Harry Liuhan

Vortex shedding—the repeated release of swirling fluid behind an object—does not only affect mean aerodynamic performance: it also creates rapidly fluctuating aerodynamic forces that can contribute to vibration, fatigue and noise. In this project, I investigated whether deliberately introducing a three-dimensional wavy trailing edge could passively suppress these fluctuations without requiring an active control system.

Using three-dimensional large-eddy simulations, I studied a NACA 0012 aerofoil at Re = 5,000 and an angle of attack of 9°. A straight trailing edge was compared with sinusoidal geometries of four wavelengths, λ/c = 0.15, 0.20, 0.30 and 0.60, allowing me to test whether the aerodynamic response depended on matching the imposed geometry to characteristic spanwise scales within the wake.

The estimated reduction varied between wavelengths under the tested conditions. The λ/c = 0.30 geometry reduced RMS lift fluctuations by 9.66%, with a 95% confidence interval of 2.98–15.46%, while producing essentially no change in the dominant shedding frequency and only a very small change in mean drag. It was the only tested wavelength for which the confidence interval of the lift-fluctuation reduction excluded zero. However, the intervals for different wavelengths overlapped, so this does not establish a uniquely best wavelength.

I then separated the fluctuating force into contributions from different sections of the wing to investigate possible explanations for the reduction, including weaker local loading and changes in how sectional forces combine. This analysis explored the mechanism behind the measured result; it does not establish a single explanation that applies across other wing designs or operating conditions.

My first erosion study showed computationally that very small geometric changes could produce unexpectedly large aerodynamic effects. I wanted to determine whether those effects would survive the transition from an idealised two-dimensional simulation to a physical, rotating propeller.

I designed the experiment from the ground up, including a three-bladed fixed-pitch propeller based on a triangular aerofoil and a purpose-built thrust-testing rig. The propellers were additively manufactured with controlled erosion to their leading edge, upper apex, or combinations of both. I then measured thrust and power across Reynolds numbers of approximately 20,000–40,000, while the updated study also incorporated validation of the thrust-measurement system and additional mixed-mode erosion data.

The experiments showed that erosion could either improve or degrade performance depending strongly on where it occurred, its magnitude and the operating Reynolds number. In these static measurements, leading-edge erosion generally reduced thrust per unit of measured power, whereas removing material from the upper edge could reduce power consumption and increase this ratio under some tested conditions. Mixed erosion generally behaved as a combination of the individual modes rather than as an entirely separate response.

These static tests show how the measured thrust-to-power ratio changed under the tested conditions. Forward-flight performance would need separate testing.

When fluid flows around a bluff body, vortices periodically detach from alternating sides and produce oscillating forces. If this vortex-shedding frequency approaches a structure’s natural frequency, the resulting vibrations can become significant. I investigated how the geometry of polygonal structures affects both the frequency and variability of this shedding.

I created computational models of squares, hexagons and octagons at a Reynolds number of 200, testing each polygon at three angular orientations. My initial hypothesis was that increasing the number of sides would make vortex shedding less sensitive to orientation because the possible movement of the flow-separation point becomes progressively smaller.

The simulations supported this: from the square to the octagon, the range over which the separation point moved fell by 83%, while the standard deviation of shedding frequency across orientations decreased by 90%. However, the hexagon produced an unexpected result: the ordering of its maximum and minimum shedding frequencies was reversed relative to the square and octagon. Investigating this anomaly led me to propose an alternative way of classifying polygon orientation based on the geometry near flow separation rather than the conventional rotational labels.

These two-dimensional simulations show how polygon geometry affected sensitivity to orientation at a Reynolds number of 200. Testing at other flow conditions and in three dimensions would be needed before applying the numerical findings to full-scale structures such as pipelines or buildings.

Small aircraft and Martian aerial vehicles operate at low Reynolds numbers, where aerodynamic performance is particularly sensitive to changes in aerofoil geometry. Yet while erosion has been extensively studied for conventional aircraft and wind turbines, its effects on low-Reynolds-number aerofoils remain much less understood. I investigated whether relatively small amounts of erosion could significantly alter the performance of a triangular aerofoil designed for this flow regime.

I developed a two-dimensional computational fluid dynamics model of the aerofoil at a Reynolds number of 10,000 and simulated erosion at different locations and severities. Using lift, drag and surface-pressure distributions, I examined not only whether performance changed, but also how the location of the erosion altered the surrounding flow.

The results showed that aerodynamic performance was highly sensitive to where erosion occurred. At a Reynolds number of 10,000, across the tested angles from −3° to +6°, lift changed by between −35.2% and +23.2%, while drag changed by −16.0% to +28.8%. Leading- and top-edge modifications produced particularly strong changes to the suction-side pressure distribution, while top-edge erosion unexpectedly improved lift-to-drag ratio at positive angles of attack above approximately 2°.

Within the tested two-dimensional model at a Reynolds number of 10,000, erosion was not uniformly detrimental: its effects depended on the geometry and angle to the airflow. This raised a broader question that motivated my later experimental work: would these computational trends persist on a real rotating propeller?