Need an account? Click here to sign up. Download Free PDF. Francesco Asdrubali. A short summary of this paper. Duranti, 67 - Perugia, Italy. E-mail: fasdruba unipg. The use of in situ techniques is promising, but their accuracy has to be duly verified, even in comparison with well-known standardized procedures. Sound insulation and reflection properties have been measured through a MLS-based technique in an outdoor test field.
The paper analyzes the procedures that mainly influence the accuracy: correction for wave spreading and time windowing. Repeatability of the in situ method for sound insulation is satisfying and its results look consistent with simple prediction models. Nevertheless, in situ data can be overestimated at low frequencies, due to the overlapping of the transmitted and diffracted components.
The method has to be carefully employed when the sample shows apertures as slits or holes, unless a different kind of sound propagation is assumed at the receiving side. A good agreement was found between in situ and laboratory sound insulation data, while in situ and laboratory absorption properties show poorer correlation.
Thanks to the recent promulgation of new national noise legislation, huge investments are expected in the next years for noise abatement along the Italian road network. The intrinsic properties, indeed, provide a more reliable characterization of noise barriers rather than the extrinsic ones i. Insertion Loss , that strongly depend on several different factors. Two kinds of measurement methods are commonly employed to evaluate the intrinsic properties of noise barriers: laboratory methods, using a diffuse sound field in a reverberation room, and in situ methods.
The latter better simulate the real operating conditions of traffic noise barriers, which are usually exposed to perpendicular or slightly oblique rather than diffuse sound fields.
The paper focuses on these latter methodologies. The overall impulse responses provide information on reflected or transmitted responses, while the free-field ones allow the extraction of the incident responses. Intrinsic properties are finally calculated as ratios of power spectra. In the framework of an agreement with the Italian Ministry for Environment, four samples of noise barriers made of novel sustainable materials have been tested at the University of Perugia in a dedicated outdoor test field Table 1.
Table 1 — Characteristics and single-number ratings of the tested noise barriers. A correction for wave spreading has to be performed, as the sound intensity decreases with the distance dSM of the receiver from the source. In order to analyze the efficacy of the spherical propagation assumption in the time domain, free-field impulse responses have been recorded at several distances between source and microphone sampling frequency: Hz, MLS order: 16, averages: 8.
In Figure 1 a the absolute amplitudes of the first peaks are correlated to their corresponding arrival times: tsample stands here for the time instant calculated multiplying the first peak sample number by the sampling frequency.
Indeed, the measurement chain, as noticed by De Geetere [4], introduces a delay corresponding to a characteristic distance, that can be evaluated by correlating the measured distance dSM and the one equivalent dpeak to the first peak arrival time. This does not occur employing the recorded time. As far as the frequency domain, solid lines in Figure 2 a show the normalised mean third-octave band spectra of the incident components extracted at several distances dSM.
From this point of view, the geometrical correction prescribed for SI seems to match the one prescribed for RI signal multiplication by the time. On the other side, noticeable discrepancy appears on the mean spectra at low frequencies: the effect of such distortions on the final results should be further investigated. In any case the corrected time Eq. Nonetheless, the fluctuations from the mean look higher in the low-frequency range as the measurement chain frequency response is poorer in this range and the hypothesis of spherical propagation may not be suitable when the distance dSM becomes comparable to the wavelength.
E 1. E Normalized mean spectrum Maximum relative error Normalized mean spectrum Maximum relative error 1. In Figure 2 b the dependence of the transmitted power spectra on the barrier- receiver distance dM has been investigated. Transmitted components were isolated from the overall responses sampling frequency: Hz, MLS order: 16, averages: 32 : unlike the incident components, a peak translation and relative amplitude attenuation were not apparent in this case.
This means that the hypothesis of spherical propagation could not be totally correct at the receiving side. The analysis of the deviations of the spectra from the mean dashed-and-dotted lines in Figure 2 b shows that is not straightforward to find the best correction method for the transmitted components.
Indeed, the transmission paths do not necessarily coincide with the direct ones and the transmitted sound field can be regarded as a superposition of the sound waves generated by each portion of the barrier. Then a plane wave hypothesis i. Later with further investigation, it was noticed that when W8L32E inline 8 cylinder was running, the environmental noise was especially high. Thus, the noise levels at various locations near the pipeline were measured in detail and it was found out that the level near the reactive silencer is the highest reach- ing about 90dB A , which is 10dB higher than the level at the stack opening.
Results are shown in Fig. One should know that these two measurements were not performed simultaneously, so there is a slight time shift for analysing. Figure 3: W8L32E speed sweep rpm.
Upper left, contour plot for in-duct noise; Lower left, contour plot for noise 1m away from reactive silencer; Upper right, total noise level in the duct line vs.
In general, from the figures to the right, one can see that the total noise inside the duct and the noise outside the duct in front of the silencer behave very similar to each other during sweep.
From the contour plots one may see that the A-weighted noise inside the duct has the dominant frequency range in Hz and Hz, while for the noise outside the duct in front of the silencer the dominant frequency range is only in Hz. The Hz range has been understood as the few-mode region of the DN pipeline [4].
Further, typically there is a transmission loss minimum in curved and doubly curved structures in this frequency range in power plant sized ducts and silencers.
Moreover, the total noise levels inside and outside the duct can sometimes easily increase by more than 8dB, marked by green arrows. These are the moments when the level for all the frequencies start to increase in the spectra, i.
The reasons for the wide band noise are very possibly to be due to the acoustic or mechanical resonances of the silencer together with the transmission loss minimum of the system. The moments of occurrence of silencer in resonance will depend at least on the cavity geometries inside the silencer, exhaust temperature and noise source features. What has been discussed yet so far is only A-weighted environmental noise, and the knowledge learnt from these tests is that the A-weighted environmental noise near the exhaust pipeline is mainly caused by pipeline transmission loss minimum at the few-mode frequency range for a DN duct, excited by exhaust gas pulsation.
When the reactive silencer is in resonance situation the metal components inside the silencer will generate wide band noise and that can easily raise up the environmental noise by 8dB. Designing a silencer avoiding resonance caused by exhaust gas pulsation in pipeline is crucial, which can in general contribute dB environmental noise reduction in certain cases.
In-duct exhaust noise For low frequencies, most of the exhaust noise energy is in the plane wave region. In case of medium speed engines with large exhaust pipeline, the plane wave region is normally below Hz, Fig. Exhaust noise is measured by a probe microphone inserting through the pre-drilled holes on the duct to obtain the in-duct noise level. The idea of measuring six locations before the first exhaust silencer is to get a decent average spatially for the levels of the octave band centre frequencies in the plane wave region.
Since the engines are 4-strokes, the noise peaks in the spectra typically will have 0. However, it has been found out also that sometimes measuring at different locations or same engine with different pipelines one may get very different results.
Reasons are standing waves exist along the pipeline and the acoustic loads are different for different pipelines so that it will influence the results of measured noise source. To get better understanding of the pipeline resonance influence to measured results, in addition to constant speed measurement one should also do the speed sweep measurement. During the sweep rpm, exhaust temperature had a variation of about 50 Celsius.
The total in- duct noise level had a variation about 16dB. For a constant load, 16dB variation is considerably large. Further, the noise spectra are plotted in Fig.
As known, the sound speed changes with temperature. Thus, with unchanged geometry of a pipeline, the resonant frequencies of standing waves will change accordingly. In the contour plot one may see the forward-slash lines representing the resonances of the pipeline, and the backslash lines are the exhaust noise peaks for various engine orders. When the engine order lines come across with pipeline resonance lines, high noise level will occur, in case of which the measured level is an overestimate.
It can also happen that the engine order lines come across with anti-resonance lines, in case of which the measured level is an underestimate. In other words, at constant load and certain speed condition, the measured noise spectrum in pipeline is not reliable due to the influence of pipeline resonances. In the peak-hold spectra one can also see that at 50 Hz as an example having the highest noise level.
This is due to exhaust noise at engine order 4. The results of simulation were later compared with measurements. Geometry of the exhaust duct line was created using NX software. The assembly consists of 13 parts. In each part, only the fluid volume was included. The parts in the assembly were meshed separately and the glue-coincident mesh mating condition was applied. The final geometry and part of the mesh is shown in Fig.
Separate acoustic models of the silencers were built and tested. First, volume geometries, meshes and acoustic models were built in Actran software. Silencer TLs transmission loss were calculated in Actran software using these models. Then silencer volume geometries were transferred in stp-format to NX, meshed there. The purpose of these operations was that the silencers, being the most complicated components in the duct line, are correctly modelled. The duct line is a major acoustic system.
To reduce solve times and facilitate the overall handling of the model, the model was built using connected acoustic subsystems. Each of the junctions was modelled as an impedance-type connection having a high porosity. In the input of the model there is a velocity constraint. It terminates the model. Several virtual acoustic sensors were placed on the model for validation purpose.
In the simulation, it was found out that up to Hz over acoustic modes exist the exhaust pipeline. Comparing to a smooth duct line, the current duct line can easily increase the exhaust sound power by 7dB up to Hz. Firstly, calculated sound power and measured peak hold sound pressure spectrum from sweep at one of the standard measurement locations before silencer are shown in Fig.
The quantities are not the same, so a comparison of these is not very rigorous. However, both curves should indicate resonances and anti-resonances in their own way. Certain similarities are indeed evident. These include the peaks between 20 and 35 Hz and the deep minimum around 40 Hz. The broad group of peaks between approx. After 80 Hz the characters become different.
Between 90 and Hz the peaks in the simulated sound power are much more distinct than in the more broadband measurement result.
0コメント