Work package 3 : Integrated prediction model
Objectives
In this work package, the results of work package 1 "Source description" and work package 2 "Reference propagation model" will be combined into an engineering model. This engineering model will be more simple, more easy and faster to handle than the reference model. The most important reason for this is that calculation times with the reference model will be very long and require detailed input data that is not easily available in real situations. The outcome of the model will be the long time averaged noise indicators specified in the European Directive : Lden and Lnight.
Whereas work package 1 deals with individual cars and trains, this work packages must deal with traffic on a statistical basis : the exact number, characteristics and operating conditions of individual cars are rarely known and unpredictable anyway. Whereas the reference model developed in work package 2 deals with point-to-point propagation under specific meteorological conditions, this work package deals with extended sources, climatological description of local meteorological conditions and must also be able to tackle the more complex geometrical situations that occur in urban and suburban areas.
An important part of this work package is the description of sound propagation, in terms of geometrical paths in complex situations, taken into account any combination of reflection, screening by accidentation, buildings, noise barriers,.. under various meteorological conditions.
A second important point is the investigation on statistics for all time varying input data. Statistics have to do with e.g. the variation of traffic conditions and propagation conditions for day and night situations or for summer and winter situations, the uncertainties on road and rail characteristics as a result of poor or correct maintenance,…
The next topic is the accuracy of the model. Whereas the accuracy of the engineering propagation model can be checked by comparison with the reference propagation model, the overall accuracy of the model still depends largely on the available input data. Simplified geometrical description of the site, statistical description of the traffic data, climatological classification of meteorological conditions, lack of accurate information for some data such as ground impedances, all affect the accuracy of the final results. The achievable accuracy must be considered in relation with the affordable complexity of the model and the requirements on calculation time for noise mapping.
The next step is to check the model as a whole by comparison with experimental data provided by work package 4. To check the improvements obtained in the Harmonoise project, calculations for some situations will also be done using the "interim methods" as suggested by the European Directive.
The final task is to prepare test maps and provide these to work package 5 where they will be included in the proposal for a future European standard on "Prediction methods for outdoor noise propagation". These test maps will enable third-party software developers to check the compliance of their products with the new prediction model.
Task 1 : Definition of propagation paths
For the accurate calculation of the noise propagation in complex situation within the limits of affordable calculation time, it is common practice to describe the propagation phenomena by means of appropriate geometrical propagation paths. Within the scope and budget of the of the Harmonoise project, it seems that this approach will be the starting point for our engineering model.
Work package 2 will produce an reference propagation model for the prediction of sound propagation from source to receiver. This means from point to point. For industrial noise, assimilated to point sources, the application of the model is straightforward. But for line sources such as traffic generated noise, this is much more complex: a road or a railway line can be described as a combination of a large number of point sources, a single line source, a segmented line source, a surface source, or even a number of surface sources. Discontinuities in the conditions for propagation, such as an edge of a noise barrier or the presence of a reflecting object or a transition in ground conditions or altitude, may require further subdivision of the source areas.
In open space, only the direct propagation path (including the ground reflection) must be considered. In more complex situations, additional diffracted paths over or around barriers and buildings and reflected paths from such obstacles must be taken into account. Also, any combinations these effects can occur in a real situation. The propagation paths can only be described accurately using three dimensional coordinates.
The definition of the propagation paths must allow the appropriate assessment of ground altitude variations and special constructions (e.g. bridges, tunnels, viaducts). In the case of reflections, the conditions for reflections, depending on the distance and size of the reflecting surface, the meteorology and the number of reflections will be examined.
For line sources, a method must be defined to find all these different calculation paths from different point sources to receiver points. Perhaps also at the receiver point no special point must be considered but a receiver area (for example the complete façade) is of interest. An alternative way is inverse ray-tracing, a common technique used in visualisation programs : rays are emitted in all directions from the receiver (the eye) and catch the (light) sources either directly or after reflections on specular or diffuse surfaces. Highly optimised algorithms use adaptive techniques to speed up such calculations, but this is beyond the scope of our project. Advantages and inconveniences of different methods will be compared.
An additional problem arises when meteorological conditions such as wind and temperature gradients have to be taken into account and curved propagation paths are to be used. One simplification is the so called "linear gradient approximation" which turns straight rays into circles.
This subtask starts with a study of literature about propagation paths. The field of interest for this investigation is larger than pure acoustics: useful information can be found in literature about visualisation techniques and electromagnetic wave propagation. Also an overview will be given of how existing calculation models and software packages handle this problem with regard to relief, specular or diffuse reflections, diffractions and meteorological conditions.
The acoustical calculations along the propagation paths will be studied with regard to the ground effect or impedance, absorption coefficients, diffraction formulae, integration over octave bands or one-third octave bands, etc. Limitations and inaccuracies in existing models will be identified and reported to work package 2 for further examination and improvement. At the end of this subtask, a first design of the integrated model will be decided that will form the framework for subtask 2.
Figure 1 : Complex propagation paths in urban situation. Ray paths between noise sources and receiver point include multiple reflexions and diffractions.
Task 2 : Implementation of the integrated model
The aim of this subtask is to integrate the propagation model from subtask 1 with the results from work packages 1 and 2. This integration is not straightforward and some simplifications will have to be made to make the model useful for engineering tasks and noise mapping.
The models developed in the separate work package 1.1 and 1.2 will provide all the useful information for the source description in terms of source location, acoustical power output and directivity. These data will be made available to work package 3 through appropriate databases.
The main tasks in this work package is to derive and/or adopt formulae (empirical, interpolated, analytical,…) or heuristical combinations of formulae for the acoustical part of the integrated model. The results from task 3.1 as well as results from work packages 1 and 2 will form the scientific basis for this part of the job. All important decisions are to be made at this point of the project and the integrated model must be fully described, so that software implementation can be done without any ambiguity. All decisions, simplifications, additional hypotheses, known limitations and inaccuracies must be clearly reported.
The next task is to implement the integrated model in a prototype software package. This includes the geometrical part as described at the end of task 3.1, the emission models and databases from work packages 1.1 and 1.2 and the acoustical formulae from this subtask. Difficulties encountered in this stage of the project will be reported and the description of the model improved.
The software implementation will be checked for conformity with the described model by doing the calculations "manually". The accuracy of the integrated model will be evaluated as compared to the reference model and observed discrepancies will be reported and possibly corrected.
The basic software will be extended so that it can run over any given set of input data such as traffic conditions, meteorological conditions, ground, road and rail characteristics. This software will be used in task 3 for tuning the parameters of the model for the purpose of the calculation of long time averaged indicators.
Task 3 : Statistics and data collection
In this work, information will be gathered about the statistical description of the input data and the influence of their variations on the outcome of the model.
For the purpose of the prediction of long time averaged noise indicators, input data is best described in terms of stochastical variables. One can of course run the software with a large amount of (statistically relevant) different input parameters and make the average over all these situations (e.g. one can simulate an averaged year by running the software on a 1 hour basis, thus making 24 x 365 calculations and calculate the long time averaged indicators from these results…), but this is inefficient because it is very time consuming. Therefore, a statistical description of the input data is necessary.
An important point is that input data are not necessarily independent from each other as is assumed in most of the existing methods. For example, yearly averaged values are used to describe traffic conditions in order to calculate a yearly averaged emission level, and yearly averaged climatological data are used for the prediction of the propagation loss. However, there is an obvious interference between both: winter conditions both have an influence on driving conditions and on meteorological effects, summer conditions in areas with great touristic attractions go hand in hand with an increase of traffic. Therefore, also combined effect has to be investigated.
Noise emission from individual cars and trains are largely a function of speed. Traffic conditions are mostly described by flow and main speed. But, because their is no linear relation between speed and noise emission, the mean speed cannot be used to calculate the averaged emission level. Here also, a more realistic statistical model is needed. Even the height of the source should be considered a stochastical variable, especially when vehicles are grouped together into larger categories.
Road surfaces, track and wheel roughness are variable in time and depend on maintenance conditions. It is not possible to predict the exact state of these, so once again the data are subject to statistics.
The influence of meteorological conditions on noise propagation is very different from one situation to another: the higher the receiver or the source, the lower will be this influence. The same thing is true for very hilly terrains where sound travels over valleys. About the meteorological influence on noise propagation in urban areas, little is known, but results from work package 2 in combination with this work package will give information about these statistical effects.
The geometrical model used for the description of relief, buildings, barriers, roads and railway platforms will be a simplification of reality : smooth relief will probably be approximated by some plane surface model and buildings will be represented by extruded rectangles or polygons, complex barriers will be replaced by infinite thin walls, etc. The influence of such simplifications on the predicted propagation losses should be described in a statistical way.
Ground impedances and reflection factors for buildings or barriers can be measured, but this task is impossible when noise mapping is to be carried out over large areas, so here also, statistical classes have to be defined that can be assessed through simple visual inspection or derived from GIS data.
The statistical models will be derived by using the test software developed in subtask 3.2. The partners in this task will make calculations with a variation of different input parameters. The results will be presented properly in graphs with noise levels at different receiver points in relation to stochastic variables such as the traffic speed, meteorological conditions, the degree of geometrical simplifications,… Combined effects will be simulated by simultaneous variation of different parameters, based on realistic long time statistics, taking into account the coherence of different parameters.
Because of the fact that in this phase prototype software is used, possible problems with the software must be reported and action must be taken to correct it ; either within the software itself, or in the description of the model.
The statistical data will be used in subtask 4 in order to estimate the accuracy of the model.
Task 4 : Accuracy of the engineering model
The engineering model will include further simplifications compared to the full reference model and complete source model. These simplifications imply for example :
Based on the statistical analysis carried out in subtask 3.3, the maximum degree of simplification will be specified so that the required accuracy can be achieved (see page 11 of this Technical annex) with minimum effort and cost. Also, the achievable accuracy of the calculations must be considered in relation to the required computer power and calculation time.
The possible simplifications depend on different situations. For example, a higher degree of simplification can be justified in rural areas as compared to urban situations. Requirements can also vary among different European regions since some of them have dominant wind or sunshine without clouds nearly 100% of the time, so there may be no need for calculations under other climatological conditions. Guidelines will be given for countries that want to implement further simplifications of the model with acceptable loss of accuracy.
Special attention will be paid to the "evening" situation and its climatological conditions. For instance, in most countries evening can be considered identical to day conditions in the summer while night conditions apply for the winter. This might lead to supplementary simplification of the calculations of LDEN and LNIGHT as annual averaged quantities.
Task 5 : Test cases
This task is to create some test cases that can be used for conformity checks by third parties that want to implement the new model in their own software packages (either for their own use, for commercial exploitation or as application service providers). The test cases should cover a sufficiently large range of input data and propagation distances. Manual checks made in subtask 3.2 may also be included as test cases.
Some typical measurement sites from work package 4 will be among the test cases. This way, the integrated model can be checked against the experimental data. This requires correct processing of the experimental data in work package 4 so that the emission part and the propagation part will be evaluated independently. Then, the prototype software will be run on the selected test cases. In work package 4, the results will be compared with the experimental data taking into account the expected accuracy of the model. If discrepancies occur between the experimental data and the results of the calculations, action must be taken for fine tuning of the integrated model or proposals should be given for further research and development of the model.
For some test cases, simulations will be carried out using the interim models recommended by the European Directive (the Dutch model for railway sources, the French model for road sources). The expected result of this test is that the new integrated model produces more accurate predictions than the existing national models. The test cases produced by this subtask will be included as an annex in the proposal for a future European standard in work package 5.