Models, Measurements and Tools for Aircraft Noise

Models, Measurements and Tools for Aircraft Noise Antonio Filippone 1 & Adrian Harwood 2 School of MACE The University of Manchester Manchester, M13 9PL Abstract This contribution addresses recent advances in the prediction of aircraft noise, including the development of new numerical models and computational tools, supported by an increasing database of noise measurements at commercial airfields. The models and tools described in this contribution are aimed at the optimisation of operational procedures, the improvement in understanding in the causes of noise, the determination of noise maps at realistic airfields, and hence the reduction of noise annoyance from commercial aviation. Examples of comparisons between predictions and field measurements are given as well as noise footprints at London Heathrow. Effects of trajectory displacements are shown for take-off and landings. Airplanes considered include the Airbus A320-211 and the Boeing B747-400. Gaps in knowledge of low-order methods for aircraft noise are discussed. 1. INTRODUCTION There is strong interest in the use of operational changes of commercial aircraft to reduce noise and to improve airport capacity, subject to minimum disruption to air traffic control and minimum overall environmental emissions. It is thought that operational changes can be implemented with a relatively limited cost, particularly at times when commercial aviation is expanding and adding new airplanes that are to be in service for several decades into the future. There are studies that address step-changes in airplane technology (conceptual designs) [1], but these efforts are unlikely to bring relief in the short term. Thus, the problem is to establish a noise prediction method that is sufficiently comprehensive and sufficiently accurate to guide decision-makers toward new flight procedures. As a result, a number of computerised noise prediction models have been proposed in recent years, although progress in applying new developments has been slow. There is a number of reasons for this. Aside from the complexity of the task, predicting aircraft noise is not a well-posed problem, since many internal and external parameters are unknown, unspecified or stochastic. This is in contrast with fully controlled laboratory experiments. Therefore, there is a problem in terms of validation and verification. At the opposite end of the simulation spectrum there are models that allow the detailed simulation of the acoustic sources, but these models rely on assumptions related to local geometry and operational conditions of the source, and hence, are often impractical. The work discussed in this paper is primarily concerned with practical applications of computer methods for noise prediction. One approach to aircraft noise prediction is through the use of database systems that are coupled with appropriate ancillary models for noise propagation (noise-power-distance), and some flight mechanics. These computer models are not entirely predictive, but are nonetheless routinely used for noise mapping at airports. Examples in this category include INM [2] and ANCON [3], the latter one used in the UK. An alternative approach has emerged in recent years, based on theoretical modelling with experimental data used when possible to identify the main sources of noise, the interference effects and the long range propagation. One important example is the development of ANOPP2 in the USA [4]. 1 School of Mechanical, Aerospace, Civil Engineering. Email: a.filippone@manchester.ac.uk (correspondent). 2 School of Mechanical, Aerospace, Civil Engineering. Email: adrian.harwood@manchester.ac.uk 1

The computer model developed for the prediction aircraft noise in this work (FLIGHT) is briefly discussed in this section. The reader is invited to refer to the published literature where several features of this model have been documented, for example [5-10]. In these references, additional literature is cited with regards to numerical methods developed and applied. In short, this is a comprehensive model that contains sub-modules for the reconstruction of the aircraft geometry, modelling of aircraft performance characteristics and also the impact of these parameters on generated noise. This software is fully documented in the published literature (see references). The approach proposed in this paper focuses on the use of low-order methods in FLIGHT. There are numerous reasons why low-order models are both necessary and appropriate. For example, during initial design when very few parameters are established and the engineer needs to have a preliminary assessment of how the system will respond from an acoustic point of view. 2. NOISE MODELS In this section we discuss advancements in numerical methods in several interrelated areas, including airframe noise, propulsion noise, noise propagation physics, system modelling, and operational optimisation. At present, in FLIGHT, we have 22 commercial airplane models (mixed turbofan and turboprop), some of which include sub-versions (nominal, extended range), and different engine options (example: CFM56 and IA V2527 for the Airbus A320-200). The noise measurement database used for verification includes 18 different airplane types and flight-data-recorder (FDR) data for two different airplanes. 2.1 Simulation Framework The noise simulation in FLIGHT is based on a network of sub-models that are connected through a higher-level driver. This strategy allows the replacement of a sub-model with another one without having to introduce major changes into the software. For example, two different jet noise models are available, as well as three noise propagation models; each of these models can be selected by the user. Figure 1: Network model of the aircraft noise simulation tool FLIGHT. Figure 1 shows the network model. Each node represents a numerical model that can be replaced if an alternative is available. Also, it represents a local point in the overall validation strategy that starts at the component level and proceeds to the level of the aircraft integration. Each network model depends on a number of parameters or assumptions (indicated by dashed lines radiating from a node); the results inevitably depend on these assumptions. For example, in absence of detailed 2

data on the fan/duct acoustic liners, default data are assumed, which may not represent the actual airplane model. The combination of noise sources in the frequency and time domain is passed to another key node in the network: the long-range noise propagation model. This model takes into account the effects of stratified atmospheres, the presence of the ground and the actual ground characteristics. There are a number of noise shielding effects; these depend on the aircraft configuration. Important effects are: jet-by-jet shielding; engine/propeller noise shielded by the fuselage or the wing; engine noise shielded/diffracted by the wing s shear layer (or wake). These effects depend on the line of sight of the receiver and various other geometrical conditions. The use of an acoustic liner is a form of noise shielding. 2.2 Flight Mechanics Airplane trajectories are generated with the flight mechanics module. In brief, the module generates trajectories in the vertical plane for both arrival and departure and includes key aspects such a standard noise abatement departure profile (NADP) trajectories, steep approaches, displaced landings, etc. Alternatively, it is possible to use airplane-generated trajectories via pre-processing of FDR data packs. In this case, the global positioning system (GPS) coordinates are parsed and converted to a local reference system on the ground. The interface between the flight mechanics and the noise model consists of a time-dependent matrix containing over 20 parameters that define the airplane position, air-speed, configuration, engine state and atmospheric conditions at each point in the flight trajectory. A typical trajectory lasts 2 to 3 minutes, with a variable time-step of the order of 0.5 to 1.0 seconds. Changes in airplane configuration (high-lift system deployment and retraction; landing gear deployment and retraction) are considered instantaneous. The turbofan and turboprop airplanes are treated with separate methods. In the latter case, the airplane is trimmed to either required thrust or required power; corresponding thrust and power coefficients are determined for the propellers; the propellers are then trimmed to these specified conditions; net thrust or shaft power are calculated; the gas turbine engine is solved in inverse mode to determine the corresponding engine state, in particular the fuel flow, the temperatures and pressures across the engine [8-9]. 2.3 External Conditions External conditions incorporated into the prediction methods include the main atmospheric properties (air temperature, pressure, relative humidity, wind speed, wind direction, air turbulence level). Wind, turbulence and humidity are considered constant throughout the airfield. As mentioned, the flight trajectories are simulated, and may not correspond to the actual flight. However, if FDR-based trajectories are used, they must be synchronised with the noise measurements and with the airfield coordinates. Some discrepancies may arise, due to the sampling rate of the FDR. An example is shown in Figure 2, where the airplane appears to follow a jagged, rather than smooth, track on landing at Manchester airport. 3

Figure 2: Flight trajectories from Flight-Data Recorder on arrival at Manchester airport (ICAO: EGCC). 2.4 Outputs The output of the noise model includes instantaneous, integral noise metrics and some relevant extrema. In the first category, we have the sound pressure level (SPL) for all the noise sources, 1, :, in the frequency domain (limited to 1/3 octave band frequencies). Among the integral metrics, FLIGHT returns the effective perceived noise level (EPNL), the sound exposure level (SEL), the time-audible (TAUD), and the equivalent continuous noise level (LAeqT). Among the extreme value, the results include the maximum SPL, the A-corrected maximum SPL (LAmax), and the maximum perceived noise level (tone-corrected) PNLTM. The code issues detailed noise reports, as well as a noise breakdown over the full trajectory (see 5). All the output files contain metadata that allow a fair comparison between simulations. 3. NOISE MEASUREMENTS Aircraft noise measurement is an important activity for the validation of numerical prediction models. Furthermore, accurate measurements of different aircraft types form a database of information for tracking trends. However, measurement of aircraft noise footprints over large areas is often an impractical pursuit, requiring a large number of distributed microphones vulnerable to theft, damage or tamper. Typically, measurements are made using a single microphone positioned either under or to one side of the flight path, depending on data required (ISO or ICAO standards). 3.1 Flyover Noise Measurements Noise data from community microphones at major airports are not considered reliable for comparisons with numerical predictions. Most of these measurements are taken automatically with systems that are not monitored and are affected by inaccuracies, background noise and random events. Specific flyover noise measurements [10] are used for the validation of near-range predictions and represent crucial phases of flight where noise is often the paramount consideration of residents. To perform such measurements a data logger is attached to a microphone mounted horizontally on a tripod at a distance 1-2 m above the ground. Logging is started when an aircraft is seen to approach. Alternatively, a trigger level may be set such that a noise level above a set threshold automatically initiates logging. Instantaneous broadband and 1/3 octave band noise data, such as instantaneous unweighted and A- weighted sound pressure level (SPL), are recorded at a logging rate of 10Hz. Integral noise metrics such as SEL are computed throughout signal capture as well as time-averaged, A-weighted SPL (LAS) using an integration period of 1s. EPNL is computed from the data using the FLIGHT program as it cannot be computed by the sound level meter directly. Background noise levels are recorded 4

periodically by logging data for several minutes in the absence of aircraft overhead. Logging is terminated once the aircraft has receded such that the local SPL has dropped to background levels. 3.2 Measurement Campaigns A recent flyover noise measurement campaign was undertaken at Manchester (ICAO:EGCC), where data were gathered for 18 different aircraft types. Three aircraft in particular provided numerous measurements of uncontaminated data: B737-800, A320-200 and the ATR72-200/-600. Results are presented for these aircraft types here. Throughout the campaign, a number of different measurement locations were used. These locations were chosen based on ease of access, centreline position and threshold distance and are illustrated in Figure 3. Figure 3: Microphone locations for flyover measurement campaign. Location 2 (red) is for departure, Locations 1,3,4 (green) are for arrival. 3.3 Measurements Results Landing and take-off measurements for the aircraft mentioned above are shown in Figure 4a-c. The associated measurement location is stated in the legend and appropriate background levels are indicated by the solid lines. Figure 4a: A320-200 arrival (left) and departure (right) measurements Figure 4b: ATR7 arrival (left) and departure (right) measurements 5

Figure 4c: B737-800 arrival (left) and departure (right) measurements Departure measurements exhibit a much larger spread than arrival measurements. This is attributed to variation in aircraft configuration and operational procedures during a take-off roll such as brakerelease gross weight and climb angle, engine and flap settings which may vary between operators. Arrival measurements are much more predictable with good coalescence of peak noise levels. In general, variations in noise level on approach are attributed to throttle and control surface adjustments during the final minute of final approach. In the case of the former, supporting evidence can be observed by listening to the recorded signals back through headphones. In Figure 4c a large peak can be seen between 20 and 30 seconds after the aircraft has passed for measurement 018. When heard back through headphones, this noise is characteristic of collapsing vortical flow structures [11] and is attributed to a particular strong tip vortex. 4. AIRCRAFT NOISE PREDICTIONS FLIGHT may be configured using known input data from the noise measurement campaign in order to predict flyover noise of aircraft. Examples are shown here for arrival/landing. Figure 5 shows the aircraft noise prediction and its comparison with the mean measured data at Manchester airport for an Airbus A320-211 powered by CFM56 turbofan engines. The measured data plotted are to be interpreted as mean data; the shaded area denotes one standard deviation above/below the mean value. It is noted how the comparison is close to the reference data, although in this instance the prediction fails to match the noise peak. For a reliable comparison, the time window should be 60 seconds (as shown), and the minimum noise level at least 40 db. Figure 5: Comparison between predictions and measurements for an Airbus A320-211 with CFM56 turbofan engines. 6

5. NOISE TOOLS FLIGHT can be coupled with a number of expanding numerical tools to adjust the prediction to specific airfields. Some of these tools are briefly discussed below. 5.1 Airfield Modelling The noise prediction models discussed in the previous sections can be used jointly with a realistic model of the airfield. The tools available to the software include: 1) determination of airfield maps; 2) association of ground impedance values at each ground receiver; 3) algorithms to estimate the ground reflection point of acoustic waves; 4) synchronisation of the airfield model with the simulated or measured flight trajectory; 5) algorithms for the determination of the travel path of reflected acoustic waves to the receiver through minor obstacles (trees, shrubs, low-level houses) and the corresponding acoustic effects; 6) partition of sub-zones for detailed noise calculation; 7) an algorithm to calculate the difference between noise footprints (e.g. calculated with two different noise propagation methods), swap travel direction, redefine grid resolution; 8) restart options (to be used in computationally demanding cases when job duration is limited). An example of airfield modelling using this tool is shown in Figure 6. This result refers to London Heathrow (ICAO: EGLL). The colours at the airfield indicate ground impedance value. A synchronised arrival trajectory is shown, with landing from the East on runway 09L/27R. The top plane shows an instantaneous noise map with iso-noise levels (the centre of gravity of the aircraft is represented by a large black dot). Figure 7 shows a detailed noise map over the region surrounding the landing point. Figure 6: Airfield model of London Heathrow (LHR), with a synchronised trajectory (Boeing B747-400) on arrival, and instantaneous noise levels (top plane). 7

Figure 7: Simulated noise map of a Boeing B747-400 landing on runway 09L/27R at London Heathrow (case of Figure 6). Another example of simulation is shown in Figure 8, where we display the effects of extended runways. In this example, the length of the runway has been doubled to 6,000 m, take-off and landings are displaced by 3,000 m. Figure 8: Effects of displaced take-off and landings on noise maps on a single extended runway. Data plotted are EPNL. 5.2 Stochastic Effects As anticipated, the aircraft itself and the external environment are rarely fully specified, or even known with sufficient precision. In order to establish the role of these uncertainties, an additional tool is included in the noise prediction program to deal with uncertain environments. Multidimensional response surfaces are generated, some with random changes of key parameters, before producing a set of noise predictions which are then analysed with statistical methods. Stochastic changes with pre-defined margins are introduced to about 50 system parameters in six categories: airframe geometry, engine geometry, engine s operational conditions, atmospheric parameters, ground parameters, airplane position with respect to a fixed ground receiver. Results may therefore be expressed within a particular confidence interval. 5.3 Trajectory Optimisation The flight trajectory can be parametrised in order to seek environmental optima, such as minimum noise level subject to minimum landing and take-off (LTO) emissions. Mathematical details of this approach are presented in a separate study. In brief, the flight trajectories are made to depend on a limited number of key parameters; an object function is selected to minimise emissions; and an external optimiser (neural network or non-dominated sorting genetic algorithm, NSGA) is coupled via an interface to generate new solution vectors. Optimal solutions are then evaluated from a Pareto frontier. Among the free parameters on arrival, this solution gives us the optimal point of landing gear deployment, the delay in the flap deployment, the sequence of flap deployments, the glide slope of the trajectory and the touch-down point. 8

6. CRITICAL DISCUSSION Unlike other tools for engineering simulation, aircraft noise programs are still at their infancy; their accuracy and consistency is still below expectations. However, there is a fundamental aspect that must be clarified: the simulation platform (e.g. the airplane and the external conditions) are not fully specified; the quality of the noise measurements is not at a level of consistency as one would have in fully controlled (and repeatable) laboratory conditions. Therefore, we face the difficulty of matching noise measurements (themselves affected by stochastic effects), with imprecise knowledge of the airplane system. The latter problem forces us to use low-order simulation models. In this instance, low-order methods must rely on a limited number of parameters. They offer some degree of confidence within the limits of their scope, but fail outside recognised bounds. There cannot be generality in methods that rely on some form of empiricism. The analysis of aircraft noise relies to a great extent on a systematic approach. In this framework, we have the opportunity to study multiple effects and their magnitude; hence, we can direct our efforts toward those airplane components that contribute most to the overall noise signature. Validation of aircraft noise prediction models is difficult due to the impracticalities of collecting reliable aircraft noise measurements in a real-world setting. The distribution of large unmanned microphone arrays requires permission to be granted for their placement and retrieval, with the possibility of locations being less than ideal, especially if the airport is situated in urbanised environments. These difficulties, amongst others, are the main reason suitable quantities of this kind of data simply do not exist. Noise footprints are a useful tool for quantifying noise impact over large areas, but validation of mid- to long-range prediction using microphone data is unlikely. Whilst speed of calculation may be a concern, it is believed that accuracy of predictions is of primary importance. In the past there has been too much emphasis on the speed of calculation, which has probably shifted the efforts toward less robust models. It is recognised that speed of calculation becomes a major challenge when considering noise footprints and multiple airplane movements and hence for methods to be considered practical there must be balance between computational effort and accuracy. There is a gap in knowledge in the area between accurate field methods (CFD and aero-acoustics) and empirical methods. This middle area, is the place for low-order methods, but must be developed in order to improve predictions for the applications discussed in this paper. Various technology forecasts indicate that a large number of current-generation airplanes will be replaced; strong growth is predicted within 20 years from today, with growth coming from more advanced technologies, that we are presently unable to model. 7. CONCLUSIONS Progress has been achieved in the numerical prediction of aircraft noise at a practical level. A software tool has been developed, and continues to be improved, for application to the current generation of commercial aircraft powered by gas turbine engines. Several features have been implemented and briefly described in this paper. Validation and verification is carried out at several levels, starting from the component level, ultimately leading to a comparison of simulated flyover trajectories with noise measurements at the airfield. Multiple noise metrics are produced (instantaneous, integral, peak values) for a full evaluation of the effects of community noise. At present, only known configurations can be modelled. Further work is required in a number of areas to allow insight into the contribution of alternative configurations and novel geometries. 9

There is still a lack of comparison between the different aircraft simulation methods developed in recent years (with very few exceptions); it is likely that the difference in predictions are important. This is deemed as an important area of research in the coming years. Acknowledgments The Author wishes to acknowledge the contribution of the EU CleanSky Systems for Green Operations, Grant Agreement 255750, to the development of this software tool. Further information of the FLIGHT computer program is available online: www.flight.mace.manchester.ac.uk. References 1. Hileman J, Spakovszky Z, Drela M, Sargeant, M. Airframe Design for "Silent Aircraft", 45th AIAA Aerospace Sciences Meeting and Exhibit, AIAA 2007-453, Reno NV, January 2007. 2. Boeker E, Dinges E, He B, Fleming G, Roof C, Gerbi P, Rapoza A, Hemann J. Integrated Noise Model, Version 7.0, FAA-AEE-08-01, FAA, January 2008. 3. Ollerhead J, Rhodes D, Viinikainen M, Monkman D, Woodley A. The UK Civil Aircraft Noise Contour Model ANCON: Improvements in Version 2, R&D 9842, ERDC Dept. CAA, Gatwick Airport. 1999. 4. Lopes L and Burley C. Design of the Next Generation Aircraft Noise Prediction Program ANOPP2, 17th AIAA/CEAS Aeroacoustics Conference, AIAA 2011-2854, Portland OR, June 2011. 5. Filippone A. Cruise Altitude Flexibility of Jet Transport Aircraft. Aerospace Science & Technology, 14, pp. 283-294, 2010. DOI:10.1016/j.ast.2010.01.003. 6. Filippone A. Aircraft Noise Prediction. Progress in Aerospace Sciences, 68, pp. 27-63, March 2014. DOI: 10.1016/j.paerosci.2014.02.001 7. Filippone A. Advanced Aircraft Flight Performance, Cambridge Univ. Press, 2012. 8. Filippone A. & Mohamed-Kassim Z. Multi-Disciplinary Simulation of Propeller-Turboprop Aircraft Flight. Aeronautical J., 116(1184), October 2012. 9. Filippone, A. Theoretical Framework for the Simulation of Transport Aircraft Flight. J. Aircraft, 47 (5) pp. 1679-1696, September 2010. DOI: 10.2514/1.C000252. 10. Filippone A. & Harwood A. Flyover Noise Measurements and Predictions of Commercial Airplanes, Journal of Aircraft, 53(2), pp. 396-405, 2016. DOI: 10.2514/1.C033370 11. Alix D, Simich P, Wassaf H, Wang F. Acoustic Characterization of Wake Vortices In-Ground-Effect, 43rd AIAA Aerospace Sciences Meeting and Exhibit, AIAA 2005-260, Reno NV, January 2005. 10