Taxi Drone Acoustic Analysis: CFD Simulation by Ansys Fluent

Description

Acoustic Analysis: Taxi Drone CFD Simulation Training

Introduction

In this project, we analyze a Taxi Drone acoustically and examine the sources of sound production. We also define receivers around the drone to observe and examine the amount of sound received by the receivers.

Taxi drones, or air taxis, are electric vertical takeoff and landing aircraft (eVTOLs) designed to transport passengers and cargo over short to medium urban distances. These vehicles are being developed with the aim of bypassing ground traffic and creating a new, sustainable network for urban air mobility (UAM).

These aircraft use multiple electric motors, enabling them to take off and land vertically (VTOL), so they do not require long airport runways. Many taxi drones are equipped with multi-sensor perception systems and preloaded 3D maps, which allow for safe navigation and autonomous or semi-autonomous flight in dense urban environments. The use of advanced batteries in these aircraft significantly reduces noise pollution and greenhouse gas emissions compared with traditional helicopters.

The geometry of the present model is three-dimensional and has been designed using SpaceClaim software. We do the meshing of the present model with Ansys Meshing software. The mesh type is Tetrahedral, and the element number is 10,595,776.

Methodology

The topic of acoustics is a very widely used and interesting topic in computational fluid dynamics. In this topic, we deal with waves and consequently with pressure.

For this project, we have used two models, BroadBand Noise and Ffowcs Williams and Hawkings (FW_H), and we have explained the settings for both models and examined the differences between the two models. First, we simulated the BroadBand Noise model steady, and after convergence and aerodynamic stability of the problem, we change the solution model to FW_H and perform the solution transient. If we activate the FW_H model from the beginning, we will hardly reach convergence and the solution may even diverge.

In the BroadBand Noise model, we extracted the Acoustic pressure level contour in decibels for the blades and in the FW_H model, we defined 7 receivers around the UAV and extracted the following results:

• Acoustic Pressure vs Time: This pressure is actually the acoustic signal calculated from the Faroukas–Williams–Hawkes (FW–H) equation, which is due to the fluctuations in the flow around the propellers.
• Sound Pressure Level: SPL is the “physical intensity of the produced sound” and is directly proportional to the sound energy, without the interference of the human ear.
• A-Weighted Sound Pressure Level: A-weighting is a filter-like function that simulates the sensitivity of the human ear. Instead of the physical SPL, the sound level in this graph is calculated to represent the “actual loudness perceived by the ear.”
• B-Weighted Sound Pressure Level: The B-weighted filter is weaker than the A-weighted filter and only attenuates a portion of the low frequencies.
• dpdt RMS: The values ​​in this contour indicate the intensity of the time-dependent pressure fluctuations at each point on the surface.

Results

In the BroadBand Noise model, we can observe the Acoustic pressure level contour in decibels. Acoustic Power Level (Lw) actually represents the sound power produced by the entire or part of the surface of an object and is expressed in decibels (dB). Comparing this contour with the Turbulent intensity contour, we find the similarity between them. In reality, the Acoustic pressure level is calculated and displayed based on the Turbulent Intensity. Therefore, wherever the Turbulent intensity is high, the acoustic pressure is also high.

Sometimes, slight differences appear between these two parameters, near the object (source area), the similarity is high and at greater distances, the patterns may differ because the intensity depends on the direction of propagation and wave attenuation.

In the FW_H model, the extracted data for one of the defined receivers can also be viewed.

SPL (Sound Pressure Level)  graph is obtained by applying a Fourier transform (FFT) to the time domain signal and shows the noise characteristics in the frequency domain. What is clearly visible are very sharp and distinct peaks at certain frequencies. These peaks, which have high sound pressure levels, indicate tonal noise.

As you can see, the first and strongest peak is at low frequencies (around 150-200 Hz). This is likely the Blade Passing Frequency (BPF). BPF is the product of the number of blades times the rotational speed (Hz) and represents the noise produced each time a blade passes in front of a fixed point.

Alongside the tonal peaks, a background noise level (below the peaks) is also observed, representing broadband noise due to flow turbulence, eddies, and other random aerodynamic phenomena. However, the tonal noise is far more dominant in this case. The overall sound pressure level decreases with increasing frequency, although tonal peaks are still evident up to high frequencies (around 4000–5000 Hz).

The graphs below show the sound pressure level in wider frequency bands (octave) and use A and B weighting filters. Both graphs have a very strong peak in the low frequency bands (around 500 Hz or less) which then decays rapidly.

Octave bands collect noise energy in specific frequency ranges rather than representing individual frequencies.

A-weighting simulates the human auditory response, meaning it reduces low and very high frequencies to make the graph more consistent with what the human ear hears. B-weighting also simulates the human auditory response, but it provides less reduction at low frequencies than A-weighting and was used more for older measurements. The most important factor in the shape of this graph is the application of the A-Weighting filter. This filter simulates the sensitivity of the human ear at different frequencies.

The human ear is most sensitive to mid-range frequencies (about 1 to 5 kHz) and hears low frequencies much less. For this reason, even if there is a lot of noise at low frequencies (below 100 Hz), the A-Weighting filter will greatly attenuate them, which is why the graph starts at very low values ​​(for example, -140 dBA). As the frequency increases, the filter attenuation decreases and the sound pressure level “perceived” by the ear rises rapidly.

The B-Weighting filter is also used to simulate the sensitivity of the human ear, but it attenuates low frequencies less than A-Weighting. For this reason, the B-Weighted graph shows slightly higher values ​​at low frequencies than the A-Weighted graph, and its slope is less steep (for example, starting at -90 dBB instead of -140 dBA). The peak is also slightly higher, because the overall attenuation of frequencies is less. B-weighting is less commonly used in modern analysis, and A-weighting and C-weighting are more common, but it gives us a similar perspective to A-weighting, with a slightly greater emphasis on lower frequencies.

Finally, the SPL is the actual sound intensity from the drone, derived from the pressure field calculated by FW–H, the A-weighted SPL (dBA) is the sound level that humans actually hear (the hearing threshold is taken into account), and the B-weighted SPL (dBB) is an approximation of the perceived sound at medium intensities. Comparing these three shows the difference between the physical energy of the sound and the perceived energy; in drones, the lower frequency (BPF) is usually clear in SPL but less so in dBA.

The Acoustic Pressure graphs show the time variations of the sound pressure ( p’(t) ) at the receiver location. This pressure is actually the acoustic signal calculated from the Ffowcs–Williams–Hawkes (FW–H) equation, which is due to the fluctuations in the flow around the propellers.

This plot shows that the acoustic field at the receiver point of interest has reached a steady or quasi-steady state after a very short transient phase at the beginning of the simulation. This stability usually occurs when the fluid flow (CFD) underlying the FW-H analysis has reached a steady state or at least is statistically stable. This means that the noise sources (pressure and velocity changes on the surface) do not change significantly over time.

This sinusoidal or nearly sinusoidal and periodic waveform indicates the presence of one or more dominant frequencies (Tonal Noise) in the noise signal. Each pressure pulse is likely to be associated with the passage of a vane or rotating component past a specific point.

The last data that we will discuss in this report is the dpdt RMS contour. The values ​​in this contour indicate the intensity of the time-dependent pressure fluctuations at each point on the surface

High dp/dt values ​​indicate rapid and severe pressure changes at those points. These rapid pressure changes are the main source of acoustic noise generation. The concentration of high dp/dt values ​​at the blade tips and edges is quite logical and expected for a rotating propeller.

This contour clearly shows the physical sources of noise that lead to the tonal and broadband peaks observed in the frequency domain plots. That is, these pressure fluctuations at the propeller surface propagate sound waves into the environment that we see in the SPL plots.

This description is a brief overview of one of the receivers defined in this project. In the training video, we analyze and explain the extracted data and compare them with each other in more detail.