- Email: [email protected]
Application of sound intensity and partial coherence to identify interior noise sources on the high speed train
Mechanical Systems and Signal Processing 46 (2014) 481–493
Contents lists available at ScienceDirect
Mechanical Systems and Signal Processing journal homepage: www.elsevier.com/locate/ymssp
Application of sound intensity and partial coherence to identify interior noise sources on the high speed train Rongping Fan a,b, Zhongqing Su b, Guang Meng a,n, Caichun He c a b c
State Key Laboratory of Mechanical System and Vibration, Shanghai Jiao Tong University, Shanghai 200240, China Department of Mechanical Engineering, The Hong Kong Polytechnic University, Hung Kom, Hong Kong Zhuzhou Times New Materials Technology Co., Ltd., Zhuzhou 412001, China
a r t i c l e i n f o
abstract
Article history: Received 4 May 2011 Received in revised form 26 November 2013 Accepted 29 November 2013 Available online 14 February 2014
In order to provide a quieter riding environment for passengers, sound quality refinement of rail vehicle is a hot issue. Identification of interior noise sources is the prerequisite condition to reduce the interior noise on high speed train. By considering contribution of noise sources such as rolling noise, mechanical equipment noise, structure-borne noise radiated by car body vibration to the interior noise, the synthesized measurement of sound intensity, sound pressure levels and vibration have been carried out in four different carriages on high speed train. The sound intensity and partial coherence methods have been used to identify the most significant interior noise sources. The statistical analysis results of sound intensity near window and floor on four carriages indicate that sound intensity near floor is higher than that near window at three traveling speeds. Ordinary and partial coherent analysis of vibro-acoustical signals show that the major internal noise source is structural-borne sound radiated by floor vibration. These findings can be utilized to facilitate the reduction of interior noise in the future. & 2013 Elsevier Ltd. All rights reserved.
Keywords: Interior noise Sound intensity High speed train Partial coherence
1. Introduction The noises within the high speed train are caused by many noise sources, such as rolling noise, mechanical equipment noise, air conditioning noise, structural-borne noise radiated by car body vibration and aerodynamic noise. With traveling speed increase and vehicle weight-lighting, the rolling noise and aerodynamic noise of high speed trains are greatly increased. In order to provide quieter riding environments for passengers, more attention [1] has been paid to reducing the steeply increased interior noise and vibration. The improvement in sound and vibration quality can be achieved in three aspects: source, transfer path, and receiver. It is the subject of this paper to identify the noise sources inside the high speed train in order to reduce the noise levels. The finite element and boundary element method were used by Mohanty et al. [2] to distinct the interior acoustic field by the natural modes and shapes of cabs. Ding [3] developed the finite element method to evaluate panel acoustic contribution. Liu [4] analyzed the panel vibration and interior noise of vehicle cavity by structure strength. Liang [5] investigated panel acoustic contribution at the specific mode to identify the major sound source based on boundary element method and finite element method. Vehicle structures were complex, so the numerical models had to be significantly simplified. However, the correction of numerical results was highly dependent on accuracy of the models.
n
Corresponding author. Tel.: þ86 2134206332x816; fax: þ 86 21 34206006. E-mail address: [email protected] (G. Meng).
0888-3270/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ymssp.2013.11.014
482
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
The acoustic arrays, sound pressure level and sound intensity have been measured to identify the main sources in the rail vehicles. Microphone arrays or combination of different sensors have proved useful to identify and predict the noise levels and location of major sources outside high speed trains [6–10]. Sound intensity measurement can offer both direction and position of sound sources. Sound intensity theory was used by Liu [10] to measure noise on high speed railway vehicle and analyzes the interior noise distribution. Sound intensity techniques were applied to identify engine front noise sources [11]. Although the force vector method and full matrix method were commonly used for noise path analysis [12,13], these two methods needed extra time and cost to deal with the problem of the connection between the sound and receiver. Spectral analysis method of partial coherence analysis [14,15] was used to identify vibro-acoustical sources of noise emitted inside vehicle compartments. The first step of signal conditioning in the partial coherence method is to rank the sources. Hilbert transform was used to check causality between signals by Fouladi [15] and Bee [16]. In this paper, the sound intensity is used not only to investigate the characteristics of sources but also to rank the potential sources. In this paper, acoustic-vibrational measurements are carried out and sound intensity and partial coherent methods are used to distinguish major interior noise sources inside a specific high speed train. The paper is organized as follows: firstly, a short description of the sound intensity theory along with partial coherence is given; in Section 2, the methodology of sound and vibration measurement within high speed train is described in details; in Section 3, the measurement data will be processed to identify the main sources of interior noise on carriage; finally, the conclusion of noise source identification will be made and the method to reduce the interior noise will be suggested. 1.1. Partial coherence analysis For the system with multi-input Xi and single output Y, input signals are correlated. Partial coherence analysis can rule out linear effects of other inputs from problem by taking advantage of conditional output method. Coherence function is an indication of linear relationship between input X and output Y for a system. Ordinary coherence analysis often confuses the contribution of each input Xi to the output. Partial coherence technique has been applied to find the contribution of sources to overall response in the vehicle interior vibrio-acoustic problems [14,15]. Partial coherent method can remove linear effects of other inputs and make a set of inputs irrelevant. For example, consider X ii 1 and X ij 1 as two partially correlated inputs. The conditioned signal X ij 1 from signal X ii 1 can be obtained using Eq. (1) X ij ¼ X ij 1 H iij 1 X ii 1 ; where
H iij 1
ð1Þ
is the frequency response function between the two signals
H iij 1 ¼
Giij 1 Giii 1
X ij 1
and
X ii 1
;
ð2Þ
where Giii 1 is autospectrum of signal X ii 1 and Giij 1 is cross-spectrum of signals X ij 1 and X ii 1 . The two sides of Eq. (1) are multiplied by conjugate X nr . The conditional power spectrum can be given by Girj ¼ Girj 1 H irj 1 Giir 1 ; Girj 1 Gijj of
Girj ,
where spectrum
and Giir 1 are signal X ij can
ð3Þ the cross spectra. If we replace j with r and insert Eq. (2) into Eq. (3), conditional autopower be obtained by
Gijj ¼ Gijj 1 ð1 r iij 1 Þ;
ð4Þ
where Gijj 1 is conditional autopower spectrum of signal X ij 1 and partial coherence value r iij 1 is defined by r iij 1 ¼
ðGiij 1 Þ2 Giii 1 Gijj 1
:
ð5Þ
X ij will rule out the linear effects of the stronger signal X ii 1 from signal X ij 1 . Then the effect of each input on output is identified by conditional output power spectrum. 1.2. Sound intensity theory Sound intensity is defined as the sound power per unit area. The sound source contribution can be determined by the sound intensity [13]. The time-averaged sound intensity I at a point A in a sound field can be expressed as Z Z 1 1 Ii dt ¼ I dt ¼ 〈pv〉t ð6Þ I¼ T T T T i where the instantaneous sound intensity Ii is the product of sound pressure p and particle velocity v, and T is the average time. Particle velocity v is related to sound pressure by the following equation: Z 1 t v¼ ∇pdτ ð7Þ ρ0 1
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
Substitution of Eq. (7) into Eq. (6) yields Z t 1 p ∇pdτ I¼ ρ0 1 t Sound intensity compound in a specific direction l is expressed as Z t 1 ∂p dτ p Il ¼ ρ0 1 ∂l t
483
ð8Þ
ð9Þ
Sound pressure gradient in the direction l is evaluated in terms of sound pressures at two adjacent points to point A ∂p p p2 1 ∂l Δr
ð10Þ
where Δr is the distance between two measured points and the point A is equally far away from these two points. The sound pressure at the midpoint A of the two microphones can be obtained by taking the arithmetic average of two sound pressure levels p
p1 þ p2 2
Placing Eqs. (10) and (11) into Eq. (9), we get Z t 1 p1 þ p2 ðp2 p1 Þdτ Il ¼ 2ρ0 Δr 1 t
ð11Þ
ð12Þ
When sound pressure is harmonic, sound intensity can be expressed as Il ¼
P A1 P A2 sin ðφ2 φ1 Þ 2ρ0 ωΔr
ð13Þ
where P A1 and P A2 are the magnitudes of sound pressure signals p1 and p2 , respectively. When the phase difference Δr between two microphones is very small, the sound intensity can be simplified as Il
P A1 P A2 ðφ2 φ1 Þ 2ρ0 ωΔr
ð14Þ
Through Fourier transform, averaged sound intensity in frequency domain can be expressed as IðωÞ ¼
ImfG12 g ρ0 ωΔr
ð15Þ
where G12 is the one-sided cross-power spectrum density function between two microphones, ρ0 is the air density and ω is the angular frequency. Compared to the partial coherence method, it is quicker and more direct to determine the radiated power of sound sources by sound intensity. 2. Methodology The main objective of this paper is to identify predominant sources of interior noise from the potential sources of noise emissions. There are many sources that influence the interior noise inside the high speed trains. Potential sources are classified into three categories: ▪ The interior noise sources like ventilation or air-conditioning system. ▪ Exterior noise sources like wheel–rail interaction, the propulsion, brakes, compressor and aerodynamics. ▪ Structural-borne noise radiated by vibrating car body. It is difficult to recognize the contribution of each noise source to the interior noise. The influence of interior noise sources like ventilation or air-conditioning system on the interior noise can be measured when the high speed train is still. In this paper, the influence of ventilation or air conditioning is not considered by closing the ventilation and air conditioning when the high speed train was traveling. Sound reaches the carriage interior through two different routes: the air-borne path and the structure-borne path. In the air-borne path, sound travels away from sound sources and transmits through the floor, wall and windows of carriage into the interior. In the structure-borne path, exterior vibration sources such as wheel– rail interaction excite the car body of carriages to vibrate and radiate structural sound into the interior. Because the running speed of high speed train does not exceed 300 km/h, aerodynamic noise is not a major source of noise emission outside the carriage. The acoustic insulation materials were installed on the wall of the carriage. It was easier for the noise to transmit from the window into the interior carriage than from the wall. The floor was connected through the second suspension system of the bogie and excited to vibrate. First of all the window and the floor are considered as potential sources of noise emissions in the test high speed train.
484
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
The high speed train traveled at a maximum running speed of 250 km/h. The layout of carriages of the high speed train is demonstrated in Fig. 1. The test high speed train was made up of two electric locomotives plus nine trailer cars, which were manufactured by two different vehicle factories. The air conditioning system and bogie equipment on carriages made by two factories were different. Measurement of noise and vibration was carried out in the first and second class carriages. The furniture inside the first and second class carriages was identical except that cable TVs were installed inside the first class carriages. These four carriages were the first class carriage C2, the dining carriage C5 and two second class carriages C4 and C6. Common characteristics of the interior noise within the high speed train could be obtained through the measurements in the four carriages. The major interior noise sources could then be identified.
Fig. 1. Schematic diagram of carriage distribution.
Fig. 2. Setup of vibration and noise measurement within each carriage in the running test
Fig. 3. Setup of acceleration transducer and microphones near bogie.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
485
To identify the major interior noise sources when the high speed train was traveling, this paper will focus on the running test. The air conditioning equipment is closed to eliminate the effect of air conditioning noise. The test train traveled at speeds of 120, 160 and 200 km/h on high speed straight level track free of rail joints. To discern the contribution of each source to the overall interior noise and the path way that noise propagates into the carriage interior, acoustical–vibrational measurements were simultaneously carried out. These measurements included the interior and exterior sound pressure levels, sound intensity, and the structural vibration of car body and bogie. The experimental setup in each one of the four carriages was identical. Six microphones denoted by six consecutive characters from A to E, referred to the standard GB/T 12816-2006 [17], were located in the carriage, as shown in Fig. 2(a). Three microphones A, B and C for seated persons were at a height of 1.2 m above the car body floor. The other three microphones D, E and F for standing persons were 1.6 m away from the floor. The microphones A and D are positioned at one end of the carriage. The microphones B and E are located in the middle of the carriage. The microphones C and F were placed in the middle of the gangway at one end of the carriage outside the glass door. Microphones H1 and H2, as shown in Fig. 2(b), were used to measure sound intensity near window. Microphones I1 and I2, as shown in Fig. 2(a) and (c), were used to measure sound intensity above the floor. In Fig. 3, microphone G was used to measure rolling noise. The wind cap around the microphone G was to prevent the extra pressure caused by the aerodynamics near the bogie. Accelerometer 1 was used to measure the vibration on the center of the bogie. Micro-accelerometer 2 attached to the window glass was to measure window vibration, as shown in Fig. 2(a) and (b). Accelerometer 3, as shown in Fig. 2(a) and (c), was vertically located on the car body floor to measure the car body floor vibration. Accelerometer 4 on the seat in the middle of the car body was used to measure seat vibration, as given in Fig. 2(a). All the vibrational and acoustical signals inside the carriage along with rolling noise were simultaneously obtained. The data were collected by a Sony data recorder. The time–frequency analysis results of sound and vibration showed that interior noise and vibration were stationary. Linear averaging used for analysis of stationary signals was applied to equalize all spectra and time records. Hanning weighting with overlap of 50% was used as a general-purpose window for continuous signals. Frequency resolution was 1.27 Hz and 50 spectral averages were implemented for analysis. 3. Results and observations Before both sound intensity and partial coherence methods are used for post-processing of the measured signals; the measured noise within four carriages is analyzed to characterize the spatial distribution of noise inside the carriage.
Fig. 4. Overall values of linear sound pressure levels at six points within four carriages: (a) 160 km/h and (b) 200 km/h.
486
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
The second step is to process the sound intensity signals near sources at different running speeds. The sources can be ranked by the level of sound intensity. The partial coherent power spectra of sources will be calculated according to the partial coherence theory. The coherence analysis between near-field and far field noise and vibration of the structure will be performed. 3.1. Analysis of sound pressure levels Although the acoustic environment within carriages was very complex and the measured carriages were a little different, remarkable characteristics of the interior noise within carriages were obtained. The acoustic spatial distribution within carriages was evaluated by noise levels at six microphones from A to E. The overall value of the sound pressure levels at six points at two speeds of 160 km/h and 200 km/h is illustrated in Fig. 4. As shown in Fig. 4, the overall value of sound pressure levels within carriages shows that the sound pressure levels at points A and B at a height of 1.2 m from the floor were about 10 dB higher than that of locations D and E at a height of 1.6 m. The sound pressure levels at C and F inside the gangway between two carriages are almost identical. The difference in the sound pressure levels inside the carriage and the gangway
Fig. 5. Linear sound pressure levels in one third octave band center frequency at three speeds: (a) 120 km/h, (b) 160 km/h, and (c) 200km/h.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
487
indicates that the sound environments of the carriage and the gangway are different from each other. The sound pressure levels at 200 km/h are higher than those at 160 km/h. It can be observed that there are similar trends in the spatial distribution of noise at the measured points. The sound pressure levels in the one-third octave band center frequency at three speeds of 120 km/h, 160 km/h, and 200 km/h are presented in Fig. 5. It can be seen in Fig. 5 that the low frequency components are dominant at three running speeds. Regardless of the level of frequency components, the shapes of frequency spectral curves of sound pressure levels at evaluation point B are reprehensive of the interior noise. To indicate the effect of rolling noise on interior noise, Fig. 6 gives sound pressure level at location B on carriage C6 and rolling noise at three traveling speeds of 120 km/h, 160 km/h and 200 km/h. It is clear in Fig. 6 that running speed has different influence on different frequency components of interior noise and rolling noise. The low-frequency components are dominant in the interior noise. With increase in traveling speed, the noise in the range of 25–4000 Hz at location B goes up. Since the running speed of the high speed train is below 300 km/h, rolling noise radiated by vibration of wheel and rail was the major external noise source. It can been observed from Fig. 6 that rolling noises at three traveling speeds are wide band noise in the span of 20–8000 Hz. The main energy is distributed in the frequency range of 20–5000 Hz. Rolling noise spectra have multi-peaks which are relevant to resonance of wheel and rail. By comparison of the interior noises at location B and rolling noises at three traveling speeds, it is obvious that car body can prevent rolling noise at middle and high frequencies from transmitting into carriages. The rolling noise is not the major noise source at low frequency within carriages.
Fig. 6. Linear sound pressure levels at (a) measurement point B and (b) rolling noise near wheel and rail on carriage C6. Table 1 Sound intensity levels (dB) near the floor and window within four carriages at three speeds. Running speed (km/h)
120 160 200
Carriage C2
Carriage C4
Carriage C5
Carriage C6
Window
Floor
Window
Floor
Window
Floor
Window
Floor
93.7 95.3 101.1
105.1 108.3 114.1
95.1 96 98.7
107.9 109.2 112
93.5 95.5 97.9
104.6 108.9 110.7
96.2 97.5 99.3
107.1 110.9 112.8
488
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
3.2. Post-processing by the sound intensity method It is obvious in Figs. 5 and 6 that interior noise energy is dominant in the low and middle frequencies. The major purpose of the interior noise source identification in this paper is to detect the potential sources causing the interior noise at low and middle frequencies. The low and middle frequency noises near the window and floor are made up of radiated noise by vibration of window and floor, as well as air noise that rolling noise transmits through window and floor into the interior. When the high speed train travels, aerodynamics cause windows to vibrate and radiate noise. Sound intensity near window and floor surfaces at three speeds is calculated using Eq. (15). Table 1 lists overall sound intensity levels near surfaces of window and floor within four carriages at three running speeds. It is clear that overall sound intensity levels near floor surface are higher than those near window within four carriages at the same running speed. As running speed increases, the sound intensity levels near window and floor rise. The sound intensity levels in the one-third octave band center frequency within carriage C6 at three running speeds are shown in Fig. 7. These curves show where overall values of sound intensity near window and floor differ. The intensity levels near window surface have the predominant components below 100 Hz. The low frequency components up to 160 Hz are dominant in the sound intensity levels near the floor. As traveling speed increases, the sound intensity levels near window and floor increase in the entire frequency range. It is quite clear that the components of sound intensity in the one-third octave center frequency range of 20–200 Hz near floor are higher than those near window. At different traveling speeds of the high speed train, the levels and frequency spectra of interior noise signals and sound intensity vary within different carriages. Sound sources make different contribution to the interior noise. However, the general trends of interior noise signals and sound intensity inside four carriages at three running speeds are very similar. Therefore, it is assumed that the interior noise and sound intensity measured within carriage C6 is reprehensive of the interior noise signals inside the high speed train at different running speeds. In the following analysis, signals measured within carriage C6 at the speed of 200 km/h will be used to identify the interior noise sources.
Fig. 7. The one third octave band spectra of the sound intensity of (a) window and (b) floor within carriage C6.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
489
To rank the noise sources of the floor and window, the linear sound pressure levels at measurement points H1 and I1 are shown in Fig. 8. It is clear that the main noise energy near window and floor is in the range of 20–200 Hz. The sound pressure levels above 200 Hz near the floor are higher than those near the window. For better understanding of the effects of sound sources, the sound intensity spectra near surfaces of window and floor in the narrowband frequency are illustrated in Fig. 9. It can be seen that the main components of sound intensity levels radiated by these two regions are in the low frequency range of 20–200 Hz. From the above analysis, it can be concluded that sound power is larger near the floor than near the windows.
Fig. 8. Linear sound pressure levels at (a) measurement point H1 near window and floor and (b) measurement point I1 near floor.
Fig. 9. The spectra of sound intensity radiated by window and floor within carriage C6: (a) window and (b) floor.
490
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
3.3. Application of partial coherence method to identify main sources of vibration and noise Analysis results of sound intensity show that noise near the floor makes major contribution to the interior noise at the measuring points within carriage. Vibration of the floor and window is mainly transmitted by the vibration of the bogie and mechanical equipment. The rail–wheel vibration is transmitted to carriage floor through primary suspension and bogie. The noise near the floor and window consists of structural-borne noise radiated by vibration of floor and window and airborne noise transmitted from rolling noise. The ordinary coherent analysis and partial coherent analysis of carriage structural vibration and near-field and far-field noise can show whether the noise near the floor and window is structural-radiated noise or rolling noise. Fig. 10 illustrates vibration frequency spectra of floor, window, seat and bogie of carriage C6 at the speed of 200 km/h. It can be seen in Fig. 10 that acceleration of bogie is in the wide frequency band range of 0–5000 Hz. The first two peaks of bogie vibration are at 57.2 Hz and 1.71 k Hz. The major energy of floor vibration is distributed within 0–3000 Hz. The first three peaks of vibration are at 76.3 Hz, 91.6 Hz, and 95.4 Hz. The vibration of the window is centered below 1000 Hz and there is a low peak around 2000 Hz. The seat vibration is below the low frequency of 100 Hz. The coherence analysis between the vibration of bogie, floor, and windows is carried out in order to check the relation between the vibration of the floor, the window, and the bogie. The coherence functions between bogie, floor and window are illustrated in Fig. 11. From both Figs. 10 and 11, it can be seen that the coherence coefficient of window and bogie is greater than 0.6 below 800 Hz where the vibration magnitudes of bogie and window are dominated. The window vibration is linearly related to bogie vibration. The coherence coefficient between bogie and floor is approaching 0.8 at main frequencies of floor vibration below 3500 Hz. The coherent values between seat and floor as well as seat and bogie are almost 0.8 at the low frequency of 50.9 Hz. It indicates that bogie vibration is the main source to excite vibrations of the floor and seat. Although aerodynamic noise is not considered as the air noise that transmits through window into interior under traveling speed of 300 km/h, vibration of window surface is possibly excited by bogie vibration and aerodynamics at the high speed of 200 km/h. Because the level of the rolling noise is quite high, it is possible that acoustical excitation of rolling noise results in floor vibration. The effect of aerodynamics and acoustical excitation of rolling sources as well as other excitation sources on the vibration of the floor and window can be obtained by application of partial coherence analysis to remove the effect of bogie vibration on the floor and window inside the carriage. Autopower spectra of partial coherence for the window and floor are shown in Fig. 12. Compared to accelerations of floor and window shown in Fig. 10, the frequency span of vibration of window and floor become narrower and vibration spectral magnitudes become tremendously low.
Fig. 10. R.M.S acceleration spectra at measurement points (a) bogie, (b) floor, (c) window and (d) seat with Carriage C6.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
491
Fig. 11. Coherence functions among the accelerations of the bogie, floor, window and seat: (a) bogie and window; (b) bogie and floor; (c) bogie and seat; (d)floor and seat.
Fig. 12. Autopower spectra for (a) window and (b) floor after partial coherence analysis.
The main energy of window is below 250 Hz and floor vibrates in the range of 0–3000 Hz. The frequency domain of floor vibration is wider than that of the window. This denotes that other vibration sources have little influence on the window. The window vibration is mainly caused by the bogie vibration. The interior noise depends not only on the vibration level of structures but also on the radiation efficiency. The noise sources radiate different levels of noise at the measuring points inside carriage. The frequency spectral curves of noise at six measuring points also reflect characteristics of major noise sources. In order to recognize the major noise sources, coherence analysis between near-field sound pressures at points of H1 and I1 and vibrations of bogie, window and floor are conducted,
492
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
Fig. 13. Coherence functions between the acceleration of floor and window and noises at two locations: (a) window and measurement point H1; (b) window and measurement point B; (c) floor and measurement point I1; (d) floor and measurement point B; (e) bogie and measurement point I1; (f) bogie and measurement point B.
as presented in Fig. 13. The coherence values of far-field between noise at evaluation point B and vibration of the bogie, floor and window are also shown in Fig. 13. The coherence coefficients of noise near the window are greater than 0.6 in the main frequency range of 20–700 Hz and around 1900 Hz. It is shown in Fig. 13(a) that noise near the window is relevant to the window vibration. It is noted in Fig. 13(c) that Floor vibration and near-field noise have peaks around 500 Hz and 1500 Hz. As shown in Fig. 13(a) and (c), the coherence value between floor vibration and noise at measurement point I1 is rather higher in the main frequency domain than that between window vibration and evaluation point H1. The vibration magnitude of floor is greater in the frequency range of 20–200 Hz than that window. Noise near floor surface is found to be radiated by floor vibration. As shown in Fig. 6, energy distribution of noise at measurement point B is similar to energy distribution of floor vibration. It can be observed in Fig. 13(b) and (d) that the coherence value between floor vibration and noise at point B is higher in the main frequency range and above 1500 Hz than that between window vibration and noise at point B. The result indicates that the floor vibration is the major far-field radiating source of measurement point B.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
493
4. Conclusions Measurements of sound intensity, sound pressure level and acceleration are carried out on four carriages on a high speed train. It is found that the energy distribution of noise is dominant in low and middle frequency domain. The low frequency components in the range of 50–160 Hz are of paramount importance. As running speed goes up, sound pressure levels within carriages and rolling noise increase. The car body structure can prevent the middle and high frequency rolling noise from transmitting into the carriage. The sound intensity levels are higher near the floor than near the window in the principal frequency range at three speeds. Analysis results of sound intensity and partial coherence show that the car body floor is the main noise source inside the carriage. The floor vibration on this type of high speed train should be damped to reduce the interior noise. Acknowledgments This work was supported by the National Hi-tech Research and Development Program of China (No. 2003AA333140) and the Programme of Introducing Talents of Discipline to Universities (No. B06012). References [1] Y. Moritoh, Y. Zenda, K. Nagakura, Noise control of high speed shinkansen, J. Sound Vib. 193 (1996) 319–334. [2] A.R. Mohanty, B.D. St. Pierre, P.S. Narayanasami, Structure-borne noise reduction in a truck cab interior using numerical techniques, Appl. Acoust. 59 (2000) 1–17. [3] W.P. Ding, H.L. Chen, Research on the interior noise contributed from a local panel0 s vibration of an elastic thin-walled cavity, Appl. Acoust. 63 (2002) 95–102. [4] Z.S. Liu, H.P. Lee, C. Lu, Passive and active interior noise control of box structures using the structural intensity method, Appl. Acoust. 67 (2006) 112–134. [5] X. Liang, P. Zhu, Z. Lin, The acoustic analysis of lightweight auto-body based on finite element method and boundary element method, J. Shanghai Jiaotong Univ. 40 (2006) 177–180. [6] A. Nordborg, J. Wedemann, L. Willenbrink, Optimum Array Microphone Configuration, Internoise, 4, Nice, France, 2000, 2474–2478. [7] S. Zhang, Noise mechanism, sound source localization and noise control of 350 km/h high-speed train, China Railw. Sci. 30 (2009) 86–90. [8] M. Kawahara, H. Hotta, M. Hiroe, Source Identification of High Speed Train by Sound Intensity, WCRR, Firenze, Italy, 1997, 1057–1060. [9] T. Okada., A study on predicting Shinkansen noise levels using the sound intensity method, JSME Int. J. 47 (2004) 508–511. [10] Y. Liu, X.P. Zhang, Application of sound intensity theory in noise measurement and analysis of high speed railway vehicles, J. China Railw. Soc. 25 (2003) 45–48. [11] J. Zhang, B. Han, Analysis of engine front noise using sound intensity techniques, Mech. Syst. Signal Process. 19 (2005) 213–221. [12] LMS International, Transfer Path Analysis, the Qualification and Quantification of Vibro-acoustic Transfer Paths, 1998. [13] MSX International, I-DEAS Noise Path Analysis, MTS System Corporation, 1998. [14] Q. Leclere, C. Pezerat, B. Laulagnet, L. Polac, Application of multi-channel spectral analysis to identify the source of a noise amplitude modulation in a diesel engine operating at idle, Appl. Acoust. 66 (2005) 779–798. [15] M.H. Fouladi, M.J. Mohd. Nor, A.K. Ariffin, Spectral analysis methods for vehicle interior vibro-acoustics identification, Mech. Syst. Signal Process. 23 (2009) 489–500. [16] B.K. Bae, K.J. Kim, A Hilbert transform approach in source identification via multiple-input single-output modeling for correlated inputs, Mech. Syst. Signal Process. 12 (4) (1998) 501–C513. [17] National Standard of the People0 s Republic of China GB/T 12816-2006, The Limiting Value and Measurement Method for the Interior Noise in the Railway Passenger Coach, 2006.
Contents lists available at ScienceDirect
Mechanical Systems and Signal Processing journal homepage: www.elsevier.com/locate/ymssp
Application of sound intensity and partial coherence to identify interior noise sources on the high speed train Rongping Fan a,b, Zhongqing Su b, Guang Meng a,n, Caichun He c a b c
State Key Laboratory of Mechanical System and Vibration, Shanghai Jiao Tong University, Shanghai 200240, China Department of Mechanical Engineering, The Hong Kong Polytechnic University, Hung Kom, Hong Kong Zhuzhou Times New Materials Technology Co., Ltd., Zhuzhou 412001, China
a r t i c l e i n f o
abstract
Article history: Received 4 May 2011 Received in revised form 26 November 2013 Accepted 29 November 2013 Available online 14 February 2014
In order to provide a quieter riding environment for passengers, sound quality refinement of rail vehicle is a hot issue. Identification of interior noise sources is the prerequisite condition to reduce the interior noise on high speed train. By considering contribution of noise sources such as rolling noise, mechanical equipment noise, structure-borne noise radiated by car body vibration to the interior noise, the synthesized measurement of sound intensity, sound pressure levels and vibration have been carried out in four different carriages on high speed train. The sound intensity and partial coherence methods have been used to identify the most significant interior noise sources. The statistical analysis results of sound intensity near window and floor on four carriages indicate that sound intensity near floor is higher than that near window at three traveling speeds. Ordinary and partial coherent analysis of vibro-acoustical signals show that the major internal noise source is structural-borne sound radiated by floor vibration. These findings can be utilized to facilitate the reduction of interior noise in the future. & 2013 Elsevier Ltd. All rights reserved.
Keywords: Interior noise Sound intensity High speed train Partial coherence
1. Introduction The noises within the high speed train are caused by many noise sources, such as rolling noise, mechanical equipment noise, air conditioning noise, structural-borne noise radiated by car body vibration and aerodynamic noise. With traveling speed increase and vehicle weight-lighting, the rolling noise and aerodynamic noise of high speed trains are greatly increased. In order to provide quieter riding environments for passengers, more attention [1] has been paid to reducing the steeply increased interior noise and vibration. The improvement in sound and vibration quality can be achieved in three aspects: source, transfer path, and receiver. It is the subject of this paper to identify the noise sources inside the high speed train in order to reduce the noise levels. The finite element and boundary element method were used by Mohanty et al. [2] to distinct the interior acoustic field by the natural modes and shapes of cabs. Ding [3] developed the finite element method to evaluate panel acoustic contribution. Liu [4] analyzed the panel vibration and interior noise of vehicle cavity by structure strength. Liang [5] investigated panel acoustic contribution at the specific mode to identify the major sound source based on boundary element method and finite element method. Vehicle structures were complex, so the numerical models had to be significantly simplified. However, the correction of numerical results was highly dependent on accuracy of the models.
n
Corresponding author. Tel.: þ86 2134206332x816; fax: þ 86 21 34206006. E-mail address: [email protected] (G. Meng).
0888-3270/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ymssp.2013.11.014
482
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
The acoustic arrays, sound pressure level and sound intensity have been measured to identify the main sources in the rail vehicles. Microphone arrays or combination of different sensors have proved useful to identify and predict the noise levels and location of major sources outside high speed trains [6–10]. Sound intensity measurement can offer both direction and position of sound sources. Sound intensity theory was used by Liu [10] to measure noise on high speed railway vehicle and analyzes the interior noise distribution. Sound intensity techniques were applied to identify engine front noise sources [11]. Although the force vector method and full matrix method were commonly used for noise path analysis [12,13], these two methods needed extra time and cost to deal with the problem of the connection between the sound and receiver. Spectral analysis method of partial coherence analysis [14,15] was used to identify vibro-acoustical sources of noise emitted inside vehicle compartments. The first step of signal conditioning in the partial coherence method is to rank the sources. Hilbert transform was used to check causality between signals by Fouladi [15] and Bee [16]. In this paper, the sound intensity is used not only to investigate the characteristics of sources but also to rank the potential sources. In this paper, acoustic-vibrational measurements are carried out and sound intensity and partial coherent methods are used to distinguish major interior noise sources inside a specific high speed train. The paper is organized as follows: firstly, a short description of the sound intensity theory along with partial coherence is given; in Section 2, the methodology of sound and vibration measurement within high speed train is described in details; in Section 3, the measurement data will be processed to identify the main sources of interior noise on carriage; finally, the conclusion of noise source identification will be made and the method to reduce the interior noise will be suggested. 1.1. Partial coherence analysis For the system with multi-input Xi and single output Y, input signals are correlated. Partial coherence analysis can rule out linear effects of other inputs from problem by taking advantage of conditional output method. Coherence function is an indication of linear relationship between input X and output Y for a system. Ordinary coherence analysis often confuses the contribution of each input Xi to the output. Partial coherence technique has been applied to find the contribution of sources to overall response in the vehicle interior vibrio-acoustic problems [14,15]. Partial coherent method can remove linear effects of other inputs and make a set of inputs irrelevant. For example, consider X ii 1 and X ij 1 as two partially correlated inputs. The conditioned signal X ij 1 from signal X ii 1 can be obtained using Eq. (1) X ij ¼ X ij 1 H iij 1 X ii 1 ; where
H iij 1
ð1Þ
is the frequency response function between the two signals
H iij 1 ¼
Giij 1 Giii 1
X ij 1
and
X ii 1
;
ð2Þ
where Giii 1 is autospectrum of signal X ii 1 and Giij 1 is cross-spectrum of signals X ij 1 and X ii 1 . The two sides of Eq. (1) are multiplied by conjugate X nr . The conditional power spectrum can be given by Girj ¼ Girj 1 H irj 1 Giir 1 ; Girj 1 Gijj of
Girj ,
where spectrum
and Giir 1 are signal X ij can
ð3Þ the cross spectra. If we replace j with r and insert Eq. (2) into Eq. (3), conditional autopower be obtained by
Gijj ¼ Gijj 1 ð1 r iij 1 Þ;
ð4Þ
where Gijj 1 is conditional autopower spectrum of signal X ij 1 and partial coherence value r iij 1 is defined by r iij 1 ¼
ðGiij 1 Þ2 Giii 1 Gijj 1
:
ð5Þ
X ij will rule out the linear effects of the stronger signal X ii 1 from signal X ij 1 . Then the effect of each input on output is identified by conditional output power spectrum. 1.2. Sound intensity theory Sound intensity is defined as the sound power per unit area. The sound source contribution can be determined by the sound intensity [13]. The time-averaged sound intensity I at a point A in a sound field can be expressed as Z Z 1 1 Ii dt ¼ I dt ¼ 〈pv〉t ð6Þ I¼ T T T T i where the instantaneous sound intensity Ii is the product of sound pressure p and particle velocity v, and T is the average time. Particle velocity v is related to sound pressure by the following equation: Z 1 t v¼ ∇pdτ ð7Þ ρ0 1
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
Substitution of Eq. (7) into Eq. (6) yields Z t 1 p ∇pdτ I¼ ρ0 1 t Sound intensity compound in a specific direction l is expressed as Z t 1 ∂p dτ p Il ¼ ρ0 1 ∂l t
483
ð8Þ
ð9Þ
Sound pressure gradient in the direction l is evaluated in terms of sound pressures at two adjacent points to point A ∂p p p2 1 ∂l Δr
ð10Þ
where Δr is the distance between two measured points and the point A is equally far away from these two points. The sound pressure at the midpoint A of the two microphones can be obtained by taking the arithmetic average of two sound pressure levels p
p1 þ p2 2
Placing Eqs. (10) and (11) into Eq. (9), we get Z t 1 p1 þ p2 ðp2 p1 Þdτ Il ¼ 2ρ0 Δr 1 t
ð11Þ
ð12Þ
When sound pressure is harmonic, sound intensity can be expressed as Il ¼
P A1 P A2 sin ðφ2 φ1 Þ 2ρ0 ωΔr
ð13Þ
where P A1 and P A2 are the magnitudes of sound pressure signals p1 and p2 , respectively. When the phase difference Δr between two microphones is very small, the sound intensity can be simplified as Il
P A1 P A2 ðφ2 φ1 Þ 2ρ0 ωΔr
ð14Þ
Through Fourier transform, averaged sound intensity in frequency domain can be expressed as IðωÞ ¼
ImfG12 g ρ0 ωΔr
ð15Þ
where G12 is the one-sided cross-power spectrum density function between two microphones, ρ0 is the air density and ω is the angular frequency. Compared to the partial coherence method, it is quicker and more direct to determine the radiated power of sound sources by sound intensity. 2. Methodology The main objective of this paper is to identify predominant sources of interior noise from the potential sources of noise emissions. There are many sources that influence the interior noise inside the high speed trains. Potential sources are classified into three categories: ▪ The interior noise sources like ventilation or air-conditioning system. ▪ Exterior noise sources like wheel–rail interaction, the propulsion, brakes, compressor and aerodynamics. ▪ Structural-borne noise radiated by vibrating car body. It is difficult to recognize the contribution of each noise source to the interior noise. The influence of interior noise sources like ventilation or air-conditioning system on the interior noise can be measured when the high speed train is still. In this paper, the influence of ventilation or air conditioning is not considered by closing the ventilation and air conditioning when the high speed train was traveling. Sound reaches the carriage interior through two different routes: the air-borne path and the structure-borne path. In the air-borne path, sound travels away from sound sources and transmits through the floor, wall and windows of carriage into the interior. In the structure-borne path, exterior vibration sources such as wheel– rail interaction excite the car body of carriages to vibrate and radiate structural sound into the interior. Because the running speed of high speed train does not exceed 300 km/h, aerodynamic noise is not a major source of noise emission outside the carriage. The acoustic insulation materials were installed on the wall of the carriage. It was easier for the noise to transmit from the window into the interior carriage than from the wall. The floor was connected through the second suspension system of the bogie and excited to vibrate. First of all the window and the floor are considered as potential sources of noise emissions in the test high speed train.
484
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
The high speed train traveled at a maximum running speed of 250 km/h. The layout of carriages of the high speed train is demonstrated in Fig. 1. The test high speed train was made up of two electric locomotives plus nine trailer cars, which were manufactured by two different vehicle factories. The air conditioning system and bogie equipment on carriages made by two factories were different. Measurement of noise and vibration was carried out in the first and second class carriages. The furniture inside the first and second class carriages was identical except that cable TVs were installed inside the first class carriages. These four carriages were the first class carriage C2, the dining carriage C5 and two second class carriages C4 and C6. Common characteristics of the interior noise within the high speed train could be obtained through the measurements in the four carriages. The major interior noise sources could then be identified.
Fig. 1. Schematic diagram of carriage distribution.
Fig. 2. Setup of vibration and noise measurement within each carriage in the running test
Fig. 3. Setup of acceleration transducer and microphones near bogie.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
485
To identify the major interior noise sources when the high speed train was traveling, this paper will focus on the running test. The air conditioning equipment is closed to eliminate the effect of air conditioning noise. The test train traveled at speeds of 120, 160 and 200 km/h on high speed straight level track free of rail joints. To discern the contribution of each source to the overall interior noise and the path way that noise propagates into the carriage interior, acoustical–vibrational measurements were simultaneously carried out. These measurements included the interior and exterior sound pressure levels, sound intensity, and the structural vibration of car body and bogie. The experimental setup in each one of the four carriages was identical. Six microphones denoted by six consecutive characters from A to E, referred to the standard GB/T 12816-2006 [17], were located in the carriage, as shown in Fig. 2(a). Three microphones A, B and C for seated persons were at a height of 1.2 m above the car body floor. The other three microphones D, E and F for standing persons were 1.6 m away from the floor. The microphones A and D are positioned at one end of the carriage. The microphones B and E are located in the middle of the carriage. The microphones C and F were placed in the middle of the gangway at one end of the carriage outside the glass door. Microphones H1 and H2, as shown in Fig. 2(b), were used to measure sound intensity near window. Microphones I1 and I2, as shown in Fig. 2(a) and (c), were used to measure sound intensity above the floor. In Fig. 3, microphone G was used to measure rolling noise. The wind cap around the microphone G was to prevent the extra pressure caused by the aerodynamics near the bogie. Accelerometer 1 was used to measure the vibration on the center of the bogie. Micro-accelerometer 2 attached to the window glass was to measure window vibration, as shown in Fig. 2(a) and (b). Accelerometer 3, as shown in Fig. 2(a) and (c), was vertically located on the car body floor to measure the car body floor vibration. Accelerometer 4 on the seat in the middle of the car body was used to measure seat vibration, as given in Fig. 2(a). All the vibrational and acoustical signals inside the carriage along with rolling noise were simultaneously obtained. The data were collected by a Sony data recorder. The time–frequency analysis results of sound and vibration showed that interior noise and vibration were stationary. Linear averaging used for analysis of stationary signals was applied to equalize all spectra and time records. Hanning weighting with overlap of 50% was used as a general-purpose window for continuous signals. Frequency resolution was 1.27 Hz and 50 spectral averages were implemented for analysis. 3. Results and observations Before both sound intensity and partial coherence methods are used for post-processing of the measured signals; the measured noise within four carriages is analyzed to characterize the spatial distribution of noise inside the carriage.
Fig. 4. Overall values of linear sound pressure levels at six points within four carriages: (a) 160 km/h and (b) 200 km/h.
486
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
The second step is to process the sound intensity signals near sources at different running speeds. The sources can be ranked by the level of sound intensity. The partial coherent power spectra of sources will be calculated according to the partial coherence theory. The coherence analysis between near-field and far field noise and vibration of the structure will be performed. 3.1. Analysis of sound pressure levels Although the acoustic environment within carriages was very complex and the measured carriages were a little different, remarkable characteristics of the interior noise within carriages were obtained. The acoustic spatial distribution within carriages was evaluated by noise levels at six microphones from A to E. The overall value of the sound pressure levels at six points at two speeds of 160 km/h and 200 km/h is illustrated in Fig. 4. As shown in Fig. 4, the overall value of sound pressure levels within carriages shows that the sound pressure levels at points A and B at a height of 1.2 m from the floor were about 10 dB higher than that of locations D and E at a height of 1.6 m. The sound pressure levels at C and F inside the gangway between two carriages are almost identical. The difference in the sound pressure levels inside the carriage and the gangway
Fig. 5. Linear sound pressure levels in one third octave band center frequency at three speeds: (a) 120 km/h, (b) 160 km/h, and (c) 200km/h.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
487
indicates that the sound environments of the carriage and the gangway are different from each other. The sound pressure levels at 200 km/h are higher than those at 160 km/h. It can be observed that there are similar trends in the spatial distribution of noise at the measured points. The sound pressure levels in the one-third octave band center frequency at three speeds of 120 km/h, 160 km/h, and 200 km/h are presented in Fig. 5. It can be seen in Fig. 5 that the low frequency components are dominant at three running speeds. Regardless of the level of frequency components, the shapes of frequency spectral curves of sound pressure levels at evaluation point B are reprehensive of the interior noise. To indicate the effect of rolling noise on interior noise, Fig. 6 gives sound pressure level at location B on carriage C6 and rolling noise at three traveling speeds of 120 km/h, 160 km/h and 200 km/h. It is clear in Fig. 6 that running speed has different influence on different frequency components of interior noise and rolling noise. The low-frequency components are dominant in the interior noise. With increase in traveling speed, the noise in the range of 25–4000 Hz at location B goes up. Since the running speed of the high speed train is below 300 km/h, rolling noise radiated by vibration of wheel and rail was the major external noise source. It can been observed from Fig. 6 that rolling noises at three traveling speeds are wide band noise in the span of 20–8000 Hz. The main energy is distributed in the frequency range of 20–5000 Hz. Rolling noise spectra have multi-peaks which are relevant to resonance of wheel and rail. By comparison of the interior noises at location B and rolling noises at three traveling speeds, it is obvious that car body can prevent rolling noise at middle and high frequencies from transmitting into carriages. The rolling noise is not the major noise source at low frequency within carriages.
Fig. 6. Linear sound pressure levels at (a) measurement point B and (b) rolling noise near wheel and rail on carriage C6. Table 1 Sound intensity levels (dB) near the floor and window within four carriages at three speeds. Running speed (km/h)
120 160 200
Carriage C2
Carriage C4
Carriage C5
Carriage C6
Window
Floor
Window
Floor
Window
Floor
Window
Floor
93.7 95.3 101.1
105.1 108.3 114.1
95.1 96 98.7
107.9 109.2 112
93.5 95.5 97.9
104.6 108.9 110.7
96.2 97.5 99.3
107.1 110.9 112.8
488
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
3.2. Post-processing by the sound intensity method It is obvious in Figs. 5 and 6 that interior noise energy is dominant in the low and middle frequencies. The major purpose of the interior noise source identification in this paper is to detect the potential sources causing the interior noise at low and middle frequencies. The low and middle frequency noises near the window and floor are made up of radiated noise by vibration of window and floor, as well as air noise that rolling noise transmits through window and floor into the interior. When the high speed train travels, aerodynamics cause windows to vibrate and radiate noise. Sound intensity near window and floor surfaces at three speeds is calculated using Eq. (15). Table 1 lists overall sound intensity levels near surfaces of window and floor within four carriages at three running speeds. It is clear that overall sound intensity levels near floor surface are higher than those near window within four carriages at the same running speed. As running speed increases, the sound intensity levels near window and floor rise. The sound intensity levels in the one-third octave band center frequency within carriage C6 at three running speeds are shown in Fig. 7. These curves show where overall values of sound intensity near window and floor differ. The intensity levels near window surface have the predominant components below 100 Hz. The low frequency components up to 160 Hz are dominant in the sound intensity levels near the floor. As traveling speed increases, the sound intensity levels near window and floor increase in the entire frequency range. It is quite clear that the components of sound intensity in the one-third octave center frequency range of 20–200 Hz near floor are higher than those near window. At different traveling speeds of the high speed train, the levels and frequency spectra of interior noise signals and sound intensity vary within different carriages. Sound sources make different contribution to the interior noise. However, the general trends of interior noise signals and sound intensity inside four carriages at three running speeds are very similar. Therefore, it is assumed that the interior noise and sound intensity measured within carriage C6 is reprehensive of the interior noise signals inside the high speed train at different running speeds. In the following analysis, signals measured within carriage C6 at the speed of 200 km/h will be used to identify the interior noise sources.
Fig. 7. The one third octave band spectra of the sound intensity of (a) window and (b) floor within carriage C6.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
489
To rank the noise sources of the floor and window, the linear sound pressure levels at measurement points H1 and I1 are shown in Fig. 8. It is clear that the main noise energy near window and floor is in the range of 20–200 Hz. The sound pressure levels above 200 Hz near the floor are higher than those near the window. For better understanding of the effects of sound sources, the sound intensity spectra near surfaces of window and floor in the narrowband frequency are illustrated in Fig. 9. It can be seen that the main components of sound intensity levels radiated by these two regions are in the low frequency range of 20–200 Hz. From the above analysis, it can be concluded that sound power is larger near the floor than near the windows.
Fig. 8. Linear sound pressure levels at (a) measurement point H1 near window and floor and (b) measurement point I1 near floor.
Fig. 9. The spectra of sound intensity radiated by window and floor within carriage C6: (a) window and (b) floor.
490
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
3.3. Application of partial coherence method to identify main sources of vibration and noise Analysis results of sound intensity show that noise near the floor makes major contribution to the interior noise at the measuring points within carriage. Vibration of the floor and window is mainly transmitted by the vibration of the bogie and mechanical equipment. The rail–wheel vibration is transmitted to carriage floor through primary suspension and bogie. The noise near the floor and window consists of structural-borne noise radiated by vibration of floor and window and airborne noise transmitted from rolling noise. The ordinary coherent analysis and partial coherent analysis of carriage structural vibration and near-field and far-field noise can show whether the noise near the floor and window is structural-radiated noise or rolling noise. Fig. 10 illustrates vibration frequency spectra of floor, window, seat and bogie of carriage C6 at the speed of 200 km/h. It can be seen in Fig. 10 that acceleration of bogie is in the wide frequency band range of 0–5000 Hz. The first two peaks of bogie vibration are at 57.2 Hz and 1.71 k Hz. The major energy of floor vibration is distributed within 0–3000 Hz. The first three peaks of vibration are at 76.3 Hz, 91.6 Hz, and 95.4 Hz. The vibration of the window is centered below 1000 Hz and there is a low peak around 2000 Hz. The seat vibration is below the low frequency of 100 Hz. The coherence analysis between the vibration of bogie, floor, and windows is carried out in order to check the relation between the vibration of the floor, the window, and the bogie. The coherence functions between bogie, floor and window are illustrated in Fig. 11. From both Figs. 10 and 11, it can be seen that the coherence coefficient of window and bogie is greater than 0.6 below 800 Hz where the vibration magnitudes of bogie and window are dominated. The window vibration is linearly related to bogie vibration. The coherence coefficient between bogie and floor is approaching 0.8 at main frequencies of floor vibration below 3500 Hz. The coherent values between seat and floor as well as seat and bogie are almost 0.8 at the low frequency of 50.9 Hz. It indicates that bogie vibration is the main source to excite vibrations of the floor and seat. Although aerodynamic noise is not considered as the air noise that transmits through window into interior under traveling speed of 300 km/h, vibration of window surface is possibly excited by bogie vibration and aerodynamics at the high speed of 200 km/h. Because the level of the rolling noise is quite high, it is possible that acoustical excitation of rolling noise results in floor vibration. The effect of aerodynamics and acoustical excitation of rolling sources as well as other excitation sources on the vibration of the floor and window can be obtained by application of partial coherence analysis to remove the effect of bogie vibration on the floor and window inside the carriage. Autopower spectra of partial coherence for the window and floor are shown in Fig. 12. Compared to accelerations of floor and window shown in Fig. 10, the frequency span of vibration of window and floor become narrower and vibration spectral magnitudes become tremendously low.
Fig. 10. R.M.S acceleration spectra at measurement points (a) bogie, (b) floor, (c) window and (d) seat with Carriage C6.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
491
Fig. 11. Coherence functions among the accelerations of the bogie, floor, window and seat: (a) bogie and window; (b) bogie and floor; (c) bogie and seat; (d)floor and seat.
Fig. 12. Autopower spectra for (a) window and (b) floor after partial coherence analysis.
The main energy of window is below 250 Hz and floor vibrates in the range of 0–3000 Hz. The frequency domain of floor vibration is wider than that of the window. This denotes that other vibration sources have little influence on the window. The window vibration is mainly caused by the bogie vibration. The interior noise depends not only on the vibration level of structures but also on the radiation efficiency. The noise sources radiate different levels of noise at the measuring points inside carriage. The frequency spectral curves of noise at six measuring points also reflect characteristics of major noise sources. In order to recognize the major noise sources, coherence analysis between near-field sound pressures at points of H1 and I1 and vibrations of bogie, window and floor are conducted,
492
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
Fig. 13. Coherence functions between the acceleration of floor and window and noises at two locations: (a) window and measurement point H1; (b) window and measurement point B; (c) floor and measurement point I1; (d) floor and measurement point B; (e) bogie and measurement point I1; (f) bogie and measurement point B.
as presented in Fig. 13. The coherence values of far-field between noise at evaluation point B and vibration of the bogie, floor and window are also shown in Fig. 13. The coherence coefficients of noise near the window are greater than 0.6 in the main frequency range of 20–700 Hz and around 1900 Hz. It is shown in Fig. 13(a) that noise near the window is relevant to the window vibration. It is noted in Fig. 13(c) that Floor vibration and near-field noise have peaks around 500 Hz and 1500 Hz. As shown in Fig. 13(a) and (c), the coherence value between floor vibration and noise at measurement point I1 is rather higher in the main frequency domain than that between window vibration and evaluation point H1. The vibration magnitude of floor is greater in the frequency range of 20–200 Hz than that window. Noise near floor surface is found to be radiated by floor vibration. As shown in Fig. 6, energy distribution of noise at measurement point B is similar to energy distribution of floor vibration. It can be observed in Fig. 13(b) and (d) that the coherence value between floor vibration and noise at point B is higher in the main frequency range and above 1500 Hz than that between window vibration and noise at point B. The result indicates that the floor vibration is the major far-field radiating source of measurement point B.
R. Fan et al. / Mechanical Systems and Signal Processing 46 (2014) 481–493
493
4. Conclusions Measurements of sound intensity, sound pressure level and acceleration are carried out on four carriages on a high speed train. It is found that the energy distribution of noise is dominant in low and middle frequency domain. The low frequency components in the range of 50–160 Hz are of paramount importance. As running speed goes up, sound pressure levels within carriages and rolling noise increase. The car body structure can prevent the middle and high frequency rolling noise from transmitting into the carriage. The sound intensity levels are higher near the floor than near the window in the principal frequency range at three speeds. Analysis results of sound intensity and partial coherence show that the car body floor is the main noise source inside the carriage. The floor vibration on this type of high speed train should be damped to reduce the interior noise. Acknowledgments This work was supported by the National Hi-tech Research and Development Program of China (No. 2003AA333140) and the Programme of Introducing Talents of Discipline to Universities (No. B06012). References [1] Y. Moritoh, Y. Zenda, K. Nagakura, Noise control of high speed shinkansen, J. Sound Vib. 193 (1996) 319–334. [2] A.R. Mohanty, B.D. St. Pierre, P.S. Narayanasami, Structure-borne noise reduction in a truck cab interior using numerical techniques, Appl. Acoust. 59 (2000) 1–17. [3] W.P. Ding, H.L. Chen, Research on the interior noise contributed from a local panel0 s vibration of an elastic thin-walled cavity, Appl. Acoust. 63 (2002) 95–102. [4] Z.S. Liu, H.P. Lee, C. Lu, Passive and active interior noise control of box structures using the structural intensity method, Appl. Acoust. 67 (2006) 112–134. [5] X. Liang, P. Zhu, Z. Lin, The acoustic analysis of lightweight auto-body based on finite element method and boundary element method, J. Shanghai Jiaotong Univ. 40 (2006) 177–180. [6] A. Nordborg, J. Wedemann, L. Willenbrink, Optimum Array Microphone Configuration, Internoise, 4, Nice, France, 2000, 2474–2478. [7] S. Zhang, Noise mechanism, sound source localization and noise control of 350 km/h high-speed train, China Railw. Sci. 30 (2009) 86–90. [8] M. Kawahara, H. Hotta, M. Hiroe, Source Identification of High Speed Train by Sound Intensity, WCRR, Firenze, Italy, 1997, 1057–1060. [9] T. Okada., A study on predicting Shinkansen noise levels using the sound intensity method, JSME Int. J. 47 (2004) 508–511. [10] Y. Liu, X.P. Zhang, Application of sound intensity theory in noise measurement and analysis of high speed railway vehicles, J. China Railw. Soc. 25 (2003) 45–48. [11] J. Zhang, B. Han, Analysis of engine front noise using sound intensity techniques, Mech. Syst. Signal Process. 19 (2005) 213–221. [12] LMS International, Transfer Path Analysis, the Qualification and Quantification of Vibro-acoustic Transfer Paths, 1998. [13] MSX International, I-DEAS Noise Path Analysis, MTS System Corporation, 1998. [14] Q. Leclere, C. Pezerat, B. Laulagnet, L. Polac, Application of multi-channel spectral analysis to identify the source of a noise amplitude modulation in a diesel engine operating at idle, Appl. Acoust. 66 (2005) 779–798. [15] M.H. Fouladi, M.J. Mohd. Nor, A.K. Ariffin, Spectral analysis methods for vehicle interior vibro-acoustics identification, Mech. Syst. Signal Process. 23 (2009) 489–500. [16] B.K. Bae, K.J. Kim, A Hilbert transform approach in source identification via multiple-input single-output modeling for correlated inputs, Mech. Syst. Signal Process. 12 (4) (1998) 501–C513. [17] National Standard of the People0 s Republic of China GB/T 12816-2006, The Limiting Value and Measurement Method for the Interior Noise in the Railway Passenger Coach, 2006.