A COMPARISON OF MEASURED AND MODELLED AIRCRAFT NOISE LEVELS FOR RTHA

Abstract

To reduce the growing distrust in aircraft noise models felt by communities around the airport, it is imperative to ensure accurate modelling methodologies validated by appropriately measured noise metrics. This is especially crucial in regions farther from the airport where L den = 45-55 dBA because the amount of affected residents in these areas is large. Currently, there is a lack of measured noise levels at such distances and uncertainty about the assumed procedures, such as the aircraft thrust settings. Regarding the latter, before comparing the model and measured noise levels, it's thus crucial to first create a robust workflow for obtaining accurate input data for the noise predictions. In this contribution , as a first step, audio files from the noise monitoring stations around Rotterdam The Hague airport (RTHA), combined with dedicated array and single microphone measurements, are considered for extracting fan rotational speed, N1. The 64-microphone array and the single microphone system were co-located with one of the monitoring stations at a distance of 1.14 km away from the RTHA runway. The engine settings are retrieved from the intensity-averaged spectrograms obtained from the microphone array. Using the derived thrust settings, the noise levels measured by the monitoring stations are compared with the single-event noise level prediction made by the European Noise model, Doc.29. The aircraft position, i.e., input for the model, is obtained from ADS-B data, which contains the position vector and velocity of the aircraft at 1-second intervals. In the framework of this study, noise predictions for both arrival and takeoff procedures for three aircraft types are presented. Finally, this case study aims to investigate the applicability of the data from monitoring stations for the aim of model-data predictions at the mentioned regions.

Full text

30th International Congress on Sound and Vibration
A COMPARISON OF MEASURED AND MODELLED AIRCRAFT
NOISE LEVELS FOR RTHA
Anandini Sravya Jayanthi
Aircraft Noise and Climate Effects, Delft University of Technology. Kluyverweg 1, 2629 HS, Delft, The Netherlands
E-mail: A.S.Jayanthi@tudelft.nl
Rebekka van der Grift
Aircraft Noise and Climate Effects, Delft University of Technology. Kluyverweg 1, 2629 HS, Delft, The Netherlands
E-mail: R.C.vanderGrift@tudelft.nl
Irene Dedoussi
Aircraft Noise and Climate Effects, Delft University of Technology. Kluyverweg 1, 2629 HS, Delft, The Netherlands
E-mail: I.C.Dedoussi@tudelft.nl
Mirjam Snellen
Aircraft Noise and Climate Effects, Delft University of Technology. Kluyverweg 1, 2629 HS, Delft, The Netherlands
E-mail: M.Snellen@tudelft.nl
To reduce the growing distrust in aircraft noise models felt by communities around the airport, it
is imperative to ensure accurate modelling methodologies validated by appropriately measured noise
metrics. This is especially crucial in regions farther from the airport where Lden = 45 -55 dBA be-
cause the amount of affected residents in these areas is large. Currently, there is a lack of measured
noise levels at such distances and uncertainty about the assumed procedures, such as the aircraft thrust
settings. Regarding the latter, before comparing the model and measured noise levels, it’s thus crucial
to first create a robust workflow for obtaining accurate input data for the noise predictions. In this con-
tribution, as a first step, audio files from the noise monitoring stations around Rotterdam The Hague
airport (RTHA), combined with dedicated array and single microphone measurements, are considered
for extracting fan rotational speed, N1. The 64-microphone array and the single microphone system
were co-located with one of the monitoring stations at a distance of 1.14 km away from the RTHA
runway. The engine settings are retrieved from the intensity-averaged spectrograms obtained from
the microphone array. Using the derived thrust settings, the noise levels measured by the monitor-
ing stations are compared with the single-event noise level prediction made by the European Noise
model, Doc.29. The aircraft position, i.e., input for the model, is obtained from ADS-B data, which
contains the position vector and velocity of the aircraft at 1-second intervals. In the framework of this
study, noise predictions for both arrival and take-off procedures for three aircraft types are presented.
Finally, this case study aims to investigate the applicability of the data from monitoring stations for
the aim of model-data predictions at the mentioned regions.
Keywords: environmental noise, aircraft noise, noise model validation, munisense microphone system
1. Introduction
The aviation industry is rapidly developing to accommodate passengers of all socioeconomic condi-
tions [1]. However, these benefits also bring many detrimental long-term climate change challenges and
1

short-term environmental challenges related to air quality and noise exposure. Community resistance
towards expanding aviation operations due to noise exposure is one of the immediate challenges faced
by the industry. The proposed aircraft noise reduction measures are threefold: improving aircraft de-
sign and producing new technologies, optimizing aircraft operations, and improving policy-making and
management [2].
Changing aircraft design is an expensive long-term solution, while the other two measures are more
applicable for resolving the current challenges, requiring precise noise exposure prediction. Hence, ac-
curately measuring and predicting aircraft noise is crucial for facilitating precise noise control limits and
evaluating noise exposure from various noise abatement procedures. The noise calculations are made
according to the guidelines prescribed in the ECAC Doc.29 [3]. They are described by cumulative noise
metrics such as the day-evening-night average level Lden, but are only applicable for average long-term
exposure. According to the 2018 WHO recommendation, major European airports must adhere to a limit
of Lden = 45dBA, as aircraft noise above this level may cause adverse health issues among the exposed
communities [4]. However, this metric is less suitable for model-data comparisons since comparing the
measured averaged exposure with corresponding model predictions cannot reveal the effects of varying
parameters, such as aircraft types and operational procedures, on the model-data agreement. This limits
the ability to determine the conditions under which the model predictions are less accurate. Hence, it is
necessary to predict single-metric noise levels for individual flyovers.
To improve and extend the prediction capabilities of Doc.29, the assumptions made must be verified,
and the input must be validated and updated. This input is typically obtained from the Aircraft Noise
and Performance (ANP) database subset, i.e., the Noise-Power-Distance (NPD) tables. These tables are
curated and extrapolated for a cluster of similar aircraft types with measurements during the aircraft cer-
tification process at various standard operation stages and atmospheric conditions. Choosing the most
accurate source of input, i.e., power and distance, is necessary to validate the listed noise levels. There-
fore, as shown in Figure 1, this paper aims to establish a robust workflow by obtaining accurate input for
predicting single-metric noise levels with an accuracy of less than 2 dBA. The noise level validation step
is carried out using two types of measuring equipment: A noise monitoring station (NMT) and a Mu-
nisense microphone. The radar and ADS-B data validate the distance, and the thrust settings are retrieved
from the processed output of a third measurement system, an acoustic array. Then, the noise levels mea-
sured during a measurement campaign around Rotterdam The Hague Airport (RTHA) are compared with
the predicted values. Furthermore, the applicability of the noise-measuring setups toward model-data
comparisons for the future is determined through qualitative and quantitative comparison.
Figure 1: Each key input, i.e., noise, power, distance, is validated with measurements from different
sources.
ICSV30, Annual Congress of International Institute of Acoustics and Vibration (IIAV), 8-11 July 2024

2. Measurement setup
2.1 Noise Monitoring Stations
While being an international airport, Rotterdam The Hague Airport (RTHA) houses only one runway
and is located in a densely populated area. Therefore, it is an appropriate case study for validating the
noise model at varying observer distances. RTHA is surrounded by six fixed NMTs. The measurements
were conducted on 5th Septemeber 2023 near Veldkersweg, Rotterdam, represented by a red circle in
Figure 2. The maximum wind speed was 4.63 m/s, and the temperatures ranged from 15◦C to 29◦C, with
relative humidity reaching a maximum of 49%. The total number of flyovers measured with the aircraft
type, engine type, and procedure are given in Table 1.
Table 1: List of flyovers measured according to Aircraft type, Engine type, and operating procedures.
Aircraft type Engine type Landing Take-off
EMB190 CF34-10E (82kN) 2 1
B737-700 CFM56-7B22 (101kN) 1 3
B737-800 CFM56-7B27 (121kN) 2 4
2.2 Munisense and microphone array
The measurements obtained from the NMT are of two types: the noise levels and the audio recordings
of the flyovers, which are processed into spectrograms. The noise levels are validated with the Munisense
microphone. Due to uncertainties in the sensitivities of the array microphones, the measurements of the
microphone array will only be used to determine the spectrogram up to higher frequencies than for
Munisense and the NMT due to the array’s higher sampling frequency; both the array and the Munisense
are shown in Figure 2. Since the spectral information obtained from the recorded flyovers by the selected
NMT contains interference due to ground reflections, spectrograms obtained from the microphone array
are used instead. The specifications of these three setups are described in Table 2.
Figure 2: (left to right) RTHA with NMTs where the red dot represents the location of the chosen NMT,
and the blue line represents the runway; Portable munisense microphone; 2D array with 64
microphones.
ICSV30, Annual Congress of International Institute of Acoustics and Vibration (IIAV), 8-11 July 2024
3

Table 2: Types of measurement equipment and their specifications.
Measurement Setup Specifications Height [m] Sample frequency [Hz]
Noise Monitoring Station Fixed 9 8000
Munisense Portable & light 1.9 48000
Phased microphone array Portable & heavy 0.05 50000
3. Doc.29 Model
In ECAC Doc.29 (Volume 2) [3], modelling guidelines for best-practice noise models are stated. This
model creates noise contours of day-evening-night averages Lden for a year. To calculate Lden, the Sound
Exposure Level (SEL) is modelled for all flights and data points around the airport grid. The SEL for
all flights is logarithmic summed and averaged for the entire year. As mentioned before, best-practice
models are based on NPD tables. For different combinations of power settings and distance between the
aircraft and the ground, noise values are given in four different metrics. For this research, the two main
metrics are the maximum A-weighted sound pressure level LA,max and the SEL of an aircraft noise event.
To estimate these metrics for different locations on the ground, a flight path is divided into segments, for
example into 1-second intervals. For each interval, a new noise level is determined using the NPD table.
At each data point in the grid, the LA,max is calculated by taking the maximum value found during the
flight path, and the SEL is calculated by a summation of the noise from each segment. Correction factors,
described in chapter 4 of Doc.29 [3], are then applied to retrieve the final values.
The power and distance input needed for the NPD tables is obtained by modelling a flight track. The
distance variable is easily obtained by calculating the slant distance dbetween the aircraft and all the
grid points on the ground for each segment. For the power input, thrust procedures are used to estimate
the power setting of the specific aircraft type. These procedures consist of several steps based on the
aircraft’s altitude, speed, and weight. In the creation of Lden contours, deriving every single flight and
computing a noise grid for it for an entire year can become computationally expensive. Hence, in best-
practice modelling, flights of similar aircraft types, routes, and procedures are bundled into a cluster. This
cluster is then divided into a main track and subtracks based on historical data. An example of methods
to analyze historical tracks and how to create airport-specific profiles can be found in Heblij et al. [5]
For year-round averages, the above-described methods for track modelling have proven to be accurate
to 1-2 dB [6]. However, when comparing single-event measurements to these general flight tracks, devi-
ations can become large. Other methods to derive accurate thrust and distance input to the noise model
are developed for this research.
4. Obtaining modelling input
4.1 Positional data
Distance is obtained from the flight positional information recorded by ADS-B transponders, which
are placed on an aircraft and retrieved by an open-source network known as OpenSky [7]. The correlated
information of individual noise events with aircraft time stamps and tail numbers is retrieved from the
Trino database in pyopensky library. The relative distance between the noise source and the microphone
array is calculated from the aircraft’s spatial position, i.e., latitude, longitude, and altitude. ADS-B data
has considerable advantages over radar track data because it provides spatial information of the flight at
a time interval of 1 s and has better accuracy in the area around the runway.
ICSV30, Annual Congress of International Institute of Acoustics and Vibration (IIAV), 8-11 July 2024

4.2 Engine settings
The engine setting required for noise prediction is the thrust setting or its related normalized fan
rotational speed. Employing Equations 4.1, thrust values are obtained by calculating N1% from the
fan’s Blade Passing Frequency (BPF). The BPF and its higher harmonics are detected from spectrograms
obtained from the noise levels recorded by the microphone array [8].
n=60 ∗BP F
B
N1% = n
nmax
(4.1)
where Bis the number of blades in the specific fan, n(rpm) is the fan rotational speed, and nmax (rpm) is
the maximum permissible rotational speed of the fan at 100% N1 (values are provided in Table 3).
Table 3: Calculated values of N1% for the given aircraft types defined by the number of blades B.
Aircraft type B nmax[r pm]Landing N1 [%] Take-off N1 [%]
EMB190 29 5955 30.42 67.057
B737-700 24 5175 50.65 88.29
B737-800 24 5175 57.62 93.866
Figure 3: Intensity-averaged spectrograms of the landing of B737-700 (top) and take-off of B737-800
(bottom) without Doppler correction (left) and with Doppler correction (right) where dashed black lines
represent the BPF and higher harmonics.
The spectral information is computed by averaging the total intensity calculated for selected micro-
phones. The spectrograms are plotted with the sampling frequency of 50000 Hz, 5000 sampling points/-
time block, and Hamming windowing. The spectrograms for all the flyovers are studied, and since no
significant spectral information in the range of 8 - 22 kHz is observed, harmonics only up to 8 kHz are
ICSV30, Annual Congress of International Institute of Acoustics and Vibration (IIAV), 8-11 July 2024
5

considered. The estimated values of N1% engine setting required as the input for Doc.29 predictions are
summarized in Table 3. Once the N1% values of the flyovers are retrieved, the thrust of the aircraft can be
derived to get the required input for the Doc.29 model. For the conversion of N1% to thrust, this research
assumes a quadratic relation between the percentage of maximum rotational speed and the percentage of
maximum thrust.
As seen in Figure 3, the measurements of landing and take-off flights exhibit Doppler shifts in fre-
quencies, and the corrected fan tone and harmonics, as described by dashed black lines, are obtained
by
f′
f=1
1 + V2ti
dc
ti=te+dmin
vc
(4.2)
where f′(Hz) and f(Hz) are the observed and emitted frequencies, the speed of sound is c= 340 m/s,
V(m/s) is the velocity, ti(s) is the emission time vector, dmi n (m) is the overhead distance between the
aircraft and the array.
5. Results
(a) Landing EMB190 (b) Take-off EMB190
(c) Landing B737-700 (d) Take-off B737-800
Figure 4: Overall Sound Pressure level values predicted by Doc.29 compared with the levels measured
by NMT and Munisense for arriving and departing flights of types EMB190 and B737-NG.
ICSV30, Annual Congress of International Institute of Acoustics and Vibration (IIAV), 8-11 July 2024

This section presents the Doc.29. prediction obtained from the given methodology. Along with the
predicted noise levels, the curves seen in Figure 4 exhibit the noise levels measured over a single flyover
by the NMT and Munisense. These graphs are representative of all the measurements taken, except for
the outlier corresponding to] the take-off of EMB190, where an under prediction of more than 10dBA of
both LA,max and SEL values is observed. The modelled values for the rest of the cases when compared
to the measurements, provide satisfactory agreement. In cases of B737-700, -800 flyovers, LA,max and
SEL values are predicted with differences of less than 3 dBA with the levels recorded by both NMT and
Munisense. To investigate the agreement of the predictions of SEL values with the noise levels measured
by the NMT and Munisense individually, Figure 5 is presented. A good agreement between the modelled
and measured SEL by the NMT is exhibited. However, the noise levels measured by Munisense are
consistently higher, especially in cases where OASP L is in the range of 60-75 dBA. This could be
attributed to the increased effect of interference by the ground. This higher measured noise level at the
Munisense microphone is most probably due to its lower setup altitude.
Figure 5: Modelled and measured S EL with NMT (left) and Munisense (right) of arriving (red) and
departing (black) flights where the blue dashed line represents correlation = 1.
6. Conclusions
This paper presents a robust methodology for predicting accurate values of LA,max and SEL with
Doc.29. The distance input is obtained from the ADS-B data, and engine settings are obtained from
the array. Although the data from NMT could be used for validation purposes, the audio they recorded
cannot be used to retrieve information concerning the engine settings from the spectrograms due to their
sample frequency of 8 kHz and the presence of noise in the background. Hence, this demands that
alternate measurement setups, such as Munisense, be made use of. Even though the measurements made
by Munisense exhibit higher noise levels than those of the NMT in this paper, this effect can be reduced
by setting it at a higher altitude for future measurements, which would reduce the ground effect. Due
to the higher sample frequency than the NMT at 48 kHz, the spectrograms from Munisense make it
more applicable to retrieve N1%. The thrust settings obtained from the array data in this paper could be
used to validate the ones that are obtained from Munisense for the given aircraft types. Additionally, the
convenience of transporting Munisense and its ability to record noise levels below LA= 60 dBA makes
it an appropriate choice for taking measurements at low-noise regions.
ICSV30, Annual Congress of International Institute of Acoustics and Vibration (IIAV), 8-11 July 2024
7

7. Acknowledgement
The authors are grateful to the EC for supporting the present work, performed within the NEEDED
project, funded by the European Union’s Horizon Europe research and innovation programme under grant
agreement no. 101095754 (NEEDED). This publication solely reflects the authors’ view and neither the
European Union, nor the funding Agency can be held responsible for the information it contains.
References
[1] AIRBUS. “Global Market Forecast 2023”. Toulouse, France, June 13, 2023. URL:https : / /
www .airbus . com/ en/ products - services/ commercial- aircraft / market/
global-market-forecast.
[2] Fedja Netjasov. “Contemporary measures for noise reduction in airport surroundings”. In: Applied
Acoustics 73 (10 Oct. 2012), pp. 1076–1085. ISSN: 0003682X. D OI:10.1016/j.apacoust.
2012.03.010.
[3] ECAC.CEAC Doc 29. Report on Standard Method of Computing Noise Contours around Civil Air-
ports Volume 2: Technical Guide. Neuilly-sur-Seine Cédex, France: European Civil Aviation Con-
ference, Dec. 7, 2016.
[4] Environmental noise guidelines for the European Region. OCLC: 1129816077. Copenhagen, Den-
mark: World Health Organization, Regional Office for Europe, 2018. 61-76. ISBN: 978-92-890-
5356-3.
[5] S.J. Heblij, J. Derei, and R.H. Hogenhuis. Toepassing ECAC Doc29 voor het bepalen van de gelu-
idbelasting van het vliegverkeer van Schiphol. Feb. 2019.
[6] Dick G. Simons et al. “Comparative assessment of measured and modelled aircraft noise around
Amsterdam Airport Schiphol”. In: Transportation Research Part D: Transport and Environment
105 (2022), p. 103216. ISS N: 1361-9209. D OI:https : / / doi . org / 10 . 1016 / j . trd .
2022.103216.URL:https://www.sciencedirect.com/science/article/pii/
S1361920922000463.
[7] Matthias Schäfer et al. “Bringing up OpenSky: A large-scale ADS-B sensor network for research”.
In: Apr. 2014, pp. 83–94. DOI:10.1109/IPSN.2014.6846743.
[8] Roberto Merino-Martínez et al. “Improving Aircraft Noise Predictions Considering Fan Rotational
Speed”. In: Journal of Aircraft 56.1 (Jan. 2019), pp. 284–294. ISS N: 0021-8669, 1533-3868. DOI:
10. 2514/1.C034849.URL:https://arc.aiaa.org/doi /10. 2514/1.C034849
(visited on 06/13/2023).
ICSV30, Annual Congress of International Institute of Acoustics and Vibration (IIAV), 8-11 July 2024