Validation and evolution of the road traffic noise prediction model NMPB-96 - Part 1: Comparison between calculation and measurement results

Transcription

1 The 2001 International Congress and Exhibition on Noise Control Engineering The Hague, The Netherlands, 2001 August Validation and evolution of the road traffic noise prediction model NMPB-96 - Part 1: Comparison between calculation and measurement results F. Besnard (1), H. Lefèvre (2), V. Zouboff (3) (1) Service d Etudes Techniques des Routes et Autoroutes 46 av. A. Briand, BP 100, BAGNEUX Cedex, France (2) Laboratoire Régional des Ponts et Chaussées de Clermont-Ferrand 8-10 rue Bernard Palissy, CLERMONT-FERRAND Cedex 2, France (3) Laboratoire Régional des Ponts et Chaussées d Angers 23 avenue de l Amiral Chauvin, LES PONTS DE CE, France Abstract The French road traffic noise prediction model NMPB-96 takes into account the influence of the real weather conditions of the site on sound propagation. In order to complement the model validation and to prepare its final version, a heavy measurement campaign took place on six sites bordering motorways. On each site sound levels, weather parameters and traffic were measured on 10 minute periods during at least two weeks so as to observe a large diversity of weather conditions. These 10 minute samples have been classified according to the weather conditions. The sound attenuation due to propagation has been compared with the predictions given by NMPB-96 for whole daytime and night-time periods. The agreement between calculation and measurement results is satisfactory, even at large distances, for microphones with a direct view of the road, and also when the road is screened off by a vertical obstacle. For microphones located below the road level (road on embankment or viaduct), the model overestimates the measurements. The work carried out to upgrade NMPB is described in part 2: Improvements based on theoretical methods. 1. Introduction The French road traffic noise prediction model "NMPB-Routes-96" [1] takes into account the influence of the real weather conditions (wind and temperature gradient) of the site on sound propagation. It was developed within a short period of time in order to meet a new regulatory requirement. During the preparation phase, the method was validated by comparison with theoretical models (boundary elements, parabolic equations) and a first series of in situ measurement results. However, the timeframe allocated then was not sufficient to undertake as many measurements as required. For this reason, it was decided to publish NMPB in an experimental form and schedule an additional validation measurement campaign. This article deals mainly with the comparison between the measured sound levels and those predicted by NMPB. The principles of the model, the measurement process and the preliminary data processing are described in detail in [2, 3] and are only recalled shortly hereafter.

2 2. Principles of NMPB As NMPB applies only to point sound sources, the first step of the calculation consists in breaking down the road into such sources. Then NMPB enables to calculate the equivalent sound level L H in homogeneous atmospheric conditions (vertical gradient of sound speed zero) and the average sound level L F in conditions said to be "favourable to propagation" (vertical gradient of sound speed positive). The long term equivalent sound level L LT is calculated by: L LT = 10 lg [p LF + (1-p) LH ] (1) where p is the proportion of time when conditions are favourable to propagation (p is between 0 and 1), and varies according to the source-receiver direction. In other words, sound levels existing in conditions "unfavourable to propagation" (vertical gradient of sound speed negative) are biased upwards by the L H level. This method overestimates the actual sound levels but takes a safe approach for the road designer. It ensures as well that L LT is not lower than L H, as required by the French regulations. 3. Measurement principle and data processing 3.1 Measurement principle Six sites were equipped in 1997 and 1998 (cross section views of the sites are given in part 4). At each measurement site, the experiment consisted in acquiring, for a minimum period of two weeks, micrometeorological and acoustic information sampled in slots of 10 minutes, the minimum representative time for meteorological conditions. The meteorological data were collected by a mast comprising three sensor levels for air temperature, wind speed and wind direction, located at 1 m, 3 m and 10 m from the ground. The sound levels LAeq were generally measured at four distances (up to 400 m from the road) and at two different heights (2 m and 5 m above the ground). An additional measuring point located immediately next to the road and at a height of 5 m provided a reference level used to eliminate variations in road noise emission when processing the measurement results. 3.2 Meteorological and acoustic characterisation of samples The space was split into a succession of directions in 20 degree steps. For each 10 minutes sample and each azimuth, the vertical gradient of the sound speed at 6 m above the ground was calculated from the meteorological data collected. Three categories of propagation conditions: "favourable" (F), "homogeneous" (H) and "unfavourable" (U) were defined below and above sound speed gradient values generally set equal to ± s -1. Therefore, a set of propagation conditions F, H or U was associated with each 10 minutes sample according to the directions of the space such as the example shown in Table 1. azimuth (degrees) conditions F F H U U U U U Table 1: example of weather characterisation of a sample. For each 10 minutes sample, there was also a corresponding L Aeq level at each receiver. In order to rule out traffic variation, the study only took into consideration the attenuation A between the reference point and each of the other points.

3 3.3 Comparison of measurement and calculation results As it is not possible experimentally to measure a long term sound level, the objective of the study was to validate (1) the calculation method in "pure" conditions i.e. identical (favourable or homogeneous) in all directions and (2) the calculation method of the long term level. As the trends observed on each site are generally the same for the three types of conditions, only the process and the results regarding the long term level are presented hereafter. The checking was based on data corresponding to complete intervals 6h - 22h or 22h - 6h, called "medium term periods". For each medium term period, the measured attenuation A MT (exp) was compared with the value A LT (calc) calculated by NMPB using the actual occurrences during the period p(exp) in each azimuth. 10 to 15 periods were selected for each site in order to corroborate the results from a statistical standpoint. Each simulation was undertaken by two different operators on two software programs applying NMPB. 4. Results The differences are shown in the form δ = A MT (exp) - A LT (calc). A positive value of δ means that the model overestimates the sound level at the receiver. Since the principle of NMPB overestimates deliberately the noise levels in conditions unfavourable to propagation, the comparison only deals with the periods in which favourable conditions occur at least 20 % of the time in at least 5 sectors of 20 degrees. The results are classified so as to distinguish as best as possible the different physical effects. The averages and the standard deviations are calculated over the results of the two software programs, i.e. twice the number of periods. It should be noted that the provisional results presented in [3] have been revised and are no longer valid. 4.1 Ground effect The relevance of the formulae for ground effect can be checked on sites B (receivers R3 to - Figure 1 left) and F (all receivers except - Figure 1 right). 10m R3 16m 59m 108m R3 3m 13m 75m 150m 225m 300m 45m 90m 180m 270m Figure 1: Cross section views of sites B (left) and F (right); the microphones are 2 m and 5 m above the ground. Tables 2 and 3 show the results respectively for sites B and F. The calculation on site B is satisfactory overall, though a little low at the intermediate receivers and. R3 Average of δ db(a) Standard deviation of δ db(a) Table 2: Site B - Results corresponding for each receiver to 8 medium term periods.

4 R3 * Average of δ db(a) Standard deviation of δ db(a) Table 3: Site F - Results corresponding to 10 medium term periods (*except : 6 periods). On site F, the microphones line was located along a rather high corn field which was probably more absorbing than the bare ground considered by NMPB. Additional short duration measurement results lead to consider that 1 to 1.5 db(a) of the overestimation at the far receivers ( to ) was due to the presence of the corn field disregarded by the model. Bearing this in mind, the calculation of the ground effect is deemed satisfactory on site F as well. 4.2 Diffraction by a vertical barrier Only site E includes a barrier (Figure 2). Its overall height (4.5 m) lowers by far the influence of the ground effect. 2.5m 2m R3 6m 12.7m 100m 150m 200m Figure 2: Cross section view of site E; the microphones are 2 m and 5 m above the ground. R3 * Average of δ db(a) Standard deviation of δ db(a) Table 4: Site E - Results corresponding to 6 medium term periods (*except : 5 periods). The averages (Table 4) are satisfactory overall. The rather high standard deviations are partly due to one specific period. If this period is ruled out from the sample, the standard deviation drops at 1.5 db(a) at R3 and lower than 1.1 db(a) at the other receivers. 4.3 Diffraction by a horizontal platform Configurations where receivers are located below the road platform level are met on sites B ( and - Figure 1 left) and C (all receivers - Figure 3). on viaduct 16m 24m 32m 265m 22m 115m R3 Figure 3: Perspective view of site C; the microphones are 2 m and 5 m above the ground.

5 On both sites, the ground effect can be considered as negligible and the calculation result is only governed by the diffraction on the edge of the platform. The noise levels in such a configuration are overestimated by far (Table 5). On site C, the results of the two software programs differ by at least 1 db(a) for 61 % of the calculations, and by at least 2 db(a) for 7 % of the calculations. However this is not enough to explain the high standard deviations observed. Site B Site C R3 * * Average of δ db(a) Standard deviation of δ db(a) Table 5: Sites B and C - Results corresponding to 8 medium term periods on site B and 9 medium term periods on site C (*except : 8 periods and : 7 periods) 4.4 Trench roads Sites A (Figure 4 left) and D (Figure 4 right) include a trench road. Here both the ground effect and diffraction govern the differences between calculations and measurements. 2m R3 16,6m 50m 100m 200m 400m 0 350m R9 200m 100m 50m Figure 4: Cross section views of sites A (left) and D (right); microphones are 2 m and 5 m above the ground. On site A (Table 6), the average differences are a bit high, mainly at receivers located 5 m high. One of the five medium term periods analysed shows atypical phenomena; when ruled out from the sample, both the averages and the standard deviations drop by more than 1 db(a) for most receivers. R3 Average of δ db(a) Standard deviation of δ db(a) Table 6: Site A - Results corresponding to 5 medium term periods. On site D, a checking using a theoretical model [4] showed that the reflection of the sound on the cutting slope, disregarded by NMPB, causes an increase by about 1.5 db(a) at the reference microphone, whereas the noise levels at the other receivers are not affected by the reflection. Therefore the noise level at cannot hold the intended function of reference. It was decided to compare exceptionally on this site the measured and calculated noise levels, instead of the attenuations. + + R3 + +R9 0 Distance - Height (m) Number of periods Average of L db(a) Std deviation of L db(a) m 50m R3 100m Table 7: Site D - Receivers of both sides of road are put together according to their location. 200m

6 Table 7 shows the results regarding the differences L = L LT (calc) - L MT (exp), putting together the receivers of both the northern and southern sides according to their position. A positive value means that the calculation overestimates the noise level. The average at the reference microphone confirms the error of 1.5 db(a) by NMPB due to the reflection on the slope. It is therefore relevant to carry out the comparison with the absolute noise levels. At 0, the model overestimates the noise level, but the measured LAeq was rather low - about 45 db(a) - thus very sensitive to background noise. At the other receivers, the average differences remain in an acceptable range. 5. Conclusion and outlook for NMPB NMPB is meant for assessing the noise impact of a road in all usual situations (distance, topography, kind of ground). On such sites the effect of weather conditions on sound propagation varies much with time and space, but such a model can only account for it roughly. Expecting an accuracy better than 2 db(a) would be therefore unrealistic. In addition, as far as the observance of mandatory noise level thresholds is concerned, a slight overestimation is better than an underestimation. The calculations of the ground effect and the diffraction by a barrier or a road cutting ridge are consequently deemed satisfactory. On the contrary, when the sound is diffracted by the edge of a horizontal platform, the noise levels at the receiver can be overestimated by 3 to 5 db(a). The first improvement studied consists in lowering the equivalent source height from 50 to 5 cm, since the latter fits better the real sources [5]. However this change questions the relevance of the formula for ground effect in favourable conditions, derived from ISO and unsuitable for so low sources. Studies are in progress as well in order to include the reflection on a road cutting slope and the effect of small barriers, disregarded by the present version of NMPB. This work is presented in [4]. Acknowledgements The measurement campaign and the data processing were carried out by the LRPC (regional laboratories of LCPC) of Angers, Blois, Clermont-Ferrand, Lille, and Strasbourg. erences 1. "Bruit des infrastructures routières - Méthode de calcul incluant les effets météorologiques (NMPB-Routes-96)", CERTU - CSTB - LCPC - SETRA (1997). ("Noise of road infrastructures - Calculation method including meteorological effects"; in French). 2. Y. GABILLET et al., "Comparison of two methods for predicting traffic noise", Proceedings Inter-noise 96 (Liverpool), pp F. BESNARD et al., "Traffic noise prediction method taking into account weather conditions: a validation measurement campaign", Proc. Inter-noise 99, pp J. DEFRANCE et al., "Validation and evolution of the road traffic noise prediction model NMPB-96 - Part 2: Improvements based on theoretical models", Proc. Inter-noise J.F. HAMET et al., "Acoustic modelling of road vehicles for traffic noise prediction: Determination of the source heights", Proc. of the 16th ICA, Seattle, USA (1998).