intensity.

Two control concepts are explored. The first, termed Geometric Texturing Pattern (GTP), uses fixed on/off duty‑cycle patterns that create sharp shear layers between active and inactive fans. A checkerboard (chessboard) arrangement is highlighted because it yields a regular array of interacting jets. The second approach employs Particle Swarm Optimization with Target‑Position‑Mutation‑Elitism (PSO‑TPME). Here, 25 independent control variables (the duty cycles of the symmetric fan subset) are optimized in real time using feedback from four hot‑wire anemometers placed at strategic locations in the test section. The cost function J combines three terms: (a) the reciprocal of the average turbulence intensity (to encourage high Tu), (b) a penalty for deviation of the local mean velocity from a prescribed target, and (c) a penalty for non‑uniformity among the four measured turbulence intensities. Weighting factors λ₁ and λ₂ are tuned so that a 1 % increase in Tu corresponds roughly to a 1 % increase in the other penalties.

Measurement techniques include constant‑temperature hot‑wire anemometry (sampling at 10 kHz for 30 s across 20 streamwise stations, x/w = 1…20) and stereoscopic particle‑image velocimetry (PIV) at a downstream plane x/H = 1 (12 Hz, 2000 image pairs). The hot‑wire probes are positioned at the centers of the four quadrants (A1‑D1) and a cluster in the lower‑left quadrant (A2‑D2) to capture spatial variations.

Results for the geometric patterns show that the small checkerboard (Pattern 1) produces a highly uniform mean‑velocity field (≈8.6 m/s) and a fairly homogeneous turbulence intensity of Tu ≈ 0.125 ± 0.035 after the jets have merged (x/w ≈ 10). The turbulence peaks early (x/w ≈ 2.5) at Tu ≈ 0.355 ± 0.065, reflecting the strong shear at the jet edges, then decays and homogenizes downstream. The larger checkerboard (Pattern 2) yields a slightly higher overall Tu (≈0.14–0.15) while preserving symmetry, making it the best candidate for a “high‑intensity yet still uniform” field. Other patterns (vertical/horizontal stripes, cross‑stripes) either increase local mean velocity or introduce pronounced anisotropy, confirming that the spatial arrangement of active fans directly controls both the magnitude and distribution of turbulence.

The PSO‑TPME optimization converges to a duty‑cycle distribution that raises the locally measured turbulence intensity to Tu ≈ 0.28, nearly double that of the best geometric pattern. However, the four hot‑wire probes report a large spread in Tu (differences > 0.1), indicating a highly non‑uniform turbulence field. The mean velocity remains close to the target (≈8 m/s), showing that the optimizer successfully balances the competing objectives but ultimately sacrifices spatial uniformity to achieve the peak Tu.

The authors discuss the trade‑off between global uniformity and local maximization. For applications that require a repeatable, isotropic turbulent environment (e.g., baseline UAV stability testing), the checkerboard GTP is preferable due to its simplicity, robustness, and ease of implementation. For scenarios where a specific region of extreme turbulence is needed (e.g., testing control algorithms under worst‑case gusts), the PSO‑based approach offers a powerful tool, provided sufficient sensor coverage and computational resources are available.

In the concluding section, the paper emphasizes that FAWGs provide active, scalable, and programmable turbulence generation, overcoming the rigidity of passive devices. The 10 × 10 prototype demonstrates that jet‑interaction physics observed in small‑scale arrays scale predictably to larger configurations (e.g., 40 × 40 arrays). Future work is suggested in three directions: (1) extending the optimization to multi‑objective formulations that include spectral shape or frequency‑domain targets, (2) scaling up the hardware while preserving the symmetric control architecture, and (3) integrating the FAWG with hardware‑in‑the‑loop flight‑control experiments for UAVs and eVTOLs. Overall, the study provides a clear framework for designing turbulence‑generation systems tailored either for uniform, high‑intensity fields or for localized extreme gusts, thereby advancing experimental capabilities for next‑generation aerial vehicle testing.