COMPARISON OF METHODOLOGIES FOR CONTINUOUS NOISE MONITORING AND AIRCRAFT DETECTION IN THE VICINITY OF AIRPORTS Alejandro Osses Vecchi Laboratory of Acoustics, Sociedad Acustical S.A., C.P. 7770563, Ñuñoa, Santiago, Chile. e-mail: [email protected]

Max Glisser Donoso, Christian Gerard Büchi Department of Engineering, Gerard Ingeniería Acústica SpA., Control Acústico, C.P. 7770563, Ñuñoa, Santiago, Chile.

Ricardo Guzmán López Directorate of Aerodromes and Aeronautical Services DASA, Directorate General for Civil Aviation DGAC, C.P.9020588, Pudahuel, Santiago, Chile. The aim of the current study is to compare the effectiveness and efficiency between two different procedures for analysing noise levels emitted by commercial aircrafts in the vicinity of airports. The measurements were made in sensitive points near the Arturo Merino Benítez International Airport (SCL), in Santiago of Chile. Both methodologies are applied over the time history of A-Weighted Sound Pressure Levels, but presenting different criteria for detecting aircraft noise events. In the first method, the identification of aircraft noise is carried out applying a moving average to the measured noise levels followed by threshold detection. The second method characterise noise events using the relationship among acoustic descriptors (Leq, Lmax and SEL) in one-minute periods. The results are compared with a high confidence method consisting in the correlation of the measured noise levels with the corresponding Flight Logs.

1.

Introduction

The present study evaluates two methodologies for processing the sound levels measured in a noise monitoring system in the vicinity of airports in absence of non-acoustical information (RADAR, flight logs, aircraft type, etc.). For these ends are presented two automated methods which consider the relationship among acoustic descriptors, the sound changing rates and the use of thresholds for performing the aircraft noise classification. The results for both methodologies are compared with the manual correlation of noise levels with the corresponding flight logs. This research was divided in two stages: (a) Continuous noise monitoring at a residential area, and (b) application and analysis of the methods under evaluation. This last stage considered the quantification of the efficiency and efficacy of the methods, considering the presence of some secondary sound sources as dogs, cars and other daily activities

ICSV18, Rio de Janeiro, Brazil, 10-14 July 2011

1

18th International Congress on Sound and Vibration, Rio de Janeiro, Brazil, 10-14 July 2011

2.

Noise Monitoring Plan

The measurements were performed in the vicinity of the international airport Arturo Merino Benítez (SCL), located in Santiago of Chile. Currently the airport has implemented a Noise monitoring system composed by 3 terminals. The noise monitoring for the current study was performed in the facilities of one of those terminals, the most sensitive point, which is located 11.3 Km to the south of the SCL airport, in the Maipú neighbourhood. At this point it is possible to detect only departure aircraft operations (take-offs). The point is located in a residential area.. The instrument used for the measurements was a Class 1 Sound Level Meter according to the standard ISO 61672-1:20021, configured with a resolution of 1/8th samples per second. The results shown in this paper were measured during three days in February 2011, nevertheless, the instrument was installed as a continuous noise monitoring for almost 2 weeks.

3.

Methodology

In this section the two selected methods for processing the Sound Levels at the noise monitoring terminal site will be explained. 3.1 Method 1 This methodology is based on the typical relationship between the Sound Exposure Level (SEL) and the maximum Sound Pressure Level (Lmax) during a fixed integration time. In the literature it is well established that a sound source could be characterised by using the SEL and Lmax 2,3, particularly considering the difference between those descriptors3. In Table 1 are shown some examples: an explosion, an aircraft flyover and a railway noise. The 3 examples have the same accumulated energy or SEL. Table 1. Comparison of different sounds having equal Sound Exposure Level.

Event Duration [s] Firecracker 0.3 Aircraft Flyover 70.0 Railway Noise 900.0 Event

SEL [dB] 100 100 100

Lmax [dB] 102 90 72

Leq |SEL - Lmax| [dB] [dB] 105 2.0 82 10.0 71 28.0

Depending on the point selected for the aircraft noise evaluation is the value for the typical difference between the SEL and Lmax, nevertheless, usually this value is around 10 dB. This difference in value could be affected by the residual sound at the monitoring point. It could also be affected by the approximation of using fixed integration times, i.e., when computing the SEL, Lmax or even the Leq in one-minute periods. The range in which the difference SEL-Lmax would be interpreted as an aircraft sound event should be determined empirically2,3. In the Van der Heijden’s article2, this range used was set from 7 and 13 dB. The method could be summarised by means of the Eq. (1). x < SEL − Lmax < y

(1)

Where x and y are values in dB and the SEL and Lmax are expressed in dB(A). In addition to this criterion, the SEL has to exceed certain threshold, also in dB(A)3. This methodology was used in Evaluation Impact studies for different airports and aerodromes of Chile: Mataveri Airport (IPC, Easter Island), Carriel Sur Airport (CCP, Concepción) and El Tepual Airport (EPC, Puerto Montt)3.

2

18th International Congress on Sound and Vibration, Rio de Janeiro, Brazil, 10-14 July 2011 3.2 Method 2 This method considers a multiple threshold-triggering for detecting a sound event. For doing this the variability of the acoustic descriptors is evaluated over the time by means of the changing rate between a current value a(t) and the hth previous sample, a(t-h), which is defined in Eq. (2). ∂ a (t ) a (t ) − a (t − h ) . ≈ ∂t h

(2)

Where h is in seconds and a(t) is an acoustic descriptor. For this work, a(t) is the Leq,1s, in dB(A). The events occurring during the first h seconds long period would not be detected. For this reason, it is necessary a previous knowledge of the events under observation4. 3.2.1 The smoothing coefficient In the presence of spurious sound sources (noisiness environment) false events could be detected. For avoiding this situation a digital filter (First-order IIR filter) could be applied, according to Eq. (3). y n = α x n −1 + (1 − α ) y n −1 .

(3)

Where α is a smoothness coefficient between 0 (0%) and 1 (100%), xn is the input and yn is the output of the filter. In Fig. 1 a sound event is shown with a smoothing coefficient α of 0%, 50% and 95%. a(t) [dB(A)]

a(t) [dB(A)] 80

80 α=0 α =0.5

75 70

70

65

65

60

60

55

55

50

50

45

45

40 28

28.5

29 t [sec]

29.5

α=0 α =0.95

75

30

40 28

28.5

29 t [sec]

29.5

30

Figure 1. Aircraft sound event precedes by impulsive sounds (dogs’ barks).

For detecting an event, the smoothing has to be applied to the derivative of the noise levels and the event start and end can be set, defining 3 thresholds: Th, TL1 and TL2, as shown in the example of the Fig. 2. The parameters Th, TL1 and TL2, depends on the acoustic descriptor in use (e.g, dB(Z), dB(A)) and the nature of the event. 3.3 Reference Method: Manual Identification of Aircraft Events This method is used for validating the Method 1 (subsection 3.1) and Method 2 (subsection 3.2). As stated in ISO 20906:20095, the processing of data for an automatic sound-monitoring system in order to classify aircraft sound events, has to fulfil 3 criteria: (a) The uncertainty of the measured cumulated exposure level of all aircraft sound events shall not exceed 3 dB; (b) At least 50 % of true aircraft sound events shall be correctly classified as aircraft sound events; (c) The number of sound events which are incorrectly classified as aircraft sound shall be less than 50 % of the true number of aircraft sound events.

3

18th International Congress on Sound and Vibration, Rio de Janeiro, Brazil, 10-14 July 2011

Figure 2. A sound event being detected by multiple threshold triggering. Picture taken from 4. In this example the thresholds were set to 0.3, -1.2 and -0.5 for Th, TL1 and TL2 respectively.

The first criterion was assumed to be fulfilled since it was used a Class 1 Sound Level Meter, the wind induce noise remained below the 65 dB(A) limit and the weather conditions were favourable (no rain and temperatures from 13° to 31° 6). The percentages in the criteria (b) and (c), consider a total number of aircraft events in function of the manual correlation between the sound levels and the related flight logs. The flight logs were provided by the Chilean Directorate General for Civil Aviation DGAC. The starting and ending for an aircraft event was considered according to ISO 20906:20095.

4.

Results and Analysis

The Methods 1 and 2 are compared with the reference methodology (manual correlation, section 3.3). In Table 2, are presented the results for the reference procedure and the total number of commercial flights (departures) for the selected hours. Table 2. Results for the reference methodology (Flight log correlation). Meas. Nº 1 2 3 4 5 6

Date 2011/02/18 2011/02/18 2011/02/19 2011/02/19 2011/02/20 2011/02/20 Total

Hours 13:00-13:59 23:00-23:59 00:00-00:59 07:00-07:59 15:00-15:59 16:00-16:59

Nº Departures 7 10 1 10 9 9 46

Aircraft 52.2 61.5 50.6 55.9 59.2 55.8 57.4

LAeq dB(A) Residual 59.9 61.1 56.6 50.0 63.6 55.8 59.6

Total 60.6 64.3 57.5 56.9 64.9 58.8 61.7

The parameters selected for the aircraft detection were determined empirically for both methods. In the Method 1, the SEL threshold was set to 75 dB(A), and the difference between the SEL and Lmax in the range from 10 and 13 dB(A). For the Method 2, the parameters Th, TL1 and TL2 were set to 0.50, -0.45, 0.30, respectively, while the derivative variables α and h to 0.95 and 20 seconds. In Table 3 are shown the final results for each method.

4

18th International Congress on Sound and Vibration, Rio de Janeiro, Brazil, 10-14 July 2011 Table 3. Results for methodologies Nº 1 and 2. Meas. Nº 1 2 3 4 5 6

Total Reference Difference

Method 1 LAeq dB(A) Aeronave Residual 59.2 55.0 56.1 63.6 52.0 56.1 52.4 55.0 62.2 61.5 51.5 57.9 57.50 59.54 57.40 59.61 -0.10 0.07

Method 2 LAeq dB(A) Aeronave Residual 57.1 58.1 61.5 61.1 50.8 56.5 56.1 49.5 61.0 62.6 56.1 55.4 58.37 58.89 57.40 59.61 0.97 -0.72

Table 4. Efficiency in the detection of aircraft events for Methods N° 1 and 2 Meas. Nº 1 2 3 4 5 6

Total %

Method 1

Method 2

N° of detected sound events

N° of detected sound events

Aircraft 4 8 1 5 7 3 28 61%

Aircraft 7 9 1 10 8 8 43 93%

Non-aircraft

0 0 3 0 1 0 4 9%

Non-aircraft

1 0 0 1 4 0 6 13%

Considering a total of 46 commercial flights during the 6 hours under observation (Table 2), it is required to detect successfully at least 23 aircrafts (over a 50%) and no more than 23 false aircrafts (less than 50% of false events). As could be seen in Table 4, Method 1 detected correctly 28 aircraft sound event, representing a 61% of the total observed departure operations, while non-acoustical events incorrectly classified as aircrafts represent a 9% (4 times). Both percentages satisfy the requirements of the ISO 20906:20095. In Fig. 3 is presented the event detection using Method 1 for one of the measurements.

Figure 3. Sound event detection using Method 1, for the measurement N° 4 (Sat 02/19/2011, 7:00-7:59). 5 Events are correctly classified and 0 false events are incorrectly classified.

5

18th International Congress on Sound and Vibration, Rio de Janeiro, Brazil, 10-14 July 2011 In the first Method 1 the quantity of detected aircraft sound events is relatively limited (approximately 6 out of 10), nevertheless, the global values for the Equivalent Sound Level Leq 1Hr related to aircrafts are very closed to the results obtained with the reference method, with a difference of 0.1 dB. L Aeq,1s dB(A)

90 80 70 60 50

( ∆ LAeq,1s dB(A) ) / h

07.00

07.10

07.20

07.10

07.20

07.30

07.40

07.50

08.00

07.30

07.40

07.50

08.00

2 1 0 -1 -2 07.00

t [hh.mm]

Figure 4. Sound event detection using Method 2, for the measurement N° 4 (Sat 02/19/2011, 7:007:59). 10 Events are correctly classified and 1 false event is incorrectly classified.

As shown in Table 4, the second method has a better efficiency, detecting correctly 43 events (93%). Meanwhile, the events with an incorrect classification increase slightly to 6 events (13%). The method also fulfils the requirements of the ISO 20906:20095. In spite of the fulfilment with the standard, the difference in the Equivalent Sound Levels respect to the reference method were affected negatively, increasing to almost 1 dB for aircraft sound events and so, the residual sound was underestimated presenting a difference near -0.8 dB. The method is highly robust to impulsive noise sources, as the typical dog barks presented in residential areas.

5.

Future work

For the analysis in the Method 1 (SEL-Lmax) only an integration time of one-minute was considered, assuming that most aircraft operations would be detected within that period, but it is necessary to compare the efficacy of the method using other integration times, e.g., 30 seconds or 2 minutes. The method is very sensitive to the changes in the background noise and so, it is difficult to set the same parameters to correctly detect aircrafts during the morning and during the night. For the second method, presented in [5], proposes the use of multiple triggering for detecting different kind of sounds, but also the use of specific frequency bands (1/1 or 1/3 Octave Bands) for improving the sound detection. In the case of aircrafts, the methodology could be applied simultaneously to the levels in dB(A) and to levels of a low frequency band in dB(Z), e.g. 125 Hz- or 250 Hz-band.

6.

Conclusions

The purpose of the present paper was to evaluate two methodologies for analysing aircraft noise, but without using non-acoustical information (RADAR, flight paths, etc.). It was shown that the self-contained information in the noise levels, in dB(A) was enough to fulfil the requirements established in the ISO 20906:20095, nevertheless, it could be possible to improve the detections using the same methods in low-frequency bands. The first method (based on SEL and Lmax) presented a higher accuracy in the final results than the second method, but is more sensible to the background noise and in some situations their parameters are difficult to be set. 6

18th International Congress on Sound and Vibration, Rio de Janeiro, Brazil, 10-14 July 2011

7.

Acknowledgements

The authors would like to acknowledge the support given by the Chilean Directorate General for Civil Aviation (DGAC), for permitting us the access to the Noise Monitoring Terminal and to the Santiago’s Airport official flight logs. We are also grateful to Richard Wright and Guillaume Goulamhoussen, from Cirrus Research plc, who developed the “Method 2” and spent some time for giving us a lot of details related to the analysed methodology. The research was granted by Sociedad Acustical S.A. and supported also by Control Acústico, making possible the publication of this paper.

REFERENCES 1.

2.

3.

4.

5.

6.

IEC, International Electrotechnical Commission: IEC 61672-1:2002: Electroacoustics – Sound Level Meters – Part 1: Specifications, 2002. J. Van der Heijden, “Recognition and quantification of aircraft noise events inside dwellings”, Internoise 2001, The Hague, The Netherlands, August 27-30 (2001). Glisser, M; Gerard, C. “Results and methodologies of airport noise studies“. J. Acoust. Soc. Am. 128, 2420 (2010). R. Wright, G. Goulamhoussen, “Improvements in Source Identification from unattended sound level measurements using threshold-triggered audio recording”, Internoise 2010, Lisbon, Portugal, June 13-16 (2010). ISO, International Standard ISO 20906:2009(E): Acoustics – Unattended monitoring of Aircraft sound in the vicinity of airports, Geneva, 2009. Meteorology Group, Department of Geophysics, University of Chile (http://infomet.dgf.uchile.cl/)

7