Final Report Summary - ARMONEA (Anotec Real-time MOdel for Noise Exposure of Aircraft)
Executive Summary:
The System for Green Operations research consortium of Clean Sky aims to demonstrate substantial reductions of environmental impacts in civil commercial mainline, regional aircraft and business jet domains. The Management of Trajectory and Mission (MTM) branch of the Systems for Green Operations research consortium aims at developing technologies to reduce chemical emissions (CO2 and NOx) and Noise. One of the main fields of research considered by MTM to reach these objectives is to optimize in-flight 4D trajectories, including the overall missions’ profiles, through mathematical optimisation. Once an optimum trajectory will be found, it will be evaluated against current state of the art route. Simulations will be performed with emissions and noise models to assess the improvement of environmental performance achieved by the trajectory of the aircraft. Since the technologies and systems developed for trajectory and mission optimisation need to be inserted in the overall economic models of the airlines, which influence these operators’ choices, the operational “cost” of trajectory will also be assessed.
Implementation of these optimisations is foreseen either on-board, in an avionics computer, or on ground, using computing tools in a laboratory or in an airline operations centre. The activities of MTM will bring implementation prototypes of these technologies to avionics systems demonstration platforms.
Some of the technologies studied in the MTM branch of Clean Sky intend to implement and experiment aircraft trajectory optimisers aboard the aircraft, in an avionics computer. ARMONEA looks after an 'environmental cost' noise model to be used with these trajectory optimisers.
The main objective of the ARMONEA project was to provide a detailed specification and design for a new level of aircraft noise prediction modelling, capable of being used in real-time applications. This model can thus be used for the minimisation of the noise impact “on-the-fly” as part of a wider optimisation for minimum environmental impact of aircraft operations, as envisaged in SGO/MTM.
Starting from the complex SOPRANO aircraft noise prediction suite a series of simplifications was assessed and the most promising options were used in the design.
The resulting ARMONEA model was implemented in a software package and assessed against a reference model. It was demonstrated that the model designed in ARMONEA is in full compliance with the requirements.
Project Context and Objectives:
Background:
The System for Green Operations research consortium of Clean Sky aims to demonstrate substantial reductions of environmental impacts in civil commercial mainline, regional aircraft and business jet domains. The Management of Trajectory and Mission (MTM) branch of the Systems for Green Operations research consortium aims at developing technologies to reduce chemical emissions (CO2 and NOx) and Noise. One of the main fields of research considered by MTM to reach these objectives is to optimize in-flight 4D trajectories, including the overall missions’ profiles, through mathematical optimisation. Once an optimum trajectory will be found, it will be evaluated against current state of the art route. Simulations will be performed with emissions and noise models to assess the improvement of environmental performance achieved by the trajectory of the aircraft. Since the technologies and systems developed for trajectory and mission optimisation need to be inserted in the overall economic models of the airlines, which influence these operators’ choices, the operational “cost” of trajectory will also be assessed.
Implementation of these optimisations is foreseen either on-board, in an avionics computer, or on ground, using computing tools in a laboratory or in an airline operations centre. The activities of MTM will bring implementation prototypes of these technologies to avionics systems demonstration platforms.
Some of the technologies studied in the MTM branch of Clean Sky intend to implement and experiment aircraft trajectory optimisers aboard the aircraft, in an avionics computer. This topic looks after an 'environmental cost' noise model to be used with these trajectory optimisers as depicted in Figure 1.
It can be deduced from this schematic that the simplified noise model will be in the inner loop of a computational intensive optimisation process and therefore that the need cannot be fulfilled by an existing complex noise model with a reasonable time delay. This model will be used by a prototype of an on-board system – such as a flight management system – to compute the best 4D trajectory according to the following contributing 'costs':
• Amount of carbon dioxide produced,
• Amount of nitrogen oxides produced,
• Fuel consumption
• Time
• Amount of noise perceived by the population surrounding an airport
ARMONEA provides the model required to support the latter point.
Aircraft noise modelling:
At present several types of aircraft noise models exist, ranging from relatively simple to very complex, each of which have their domains of application.
For noise zoning and strategic noise mapping mostly so-called integrated methods are used. These methods usually are based on Noise-Power-Distance (NPD) relationships for a variety of aircraft types, stored in a database, in which source noise and propagation are combined. The contribution of each flightpath segment to the total noise at an observer on the ground is determined by means of interpolation in this database at the relevant distance and powersetting. This methodology (as described in a.o. ECAC Doc 29 3rd ed) has now been globally harmonised and its software implementation in programs like INM is generally accepted as a suitable way of determining noise maps around airports and are therefore used worldwide for mapping purposes. These methods are computationally light and need only a limited amount of input, mainly on air traffic and flight tracks. Since they are usually used for long term calculations (typically one year), the errors introduced due to the simplifications made tend to compensate eachother and a very reasonable accuracy in the final noise contours can be expected.
However, the methods described above are not very suitable to accurately predict the noise of single events (in fact this has never been their purpose). Especially when attempting to optimise a flight procedure for minimum noise impact these methods fall short, both because of the way they calculate the noise as well as due to the lack of sufficient data in the NPD (especially multi-configuration data for aircraft states or separate airframe noise data). To this end more complex source-based models are available. These methods usually perform a calculation of the various engine- and airframe noise sources and relevant installation effects, at many discrete points of the flight trajectory, after which the resulting noise is propagated to the ground in a more or less sophisticated manner. These methods allow for an accurate prediction of the noise received on the ground for a variety of flight procedures and engine- and aircraft states. However, they usually need very detailed input data and they are quite computational intensive. In general these are proprietary models of the main aircraft- and engine manufacturers.
ARMONEA provides a new, until now non-existing, intermediate level of noise model, combining the speed and reduced data requirements of simplified models with the accuracy of the complex software packages, thus enabling to target real-time applications such as those envisaged in SGO-MTM.
Project Results:
1. Introduction
The main objective of the ARMONEA project was to provide a detailed specification and design for a new level of aircraft noise prediction modelling, capable of being used in real-time applications. This model can thus be used for the minimisation of the noise impact “on-the-fly” as part of a wider optimisation for minimum environmental impact of aircraft operations, as envisaged in SGO/MTM.
The ARMONEA model was implemented in a software package and assessed against a reference model in order to demonstrate compliance of the design with the requirements.
The ARMONEA project consisted of 3 main parts:
Requirements and specifications
• Definition of the requirements for the model to be designed
• Elaboration of the specifications for the model to be designed
Design and development
• Elaboration of the detailed design of the simplified noise model
• Development of software package
Verification and validation
• Definition of a test plan
• Elaboration of a reference dataset
• Verification and acceptance tests of software package
2. Requirements and specifications
At the start of the project the following list of requirements was elaborated.
2.1 Model capabilities
The following needs are dealing with the main functions of the model:
- The model shall take into account each relevant noise source (airframe and engine noise at least) for noise computation.
- The model shall assess noise impact on discrete microphone locations around an airport. The model shall be able to manage from 1 to, at least, 4 microphones.
- The positions of microphones shall be configurable
- Outputs: The model shall compute the effect of airframe and engine noise perceived (spreading across the audible spectrum) on the ground. This effect has to be quantified according to standard noise metrics
- Inputs: The model shall take as input all necessary and available data:
• Aircraft trajectory as a function of time: aircraft flight phase (take-off/climb, descent/approach/landing, below 15000 ft AGL), aircraft position (latitude, longitude, altitude), aircraft attitude (roll, pitch, yaw), aircraft speed (true airspeed, ground speed)
• Aircraft configuration as a function of time: landing gear and flaps configuration.
• Aircraft engine state as a function of time: thrust level, fan speed
• Atmospheric conditions around the airport (according to position): static air temperature, pressure, wind, relative humidity, … The feasibility of the use of wind and temperature profiles shall be assessed
• Observer characteristics (height above ground, ground acoustic properties)
- Required data, their format, resolutions, time samples or volumes, and the way they are provided to the model (interface), are to be detailed.
2.2 Model setting
The following needs are focused on the parameterization of the model:
- The model shall allow the user to define the following initialization parameters:
• Choice of aircraft and engine types
• Choice of airport configuration (number of microphones, positions of microphones...)
- The model shall allow the user to define the required output metrics
- Setting definition can be performed via loaded data files. The format and data shall be detailed.
2.3 Future utilization and maintenance
The following needs are dealing with the potential evolutions, adaptation and maintenance of the model:
- It shall be possible to adapt, optimize, and produce updated versions (via algorithm and model description documents)
- The detailed model specification should be provided in a way to allow reimplementation in any high level language. Consequently the model itself should not be tied to any computer language specific paradigm.
- User shall be able to prepare and upload new aircraft database. The model shall be adaptable to different commercial aircrafts types:
• Various airframes types, configurations and geometries,
• Various engine types (turbofan, turboprop, propfan), size, behavior and thrust settings.
2.4 Constraints
The constraints that will impact the development of the model are:
Hardware:
- The model shall be adapted to limited processing power, memory and data storage.
- Equivalent to PC from generation 2007: core 2 Duo (2 cores) E6850 3GHz with 4Gb of RAM, with no computational support from the GPU
- Parallel computing should be avoided whenever possible, since current generation of avionics relies on time sharing operating systems, with strict time allocation task schedulers, on single CPU boards
Software
- The model shall demonstrate predictable convergence and computation time of the algorithm
- Software architecture allowing certification: constraints can be restrained to the algorithm, which for instance, should not (as much as possible) rely on:
• Dynamic allocation (risk of ghost references, “memory leaks”, performance penalty due to allocation within loops)
• Garbage collectors (drop of performance and lack of determinism)
• Implicit declaration
• Recursions
• Parallel computing
- It is noted that these restrictions do not apply to the prototype code.
Performance:
- For the noise computation of one trajectory, the algorithm shall ensure a computation time of less than 20 milliseconds, in the following conditions
• Trajectory is a take-off/climb or approach/landing phase, from 0 to 15000ft (probably consisting of various flightpath segments)
• Noise computation on 1 microphone
- On a current generation PC: Core i7-2600K (4 cores) 3.4GHz with 4Gb of RAM, with no computational support from the GPU
- The target maximum error is in the range of 1dB (reference vs. simplified - i.e. not between prediction and reality) in order to be consistent with gain margins expected in Clean Sky. (Note: Whereas this requirement applies to all noise metrics, instantaneous metrics like LAmax will most likely be the more restrictive ones with respect to this criterion).
Mock-up
- The mock-up shall be delivered as a software package running on Microsoft Windows XP Operating System
- The mock-up shall run on a current generation PC, and share processing resources with other applications
- Validation of the model mock-up will have to be performed for several aircraft types, preferably ATR 42-500 / A320CFM56-5B /A330 CF6.
3. Model design
The strategy selected in ARMONEA was to simplify an existing complex noise prediction suite (SOPRANO) in order to derive a model which complies with all imposed requirements and constraints.
Simplifications have been explored for:
• Source noise predictions
• Propagation effects
• Geometrical calculations
3.1 Simplification of source noise predictions
Various alternatives for simplifying a complex aircraft model were considered potentially viable. Two main options were distinguished:
1. In a pre-processing step (performed ‘off-line’) a set of datafiles can be generated with a sophisticated tool (complex prediction software, measurements, etc.) by determining source noise for a discrete number of datapoints. These datafiles (look-up tables) would then replace the more complex source models, thus requiring a rather light embedded model to handle these files to obtain the required noise data at the observer.
2. Complex noise models require a significant set of input parameters, of which many usually are unknown, especially in the case of real-time application as envisaged here. Reduction of the amount of required input is thus a potential way of simplification. The first pathway is to implement a generic model of an engine (engine deck), which only requires a limited number of input parameters to generate the values of parameters at intermediate stages, required for the source noise models. A second pathway would be to simplify the source noise models themselves, thus requiring less parameters.
After prototyping of the various options and initial testing against the requirements, the following alternatives have been retained for the ARMONEA model.
3.1.1 Look-up tables
Source noise for a fixed distance is stored in a multidimensional lookup table, with the noise spectrum (in 1/3 octave bands) is provided for a range of directivity angles and for one or more independent parameters like power setting, speed, aircraft configuration, etc. (see Figure 2).
These lookup tables may contain any number of axes but usually there will be two cases of use: Engine noise and Airframe Noise (see Figure 3). However, it is noted that both main noise sources may also be combined in a single look-up table, representing the noise of the whole aircraft.
3.1.2 Spherical harmonics
Look-up tables can become quite large if many (engine and/or aircraft) conditions are included. Therefore, apart from the abovementioned look-up tables, ARMONEA also includes an elegant way of reducing the source noise description, by means of spherical harmonics (see Figure 4). For this method only a set of coefficients has to be stored in the look-up table, thus significantly reducing memory- and disk footprint.
3.2 Propagation effects
Atmospheric Absorption
The model had at least to be capable of calculating atmospheric absorption for a homogeneous atmosphere with user definable atmospheric conditions (temperature, relative humidity, etc), by using one of the well-known existing models, all available in ARMONEA:
- ISO 9613-1
- SAE ARP 866A
- Sutherland ANSI S1.26
Increased accuracy was envisaged when atmospheric absorption is calculated for an atmosphere divided in finite layers of more or less constant properties (temperature and relative humidity). Therefore a layered atmosphere was implemented in ARMONEA, by reading an ASCII file with all relevant information.
Diffraction effects
Usually wind and/or temperature gradients are present in the atmosphere. These gradients may cause diffraction effects which might alter the noise propagation considerably and it was therefore to be investigated if the model would be capable of taking these effects into account. Prediction models based on e.g. ray-tracing are considered the most suitable ones for these cases, but required calculation times are most probably outside the performance requirements sought. Therefore a simplified method, based on a set of look-up tables with diffraction effects for pre-defined atmospheric conditions was considered to be more appropriate for use in ARMONEA. From this set, a specific tool (“ray tracer”) selects the look-up table for the conditions closest to the ones for which the predictions have to be made.
Ground reflection
The interference of direct and reflected sound waves at the observer location will cause dips and peaks in the resulting received spectrum. To calculate this effect the well-known Chien-Soroka method was selected.
Additional effects
In specific situations it might be of interest to correct predicted noise levels for certain additional effects, not taken into account in the model (e.g. significant reflection from a building close to a microphone). To this end an additional function was implemented, based on a look-up table with the correction values.
It is noted that the process by which these correction values are to be determined did not form part of the current project. Most probably a statistical analysis of long-term measurements at the specific site would be necessary to provide the required information.
3.3 Simplified geometrical calculations
It was envisaged that an important reduction in calculation time could be obtained when considering only those points of the flight trajectory which contribute to the noise metric(s) at the observer location(s). Several options were identified for this, which were all to be investigated on their merits and their disadvantages. Finally a method where a first estimate of the 10dB down time (using fixed angles) is combined with a refinement of a simplified noise prediction at intermediate adaptive time steps, was found to be the optimal procedure, taking into account both calculation time and accuracy of calculated noise level.
3.4 Final design
Based on the results of the downselection procedure the design depicted in Figure 5 was found the most promising for the ARMONEA model.
4. Software development
In order to be able to verify and validate that the designed model complies with the requirements and specifications, a mock-up of the model, designed in the former section, was developed.
In order to facilitate future integration and also to maximise performance, the core modules (Geometry, Source noise and Propagation), enclosed by the dashed line in Figure 5, have been implemented as an Application Programming Interface (API) in FORTRAN 95 that can easily be called from programming languages like C/C++, Python, Matlab, VB.net C# and FORTRAN.
The API and main program source code are able to be compiled in at least Intel and GNU (gfortran) compilers. If differences are found in the “state of art” of each compiler, pre-processor directives may be used to ensure compatibility of the source code.
The other elements of the model have been implemented in a console program in FORTRAN 95 to easily debug both main and API. In future applications beyond ARMONEA, it is supposed that these elements will be developed by the developing party.
Look-up tables with source noise for the following aircraft types are provided as part of the software package:
- A320-214/CFM56-5B4P (provided by Airbus)
- A330-201/CF6-80E1A2 (provided by Airbus)
- ATR 42-500 (developed by reverse engineering NPD data)
- Cessna Citation (provided by Anotec, from the FP6 IMAGINE project)
5. Verification and Validation
In order to be able to verify and validate that the designed model complies with the requirements and specifications a V&V process was defined.
5.1 Test Plan
The main purpose of the Test Plan was to describe the manner in which the simplified noise model was to be tested in order to verify and validate compliance with the requirements and specifications described in section 2 above. To this end test cases and procedures were defined. The envelope within which the model will be validated were also defined.
The tests cover all elements of the model and its implementation in a mock-up:
- Component testing
- Integration testing
- Interface testing
- Execution of reference test cases
- Performance checks
- Final acceptance tests
The correct implementation of the algorithm in the mock-up was verified based on the component, integration and interface testing.
In order to verify compliance of the model with the main requirements (accuracy and performance) a series of reference test cases was defined (see Table 1). Although an accuracy requirement in noise level is only imposed for the noise certification positions, some test cases used a grid of points far beyond these positions, in order to provide an indication of the accuracy that can reasonably be expected in the area affected by an actual aircraft operation.
5.2 Reference Model
The reference test cases described above were executed with the so-called Reference Model, in order to provide a reference dataset, to compare the ARMONEA results with.
Although it was originally planned to use an aircraft noise prediction software, already developed in SGO in another project, as the Reference Model, this appeared not feasible.
In the same period Airbus provided detailed source noise tables for 2 of the 3 aircraft types envisaged (A320 and A330) to be tested, which enabled the use of the SOPRANO platform as the Reference Model. Using the same source noise data with both models (SOPRANO and ARMONEA) allowed for a unequivocal verification of the behavior due to the simplifications implemented in the latter.
It was therefore decided to use SOPRANO as the Reference Model.
The reference testcases defined in 5.1 were then executed with SOPRANO, thus yielding the Reference Dataset.
5.3 Verification and acceptance tests of software package
As mentioned above, the strategy to verify and validate the model was to compare the ARMONEA results for the reference testcases with those from the Reference Model. Both models provide sufficient intermediate results to be able to verify correct calculation of relevant variables throughout the calculation process, such as:
- Geometrical parameters (directivity angles, etc.)
- Source noise levels
- Propagation effects (atmospheric absorption, ground reflection, etc.)
- Noise levels received at the microphone positions.
Especially the latter step provided the information required to assess the accuracy obtained with the ARMONEA model. An example of a comparison between the noise levels predicted by ARMONEA and SOPRANO is given in Figure 6.
The reference test cases were also be used to provide the information required on the performance of the ARMONEA model in typical calculations.
Calculation speed was checked by timing the various steps in the process. Memory and disk usage was also checked during program execution.
Based on the tests performed, the following conclusions can be drawn:
The accuracy determined for the various test cases was in the order of 0.1-0.2 dB, well within the 1 dB required.
It has been verified that for a typical 3 mic case the typical memory usage (RAM) is 1.7 MB, far below the maximum of 50 MB allowed.
It has been verified that for the reference test cases the maximum disk space used was 15MB (Gad files), well below the maximum of 1 GB allowed.
It has been verified that for the reference test cases computation time is in the order of 1-10 ms per microphone, well below the maximum of 20 ms allowed.
It can thus be concluded that the ARMONEA model designed in the project fully complies with the requirements and specifications defined at the start of the project.
Potential Impact:
The main objective of the work described here is to provide a detailed specification and design for a new level of aircraft noise prediction modelling, capable of being used in real-time applications. This model can thus be used for the minimisation of the noise impact “on-the-fly” as part of a wider optimisation for minimum environmental impact of aircraft operations, as envisaged in SGO/MTM.
Successful completion of ARMONEA is decisive for enabling future application of the above described development.
In recent years increased effort has been dedicated to better understand the interdependencies between noise and emissions. In Europe initiatives like TEAM-PLAY have been setup in order to create a toolsuite with which these interdependencies can be assessed. However, as with the integrated noise models usually used for airport noise impact studies, these efforts are mainly directed towards long term assessments, usually for policy support, informing the rulemaking process, e.g. for the assessment of stringency options or elaboration of action plans. These methods do not intend to be accurate at single event level. ARMONEA, however, provides a tool with which each individual aircraft operation can be optimised for minimum noise impact in dependence of current air traffic/meteo situations, thus significantly contributing to minimising the environmental impact of air traffic.
Another result of ARMONEA is the availability of a design for a noise prediction model which could be leveraged to other applications and domains (stand-alone or as part of a bigger tool) where on-line noise predictions may be required.
Most of the existing complex aircraft noise prediction models are proprietary information of aerospace manufacturers. Although obviously the manufacturer is usually the one who best knows its own product, there is still a significant part in the whole calculation chain which is not source related, but may influence the end result considerably (e.g. propagation through the atmosphere). It is known that each manufacturer gives its own interpretation to these ‘auxiliar’ parts. In the FP5 project SILENCER, ANOTEC developed the SOPRANO platform with the aim to provide a common aircraft noise prediction suite, providing a harmonised platform for aircraft noise predictions, thus allowing for noise assessments with single, agreed criteria. In addition it is able to handle manufaturers’ proprietary methods, thus ensuring compatibility with their in-house methods, if so required. Over the years SOPRANO has been further developed in a variety of EU projects and is now the noise engine of multi-disciplinary optimisation tools like TERA, used in projects like VITAL and DREAM. For rotorcraft the HELENA software has been derived from SOPRANO, and is now being used in the Clean-Sky GRC and Technology Evaluator as the common helicopter noise model. Since the model proposed in the present project is also based on SOPRANO, ARMONEA will bring the harmonisation of aircraft noise predictions in Europe a step further.
List of Websites:
Contact:
Anotec Consulting SL
Nico van Oosten
nico@anotecc.com
tel +34958620631
