Hybrid Rotor Noise Optimization ^†

Abstract

This study examines noise reduction strategies for unmanned aerial vehicles (UAVs) in urban air mobility applications, with a particular focus on package delivery. By employing a combination of low-, mid- and high-fidelity aerodynamic and aeroacoustic analyses, this research investigates the impact of rotor design parameters, including blade spacing and rotational speed, on noise emissions. The results demonstrate that an increase in rotor spacing results in a reduction in noise emissions. By adjusting the blade chord and twist within an optimization loop, it was possible to decrease tonal noise, yielding a Sound Pressure Level (SPL) reduction of about 3.51 dB while improving propulsive efficiency by 39%. These findings highlight the importance of rotor geometry optimisation during the early design stages in order to meet both performance and noise requirements.

1. Introduction

Urban air mobility (UAM) envisions the use of electric aircraft with Vertical Take-Off and Landing (VTOL) capability to address urban transportation challenges, such as congestion and pollution. One promising application is package delivery in dense urban environments, where means of ground transportation struggle. Autonomous Unmanned Aerial Vehicles (UAVs), commonly known as drones, offer a potential solution due to their ability to bypass traffic, enabling faster deliveries. However, challenges, e.g., noise emissions, remain significant. The VTOL capability is achieved through multirotor configurations, including multicopters and hybrid designs, with noise primarily generated by rotor aerodynamics and aeroacoustics. Thus, minimizing noise during the design phase is significant, necessitating early noise prediction. Recent research emphasizes the complexity of noise sources in small-scale UAVs, including tonal self-noise, blade vortex interactions (BVIs) and turbulence interaction with blade edges [1]. These sources are particularly challenging for low-Reynolds-number UAV rotors, where phenomena including flow separation and re-circulation zones impact noise emissions [2]. This study builds on recent advancements by developing a UAV optimized for noise minimization from the outset, exploring rotor design innovations and noise-predictive models. The objective is to determine how rotor parameters, such as rotational speed and rotor blade spacing, affect noise generation, using a combination of low-fidelity aerodynamic analysis for configuration design and mid-fidelity aeroacoustic analysis for noise evaluation using the FLOWUnsteady code [3]. In this paper, the terms ‘rotor’ and ‘propeller’ are used interchangeably. Initial results indicate that increasing rotor spacing significantly reduces noise emissions, with the optimal configuration featuring a hub-to-hub distance of twice the diameter length (2D).

2. State of the Art

In rotor aerodynamics, low- and mid-fidelity tools such as Blade Element Momentum Theory (BEMT) are standard for the conceptual design phase. BEMT allows for rapid analysis of rotor designs and optimization [4,5], making it well suited for UAV design [6]. Several studies have used BEMT for rotor performance analysis for low Reynolds numbers (Re), relevant for small-scale UAVs [7,8,9,10]. Additionally, new research highlights the role of recirculation zones and unsteady flow separation at noise generation in these small UAV rotors [2]. For aeroacoustic analysis, the Hanson approach, based on Goldstein’s acoustic analogy, is commonly used in preliminary design to predict noise emissions [11,12,13]. Recent studies have shown that closer rotor proximity increases noise emissions, especially in multirotor configurations, with tip-to-tip distances having a significant impact on noise levels (see refs. [14,15,16]). Moreover, computational methods such as the Ffowcs Williams–Hawkings (FW-H) model have proven effective for predicting loading noise in rotor configurations, while Vortex Particle Methods (VPMs) improve wake interaction modelling, resulting in more accurate noise predictions [2]. Recent advancements have also introduced innovative rotor designs, such as toroidal rotors, aimed at reducing broadband noise while maintaining aerodynamic efficiency [17]. These developments suggest a promising direction for future UAV rotor designs, balancing performance with acoustic considerations.

3. Methodology

3.1. Conceptual Design Methodology

This section outlines a conceptual design methodology for various aerial vehicle configurations in terms of UAVs and air taxis, developed by the Institute of Aerospace Systems at RWTH Aachen University [6,18]. The approach uses a database of configurations (e.g., fixed-wing, VTOL) to establish design relationships, addressing the limitations of aircraft-derived equations. User-defined inputs, such as Top-Level Aircraft Requirements (TLARs), configuration, mission profile and regulatory constraints, guide an iterative process. Initial sizing and aerodynamic analysis using XFOIL are followed by mission analysis to estimate power and energy demands. Mass estimation compares semi-empirical equations with statistical data until convergence on Maximum Take-Off Mass (MTOM) is reached and design variables are updated. A detailed methodology is provided in [18].

3.2. Numerical Approach

A toolchain for analyzing and optimizing local and global aerodynamic and acoustic behaviour of isolated and distributed propellers is introduced to achieve high aerodynamic efficiency and low propeller noise. The toolchain is based on a numerical approach, where aerodynamic and aeroacoustic analyses are performed using low-, mid- and high-fidelity methods based on the given geometry and mission profiles from the conceptual design stage. The numerical approach uses BEMT methods to investigate and optimize propeller shapes. A detailed description of the methodology is found in Section 4.

3.3. Aerodynamic Validation

For aerodynamic validation against experimental data, a comprehensive analysis was conducted, focusing on both isolated and distributed propellers. To ensure detailed validation, two distinct propeller geometries in tandem arrangement were selected: (i) the DJI9443 propeller (only acoustic validation) and (ii) the DJINing propeller, a modified propeller based on the DJI9443 geometry with slightly differing geometry parameters.

3.3.1. Rotor–Rotor Interaction

When focusing on rotors operating in complex environments, such as those involving interaction effects from nearby rotors, it is essential to investigate the thrust behaviour in relation to the isolated propeller case. By adjusting the separation distance between the rotors, a more compact vehicle design becomes possible, fulfilling the urban setting requirements. However, this leads to complex aerodynamic phenomena between adjacent rotor tips. These phenomena significantly influence the wake structures (like wake interactional effects) and flow field, directly affecting the aerodynamic performance and noise level of the multirotor [19]. Further investigation is necessary to identify a compromise between vehicle compactness and aerodynamic performance. The validation of the current methodology by using FLOWUnsteady is conducted for a side-by-side rotor configuration using the DJINing propeller [3]. This analysis is conducted by comparing the resulting thrust coefficient of the rotor ($c_t$) to the thrust of the isolated rotor ($c_{t_{iso}}$) across different separation distances with experimental data.

Figure 1 shows the thrust coefficient for the DJIing propeller mounted in a tandem array at 4680 RPM, comparing varying rotor separation distances with the mid-fidelity results and experimental data in Zhou et al. [16]. The results indicate that reducing the tip distance of the DJI rotor leads to a notable thrust decrease of approximately 4% compared to the 2D case. This thrust reduction is strongly influenced by rotor interaction effects and the complex aerodynamic environment, in which the rotor operates. The slight discrepancy between the experimental data and the data obtained through FLOWUnsteady could be linked to the lack of blade geometry parameters [14]. For this reason, the behaviour is analysed qualitatively.

3.4. Acoustic Validation

Following the presented successful validation of aerodynamic parameters, this section briefly introduces the acoustic workflow used within the mid-fidelity approach and the acoustic validation. Despite the validation of the acoustic approach via PSU-WOPWOP [20] against the experimental results by Alvarez et al. [14], an additional validation of this methodology tailored to this work is carried out. This validation encompasses an acoustic analysis of the two DJI propeller types introduced before. The acoustic analysis follows a procedure similar to the aerodynamics analysis. This means that the evaluation of sound parameters was performed for both the isolated and distributed propeller cases. By using the aerodynamic solution obtained within FLOWUnsteady presented in Section 3.3.1, the time-resolved loading can be transferred to PSU-WOPWOP [20] and BPM (Brooks, Pope and Marcolini method) [21] to compute noise. PSU-WOPWOP deals with tonal noise, whereas PBM targets broadband noise. The aim of Section 3.4 here is to validate the introduced workflow for the isolated DJI9443 propeller in hover at 5400 RPM as well as for the DJINing propeller at a tip-to-tip distance of 0.5D at 4860 RPM to refs. [16,22], by focusing on tonal and broadband noise. The tonal SPL is evaluated for the 1st Blade Passing Frequency (BPF). The Overall Sound Pressure Level (OASPL) is related to the sum of tonal and broadband noise sources. On the one hand, Figure 2 demonstrates reasonable agreement for the isolated DJI9443 propeller when comparing the results obtained from FLOWUnsteady and PSU-WOPWOP with experimental data. On the other hand, it depicts the most complex scenario in terms of aeroacoustic interaction effects for the DJINing propeller in the tandem arrangement. This is due to high tip interactions resulting in highly unsteady behaviour such as vortex shedding. In highly unsteady conditions, a comparison of the SPL and OASPL values can be performed for experiments, confirming the high accuracy of the introduced workflow. With the successful aerodynamic and acoustic validation of the methodology, Section 4 focuses on the results obtained from the application of this methodology to a case study. The selected case study is a small quadcopter configuration in hover developed with the methodology described in Section 3.

4. Results

4.1. Aerodynamic Optimization

As briefly mentioned in Section 3.2, the low-fidelity approach incorporates an optimization loop based on BEMT [5] and the well-established design methods of Borst [23]. Glauert’s propeller theory [24] and Betz’s condition [25] for minimizing induced loss, where the contraction of the wake is neglected to achieve minimal energy loss, served as a foundation for subsequent work. Adkins and Liebeck [5] used these principles to develop a design algorithm for determining the propeller shape that minimizes induced loss (MIL) while achieving a specified thrust. This approach maximizes aerodynamic efficiency [26]. Additionally, a Python-based framework was introduced in [26,27,28] for implementing these methods. This tool, developed by the Chair of Aerodynamics and Fluid Mechanics at the Technical University of Munich (TUM), serves as the foundation for the optimization loop designed for local airfoil shapes. While the aerodynamic optimization is based on section-wise 2D MIL/BEMT methods, the acoustic optimization follows the Hanson approach [11].

The optimization algorithm significantly reduces the tonal Sound Pressure Level (SPL) at a predefined observer position while maximizing the sectional local glide ratio (epsilon) by adjusting the radial blade geometry parameters, starting from the initial geometry as displayed in Figure 3. A local optimum for the sectional blade aerodynamics behaviour and the resulting SPL level is found through the tool. The local aerodynamic optimum is constrained by geometric parameters, specifically the chord and twist distribution. The chord at the propeller root is fixed at r/R = 0.125 and at the tip at r/R = 0.025 to meet manufacturing and structural requirements. It should be noted that structural behaviour is not considered within this local optimization loop. Starting from a target thrust of the isolated propeller in hover conditions of 12.4 N for the initial quadcopter configuration, a glide number error called ${\epsilon }_{err}$ is set. The calculation loop starts if the target error is not met. Within the loop, for a given number of iterations, the aerodynamic calculations following the MIL approach are conducted. After each iteration, the SPL for a given observer position underneath the rotor at an angle of −45° with a normalized distance of ${\displaystyle \frac{r_{Propeller}}{r_{Microphon{e}_{array}}}}=0.1413$ is calculated using the Hanson approach [11]. If the calculated SPL value exceeds the artificially set SPL limit, the effective angle of attack (AoA) within each section, resulting from the current pitch and flow angle, is artificially increased by 0.12°. The chord is then decreased by 20%, resulting in a reduction in tonal noise. This iterative process continues until either the SPL or the glide ratio limit is achieved as shown in Figure 4.

Within the optimization process for a propeller blade at 10 m/s axial inflow, it was found that the initial RPM given by RWTH Aachen for the drone reference configuration could be reduced by approximately 10% in the hover condition without increasing the risk of not meeting the thrust requirements. As a result of the BEMT in hover, the reduction in rotational speed resulted in a significant reduction in SPL of 3.51 dB while increasing propulsive efficiency $\eta$ by 39%.

4.2. Preliminary Drone Configuration—Initial and Optimized Propeller Investigations

This section introduces the designed configuration, which constitutes the case study for the aeroacoustics prediction. The results of the rotor and rotor-on-structure interactional effects are presented comprehensively for the initial and optimized propeller geometry. A quadcopter configuration with single open rotors and minimal interaction between the fuselage and the rotors is designed.

This approach was chosen to begin the analysis with a simpler configuration, enabling an in-depth examination of design- and configuration-specific parameters involved in aeroacoustic phenomena. Specifically, in simple configurations, high rotor-on-rotor and rotor-on-structure interaction effects are minimized, allowing for a detailed study of the dominant aerodynamic and aeroacoustic effects. Additionally, this provides a basis for further analysis considering more complex configurations. The TLARs set for designing an electric UAV with VTOL for delivery in urban settings are max. payload of 2 kg and max. range of 20 km. First, a detailed analysis of the propeller is conducted for an isolated and distributed propeller case in hover, both for FLOWUnsteady and partially URANS results. The labeling of the rotors for both mid- and high-fidelity can be taken from Figure 5. The need for a compact design is deemed a constraint for operation in urban settings. A conventional quadcopter configuration was chosen (see Figure 5) to perform a mission encompassing take-off, vertical climb, forward flight, descent and landing. In case of emergency, the drone can hover for 5 min. The application of the conceptual design methodology provides the mass breakdown, power and energy requests and geometry of the envisaged configuration. In particular, geometry data encompass size and dimensions, rotor arrangement and rotor blade geometry, including chord and twist angle distribution. With these geometry parameters, a sketch of the UAV can be derived. Outputs of the conceptual design tool also include the values of thrust and torque for the given design mission.

4.3. Isolated and Side-by-Side Aerodynamic and Acoustic Effects

In the following section, detailed hover investigations in terms of aerodynamics and acoustics analysis are performed using the FLOWUnsteady mid-fidelity as well as the CFD high-fidelity approach. The investigations are conducted for an air density of 1.21 $\rho$ [kg/m³], axial velocity U of 0 [m/s], a rotational speed of 5730 RPM [min⁻¹] for the initial and 5200 RPM [min⁻¹] for the optimized propeller at an AoA of 0 [∘] and a target thrust $T_{target}$ of 12.4 [N]. Various rotor arrangements are considered in the aerodynamics investigation, where this work focuses on the hover case for isolated and side-by-side propeller arrangements. The aerodynamic characteristics of the initial and optimized propeller geometry in hover for the isolated and distributed propeller are depicted in Figure 6.

Figure 6 compares the thrust coefficient $c_t$ at 5730 RPM and 5200 RPM for various horizontal and vertical (horizontal distance between hubs = 1.725D) separation distances. A partial comparison between mid-fidelity and high-fidelity URANS results indicates that FLOWUnsteady predicts the overall thrust with high accuracy when using CFD as the validation reference. URANS results are plotted as averaged values. The high thrust coefficient of the isolated optimized rotor in hover is attributed to the increased efficiency described in Section 3.2 and the specified flight conditions, for which the propeller was optimized. Even in more complex environments, such as the 1.725D case, the results from FLOWUnsteady and URANS are comparable. Figure 6 reveals a clear trend for the thrust coefficient in both FLOWUnsteady and URANS results, showing that reducing the propeller spacing leads to a decrease in thrust of up to 2.2% and an increase in thrust fluctuations. Figure 7 refers to the analysis of the initial propeller geometry by comparing various propeller arrangements by focusing on FLOWUnsteady and URANS results. When comparing the “4R” configuration to the “fuselage” and “Pull” drone configuration, a clear impact on the thrust coefficient $c_t$ and its behaviour in terms of unsteadiness can be seen. While the “4R” configuration shows comparable thrust values referred to the isolated propeller case, the several drone configurations lead to a clear decrease in thrust and an increase in thrust variation. The highest variations can be seen for the “Pull” configuration in hover. For the acoustic analysis, Figure 8a,b present the SPL values for the first BPF obtained using the PSU-WOPWOP (FLOWUnsteady) and FW-H (URANS) methods, focusing on the isolated initial and optimized propeller geometry as well as the full drone reference configuration (see Figure 5) in hover. The observer position for the aeroacoustic analysis in hover was set at the same parameters set in Section 4.1, resulting in a microphone array radius of 2.83 m. When comparing the isolated initial propeller (5730 RPM) to the optimized propeller (5200 RPM) results, a clear noise reduction of about 3.5 dB while meeting the same thrust requirements is achieved. Given that result, it can be concluded that the optimized propeller has high potential to be used in various propeller arrangements for further analysis.

5. Conclusions

This study examines the optimisation of the aerodynamic and aeroacoustic characteristics of a delivery UAV. By employing a combination of low-, mid- and high-fidelity aeroacoustic analyses, it was demonstrated how blade spacing, rotor geometry and rotational speed can influence noise emissions and aerodynamic efficiency. Validation of the aerodynamic and aeroacoustic characteristics was conducted using FLOWUnsteady and PSU-WOPWOP. An increase in rotor hub-to-hub spacing is an effective method for reducing noise emissions. The optimisation process on the chord and twist angle provided a noise reduction of 3.51 dB at the predefined observer position. Moreover, this optimisation also enhanced propulsive efficiency by 39%, thereby demonstrating that noise reduction can be achieved without affecting aerodynamic performance. Additionally, this study investigated the interactions between rotors and their surrounding structures. Reducing the separation distance between rotors to 1.1D can result in a thrust reduction of up to 2.2%. The optimised propeller demonstrated a notable enhancement in both thrust and noise performance when positioned side-by-side. These findings emphasise the importance of optimizing rotor geometry to achieve noise mitigation from the early stage of the UAV design process.