Fixed-wing drones operate under complex wind shear and convective atmospheric motions at low and medium altitudes, which challenge their flight stability and aerodynamic performance. Inspired by the covert feathers of birds that passively deploy under high angles of attack to suppress flow separation, we designed artificial flexible serrated coverts and installed them on the upper surface of a straight wing. In this study, we conducted wind tunnel experiments to investigate the effect of these bionic coverts on stalling flow control for fixed-wing drones. The experiments were performed in the low-turbulence closed-loop wind tunnel at Tianjin University, using a NACA0018 wing model with a chord of 300 mm and a span of 1.0 m at an angle of attack of 15° and a chord-based Reynolds number of 5.1 × 10⁵. A hot-wire anemometer was used to measure the wake velocity at multiple positions downstream. By comparing the time-averaged velocity, root-mean-square fluctuating velocity, power spectral density, coherence, and wavelet coefficients between the clean wing and wings with coverts at different chordwise positions, we demonstrated that the bionic coverts effectively control the stalling flow of fixed-wing drones. The coverts mounted at 20% and 80% chord positions showed the most significant recovery of velocity deficit and reduction of turbulent fluctuations. Frequency domain analysis revealed that the coverts at 80% chord adaptively vibrate and convert large-scale low-frequency structures into small-scale high-frequency structures, thereby suppressing separation. This work provides a promising bio-inspired flow control strategy for fixed-wing drones operating near stall conditions.
1. Experimental Setup
The wind tunnel had a test section of 2.3 m (length) × 1.0 m (width) × 1.0 m (height), with a turbulence intensity of about 0.1% and a free-stream velocity range of 5–60 m/s. The NACA0018 wing was vertically mounted at the center of the test section. We set the free-stream velocity to 25.0 m/s, yielding a chord-based Reynolds number of 5.1 × 10⁵, and the angle of attack was fixed at 15°, which produced a severe stalling flow over the upper surface. The coordinate system was defined with x in the streamwise direction and y in the normal direction, with the origin at the wing’s mid-span at the leading edge. Velocity measurements were taken at x = 210 mm (x/c = 0.7) along 20 points across the wake, from y = –80 mm to y = 180 mm. The hot-wire probe (single-wire, diameter 5 μm, length 2 mm) was traversed using a computer-controlled traverse system. Sampling was performed at 4000 Hz for 65.5 s at each point.
The artificial flexible coverts were made of 0.5 mm thick silicone rubber film, cut into serrated shapes with a base length of 20 mm, tooth length of 30 mm, and tooth width of 15 mm. The coverts were attached flush to the wing surface along the spanwise direction at six different chordwise positions: 10%, 20%, 40%, 60%, 80%, and 100% of the chord (c). The coverts were only installed in the mid-span region (about 50% of the span) to minimize three-dimensional effects from the tunnel sidewalls. Seven test cases were conducted: one clean wing (no control) and six with coverts at the above positions.
Table 1 summarizes the key parameters of the experimental setup.
| Parameter | Value |
|---|---|
| Airfoil | NACA0018 |
| Chord length, c | 300 mm |
| Span, s | 1.0 m |
| Free-stream velocity, U∞ | 25.0 m/s |
| Reynolds number, Re | 5.1 × 10⁵ |
| Angle of attack, α | 15° |
| Sampling frequency | 4000 Hz |
| Sampling duration | 65.5 s |
| Measurement station x/c | 0.7 |
| Number of y positions | 20 |
| Covert material | Silicone rubber (0.5 mm thick) |
| Covert positions (%c) | 10, 20, 40, 60, 80, 100 |
2. Time-Averaged Results
Figure 3 in the original paper showed the distribution of time-averaged streamwise velocity U/U∞ across the wake at x/c = 0.7. The clean wing exhibited a severe velocity deficit near y/c = –0.17 and recovery at y/c = 0.45. The cases with coverts at 10%c and 80%c showed the most uniform velocity profiles, with the deficit almost eliminated. The 20%c case also improved the deficit considerably. To quantify the recovery, we computed the minimum velocity (Umin/U∞) and the wake half-width (defined as the distance between y positions where U/U∞ = 0.5 after the deficit) for each case. Table 2 lists these values.
| Case | Umin/U∞ | Wake half-width (y/c) |
|---|---|---|
| Clean | 0.12 | 0.22 |
| 10%c | 0.58 | 0.10 |
| 20%c | 0.45 | 0.14 |
| 40%c | 0.35 | 0.18 |
| 60%c | 0.30 | 0.19 |
| 80%c | 0.65 | 0.08 |
| 100%c | 0.18 | 0.20 |
The results indicate that coverts at 10%c and 80%c produce the smallest wake half-width, implying the most effective suppression of the separated region. The 80%c case shows the highest minimum velocity, suggesting nearly complete attachment of the flow.
Figure 4 (original) presented the distribution of the root-mean-square fluctuating velocity Urms/U∞. The clean wing had two peaks corresponding to the leading-edge shear layer (y/c ≈ –0.12) and the trailing-edge shear layer (y/c ≈ 0.27). For the 10%c and 20%c cases, the two peaks merged into a single broad peak, while the 80%c case showed two distinct but very low peaks. Table 3 summarizes the peak values and their locations.
| Case | Leading-edge peak (y/c) | Leading-edge Urms/U∞ | Trailing-edge peak (y/c) | Trailing-edge Urms/U∞ |
|---|---|---|---|---|
| Clean | –0.12 | 0.22 | 0.27 | 0.18 |
| 10%c | –0.05 | 0.12 | – | – |
| 20%c | –0.10 | 0.15 | 0.10 | 0.12 |
| 40%c | –0.15 | 0.16 | 0.20 | 0.14 |
| 60%c | –0.18 | 0.14 | 0.22 | 0.13 |
| 80%c | –0.08 | 0.08 | 0.10 | 0.07 |
| 100%c | –0.12 | 0.21 | 0.27 | 0.17 |
Clearly, the 80%c case reduces the turbulent fluctuations by more than 60% compared to the clean wing, indicating a dramatic suppression of the separated shear layers.
3. Frequency Domain Analysis
To understand the energy distribution across frequencies, we calculated the power spectral density (PSD) of the velocity fluctuations at the peak locations identified above. The dimensionless form is P/(cU∞) vs. fc/U∞. Figure 5 (original) showed that for the leading-edge shear layer, the clean wing had a broad low-frequency hump, while the 20%c case showed a peak at fc/U∞ ≈ 0.3–0.4 and another weaker peak at ≈0.8. The 80%c case shifted the energy to high frequencies, with a prominent peak at fc/U∞ ≈ 3.0. For the trailing-edge shear layer, similar behavior was observed: the 80%c case exhibited the highest peak frequency. Table 4 lists the dominant dimensionless frequencies for each case.
| Case | Leading-edge (peak 1) | Leading-edge (peak 2) | Trailing-edge (peak) |
|---|---|---|---|
| Clean | – | – | – |
| 10%c | 0.5 | – | 0.5 |
| 20%c | 0.35 | 0.75 | 0.35 |
| 40%c | 0.4 | 1.0 | 0.6 |
| 60%c | 0.5 | 1.2 | 0.8 |
| 80%c | 3.0 | – | 2.8 |
| 100%c | – | – | – |
The 80%c case uniquely shifts the dominant energy to high frequencies (≈3), indicating that the bionic coverts break up large-scale structures into smaller ones. This is beneficial for fixed-wing drones because small-scale turbulence dissipates more rapidly and reduces unsteady loads.
4. Coherence Analysis
We computed the coherence function γ between the leading-edge and trailing-edge shear layer signals to assess their interaction. The coherence is defined as:
$$
\gamma = \frac{|P_{xy}(f)|^2}{P_{xx}(f) P_{yy}(f)}
$$
where Pxx and Pyy are the auto-spectral densities of the two signals, and Pxy is the cross-spectral density. A high coherence indicates strong correlation between the two shear layers. Figure 7 (original) showed that for the clean wing, coherence was low across all frequencies. For the 20%c and 80%c cases, distinct peaks appeared. In particular, the 80%c case exhibited multiple peaks at both low and high frequencies, with a notable peak at fc/U∞ ≈ 4.5. The 20%c case had a sharp high-frequency peak but lacked low-frequency coherence. Table 5 summarizes the coherence peak frequencies and magnitudes.
| Case | Peak frequencies (fc/U∞) | Maximum γ |
|---|---|---|
| Clean | – | <0.01 |
| 20%c | 4.5 | 0.045 |
| 80%c | 0.2, 0.6, 2.0, 4.5 | 0.055 |
The multiple coherent peaks in the 80%c case suggest that the coverts promote synchronization between the leading-edge and trailing-edge shear layers across a range of scales, which may enhance the cancellation of large vortices and reduce wake unsteadiness.
5. Wavelet Analysis of Multi-Scale Structures
We performed discrete wavelet decomposition on the velocity signals at the leading-edge shear layer position. The mother wavelet was the Morlet wavelet. The signal was decomposed into 10 frequency scales, with corresponding dimensionless frequency ranges as listed in Table 6.
| Scale s | Frequency range (Hz) | fc/U∞ range |
|---|---|---|
| 1 | 1.95–3.91 | 0.023–0.047 |
| 2 | 3.91–7.81 | 0.047–0.094 |
| 3 | 7.81–15.63 | 0.094–0.188 |
| 4 | 15.63–31.25 | 0.188–0.375 |
| 5 | 31.25–62.50 | 0.375–0.750 |
| 6 | 62.50–125.00 | 0.750–1.500 |
| 7 | 125–250 | 1.5–3.0 |
| 8 | 250–500 | 3–6 |
| 9 | 500–1000 | 6–12 |
| 10 | 1000–2000 | 12–24 |
We extracted the wavelet coefficients for the first 16 time units (tU∞/c) and plotted contours. The clean wing showed multiple “U-shaped” structures indicating cascades from low to high frequencies. For the 20%c case, the number of cascade steps reduced from three to two, and energy was less concentrated at low frequencies. For the 80%c case, the “U-shaped” structures became almost absent, and the highest energy appeared at high frequencies (fc/U∞ ≈ 3). This confirms that the coverts at 80%c efficiently transfer energy from large, low-frequency coherent structures to small, high-frequency ones, thereby controlling stall.
Table 7 summarizes the wavelet energy distribution across scales for the three selected cases (clean, 20%c, and 80%c). The energy percentage in each scale was computed by integrating the squared wavelet coefficients over time and normalizing by total energy.
| Scale s (fc/U∞ range) | Clean | 20%c | 80%c |
|---|---|---|---|
| 1–4 (low) | 12 | 8 | 3 |
| 5–6 (medium) | 55 | 42 | 20 |
| 7–8 (high) | 28 | 38 | 55 |
| 9–10 (very high) | 5 | 12 | 22 |
The 80%c case concentrates more than 55% of the energy in high-frequency scales (7–8), compared to only 28% for the clean wing, corroborating the frequency shift mechanism.
6. Mechanism Discussion
The bionic coverts act as passive adaptive devices driven by the incoming flow energy. When installed at 20%c, they undergo large-amplitude flapping because they are located in the highly unsteady leading-edge separation region. This flapping creates large-scale vortical structures that still persist downstream, but the velocity deficit is partially recovered. In contrast, at 80%c, the coverts are near the trailing edge where the flow is quasi-attached; they vibrate with small amplitude around a steady deflected position. This small vibration effectively “shreds” the large-scale shear layer structures into smaller scales, enhancing mixing and reducing the wake width. The mechanism can be summarized as:
$$
\text{Large-scale low-frequency disturbances} \xrightarrow{\text{coverts vibration}} \text{small-scale high-frequency turbulence}
$$
This conversion reduces the unsteady lift fluctuations and lowers the risk of deep stall for fixed-wing drones. The experimental results show that the 80%c position is the optimal location for stall control among the tested configurations.
7. Conclusions
In this work, we conducted a systematic wind tunnel study on the control of stalling flow over a NACA0018 wing using bio-inspired flexible serrated coverts, aiming to improve the performance of fixed-wing drones at high angles of attack. The main conclusions are:
- The bionic coverts significantly reduce the wake velocity deficit and turbulent fluctuations. The best performance is achieved when the coverts are placed at 80% chord, where the time-averaged velocity recovery is maximum and the wake half-width is minimized.
- Frequency analysis shows that coverts at 80% chord convert low-frequency large-scale structures into high-frequency small-scale structures, as evidenced by the shift of the dominant spectral peak from fc/U∞ ≈ 0.3 to ≈ 3.0. This conversion is beneficial for reducing unsteady loads and potential noise.
- Coherence analysis reveals that the 80%c coverts enhance the correlation between the leading-edge and trailing-edge shear layers across a broad frequency range, indicating a more synchronized wake.
- Wavelet decomposition demonstrates that the “U-shaped” cascade typical of separated flows is suppressed by the coverts, especially at 80%c, where energy is concentrated in high-frequency scales.
- The adaptive vibration of the flexible coverts, which depends on the local flow environment, is the key mechanism for energy redistribution. For fixed-wing drones, the integration of such bionic elements offers a simple, passive, and effective method to delay stall and improve flight safety.
Our findings provide a pathway for designing next-generation flow control devices for fixed-wing drones by mimicking bird plumage. Future work will investigate the scalability of these coverts for different wing planforms and Reynolds numbers, as well as their impact on lift and drag forces.