Abstract

This contribution reviews progress in the technology for assessing the silencing performance of piston engine intake and exhaust systems on a running engine, with particular reference to flow noise. The new developments include the application of selective averaging, order tracking and optimised sampling rate methods to clearly identify and quantify the small fraction of the total fluctuating wave energy that is being propagated across each discontinuity along the flow path through a highly reactive flow duct system. Such measurements then quantify the interaction between the aeroacoustic sources of excitation and the associated local acoustic characteristics that govern the generation, transfer, propagation and emission of wave energy to the environment. INTRODUCTION Intake and exhaust systems are designed to reduce engine breathing noise emissions to levels that comply with legislation, provide a sound quality and internal vehicle climate that meets customer expectations, while maintaining optimum fuel efficiency and vehicle performance. This diversity of function, coupled with operational, system layout and space allocation constraints, together with the benefits of rapid prototyping, underline the practical advantages of a design methodology that is firmly based on realistic predictive numerical modelling [1] of acoustic and operational performance. Appropriate existing software [2] can adequately describe the passive acoustic and resonant behaviour of geometrically complex flow duct systems, but does not yet include any influence of flow noise sources on predicted acoustic performance. This deficiency is of practical importance, since flow generated noise is commonly the major contributor [1][2][3] at higher engine speeds to IC engine breathing noise spectra above 200 Hz. FLOW NOISE AND ITS MEASUREMENT The sustained excitation of a tuned resonator by shed vorticity in a separating shear layer [4] has been exploited empirically for making musical and other sounds from time immemorial. In the normally highly acoustically reactive intake and exhaust systems such mechanisms [1][2][3][4] are seen to generate both coherent and broadband sound, or noise, by a flow driven generator that is highly nonlinear, exciting a resonant system with effectively linear acoustic behaviour. Similar mechanisms [1, 2] can also cause selective amplification of incident sound. The dominant aeroacoustic mechanism of concern here [4] is the action of a fluctuating Coriolis force working on the fluctuating flow. However, with some specific exceptions [1, 4], the relevant controlling features of the flow cannot be specified significantly explicitly, due to current ignorance of many essential details of the associated vortical or turbulent fluid motion. Thus one must normally rely on measurement to establish quantitative descriptions of aeroacoustic sources in relation to the associated boundary geometry, acoustic climate and fluid motion. Intake and exhaust system geometry consists essentially of silencing and other component elements connected in sequence by lengths of uniform pipe. Wave reflections at the junctions, combined with any local sound generation, produces an acoustic climate comprised of both standing and progressive waves. Superimposed on the time averaged flow, the unsteady fluid motion includes both the acoustically related potential fluctuations combined with vortical disturbances cast off with the separating shear layers [1, 4] or generated at the boundaries, that travel with the mean flow. New robust procedures [2-4] using cross-power spectral analysis of the signals from pairs of flush wall mounted pressure transducers were used to quantify the incident and reflected wave amplitude spectra at appropriate positions along the flow path. The application of selective averaging, order tracking SESSIONS and optimised sampling rate methods [2][3][4] provided sufficient precision to identify clearly the small fraction of the total fluctuating wave energy that was being propagated through the system. The location, strength and spectral structure of any flow noise sources/sinks was then derived from the small gains/losses in power flux at the relevant sites. Pilot experiments [2, 4] revealed that the flow generated sound spectra were always closely related to the system acoustic resonances. Earlier experimental studies of flow generated sound [1, Similarly complex nonlinear behaviour was observed with other typical exhaust system component geometries. Measurements were also made with strongly excited expansion chambers [2, 3], either on the test bench, or in the exhaust line of a running engine accelerating to full speed on a test bed with wide open throttle. The observations provided clear evidence of reverberant amplification or attenuation of the incident sound at specific frequencies, although the relationship with the associated boundary geometry, acoustic climate and fluid motion has not been established yet. Otherwise, the relative observed spectral behaviour corresponded to that predicted by one dimensional linear acoustic models. This was in spite of the fact that the observed amplitude of some spectral components of the cyclic excitation exceeded 3.5 kPa, corresponding to 165 dB. With purely flow excited systems on the test bench [2, 4] the maximum sound pressure levels recorded were some 20 dB lower than this, although the corresponding overall fluctuating pressure levels were close to 160 dB. Otherwise, acoustic characteristics calculated with linear acoustic models were found to be in good agreement with measurements, indicating that wave steepening was not yet a significant factor at the conditions of these experiments. DISCUSSION Pilot studies [2][3][4] have established and validated robust and sufficiently precise experimental technology for establishing the position strength and spectral characteristics of flow noise sources in relation to the boundary geometry, the mass flow and the associated acoustic climate. This includes the influence of regenerative amplification by acoustic feedback [2, 3] INTRODUCTION According to the improvement of automobile engine to provide high speed and high power, the amount of working gas increases. When a high flow rate of gas exists, a large flow induced noise radiates from the open end of the muffler. This flow induced noise sometimes sacrifices the noise reduction effect of muffler predicted through the acoustic theory. In the previous papers [1,2], the flow induced noise generated from the simple cavity type and the inserted type mufflers with steady flow was experimentally studied. In this paper, the characteristics of flow noise generated from simple expansion cavity type muffler with pulsating flow were studied. The noise sources inside the muffler were also discussed in comparison of the case with steady flow. EXPERIMENT AND DISCUSSION The experimental set up is illustrated in Extraction of Flow Induced Noise from Total Radiated Noise of Muffler In the experiments for the pulsating flow induced noise, the radiated noise contains not only the flow noise but also the mechanical noise from pulsation generator. It is necessary to extract the noise generated by pulsating flow from the total noise of muffler. Then, the noise generation model of muffler was introduced. The noise power radiated from the open end of the muffler W m is recognized as a sum of two kinds of noise power. The noise carried into a muffler from the upstream side is attenuated by the filter effect of muffler and carried out. This noise power can be evaluated as W p eF, where W p is the noise power radiated from the open end of the straight pipe (without muffler), and F is the insertion loss of the muffler. The other one is the noise power induced by the air flow W f inside the muffler, then Effects of Cavity Walls on Flow Induced Noise Two types of noise source can be considered for the flow noise generated inside the muffler. One is the turbulence of the flow and the other is the pressure fluctuation at the solid boundaries. Curle [3] proposed the relationship between these two types of noise source and generated sound pressure. Introducing the adiabatic process, the sound pressure formed at a point SESSIONS apart r from these noise sources is derived as; The first term of this formula expresses the sound pressure component generated by the disturbance of the flow, and the second term expresses the component generated by the fluctuation of the pressure at the solid boundaries such as inner walls of muffler. From the results of order estimation of these two terms, the first term was two orders smaller than the second term in as low Mach number flow as employed in this experiment. Then, the flow induced noise generated inside the muffler has to be considered in the relationship with the second term of Eq When we introduce the steady flow to the muffler, the pressure fluctuation on the side wall becomes small in high frequency range as shown in From these results of the order estimation and the experiments, the following two matters were clarified. Effect of the tabs on the acoustic characteristics of resonance generated in short enlargement muffler (cavity) was examined of 0.3 and 0.5. As the results, tabs are very effective to decrease feedback resonance generated in the muffler. The reduction depends on de-correlate of the azimuth coherence vortices. For an area blockage of tabs to the nozzle exit greater than 3 %, feedback resonance is reduced considerably. INTRODUCTION When the jet vortex hits directly against the training edge of a short enlargement muffler, then the pressure wave is generated, and the wave returns to the leading edge of the muffler. This phenomena is repeated itself many times, then feed-back resonance occurs. To decrease feedback resonance, many techniques inclining the downstream wall of the muffler have been tried 1 . But the problem of these methods is that the muffler structure becomes too long. Tabs have been shown to produce stream-wise vortices on both sides of the tabs. This device has been applied for reducing jet noise. Therefore, it is considered that the application of tabs to the reduction of feedback resonance generated in the enlargement muffler is effective. The present paper examines the effect of tabs on the reduction of feedback resonance. Results and Discussion Schematic diagram of experimental apparatus is shown in The sound pressure spectra of the noise generated aerodynamically in the muffler was measured at a position of R=15cm and q =60º by a 1/4 inch microphone. The sound pressure spectra without a tab showed a lot of the complicated discrete frequencies. The dominant frequency is called cavity tone frequency in the present experiment. The cavity tone frequencies for Mach number 0.5 are shown in 2 ) and the cavity length mode shows a similar frequency, therefore it is difficult to separate these frequencies. Tam & Ahuja 3 stated that the feedback doesn't occur below Mach number of 0.4. But in the present experiment the feedback was produced at Mach number 0.3. These phenomena might depend on the difference of experimental apparatus whether the box is installed or not. Varying the shear layer by fitting tabs to the nozzle, jet vortex structure changes. As the vortex becomes smaller than before fitting the tabs, the hit of the vortex is alleviated. The pressure wave produced by the hit is decreased, and then noise generation is reduced. The effect of tabs on the overall sound pressure level (SPL) is shown in To examine the reason reducing the cavity tone, the coherence function of the velocity fluctuation in the shear layer was measured using 2 hot wires. In this case, the side of the enlargement muffler was removed to do the measurement of coherence easily. As the results, the values of coherence function of the velocity fluctuation with tabs were lower than that without tab. It is found that tabs de-correlate the azimuth coherence vortices. Conclusions Tabs are very effective to decrease the cavity tone generated in the short enlargement muffler. The reduction depends on de-correlation of the azimuth coherence vortices. In this paper, the problem of detecting the acoustic signals contaminated by random background and wind-induced noises is discussed from the statistical viewpoint. For evaluating the effects of wind-induced noise, we will pay our special attention to the measurement of wind velocities which cause the wind-induced noise. By grasping the statistical relationship between wind velocities and wind-induced noise in the form of generalized regression model, a new type wide-sense digital filter for detecting the acoustic signals under wind-induced noises is proposed from Bayesian viewpoint theoretically and experimentally. INTRODUCTION A wind-induced noise influences the acoustic measurements in the outdoor. The wind-induced noise fluctuates randomly owing to the temporal changes of wind velocity at the observation point and shows arbitrary probability distribution forms of non-Gaussian type. It is necessary to establish some systematic countermeasure methods for the wind-induced noise especially when measuring the low frequency acoustic signals, because the wind-induced noise includes components in a low frequency range which can not be reduced with use of the usual methods ( such as the usage of wind screen and others). This is due to the fact that the physical mechanism of wind-induced noise is not known and the stationary property of wind-induced noise can not be assumed. Here, we will pay our attention to the fact that the windinduced noise is generated from the temporal change of wind velocity, which can be measured separately when observing the acoustic data. The wind velocity changes relatively slower than the changes of the wind-induced noise and, in addition to the amplitude of wind velocities, the directions and the frequency components of them are also important. Then, for predicting the whole probability form of wind-induced noise, we should observe the velocities in different times and/or positions, and then consider the nonlinear multivariate correlation informations between the wind-induced noise and the wind velocities. In this paper, from the above viewpoint, a new trial of dynamical countermeasure method for the wind-induced noise will be proposed by considering how to use effectively the multivariate information of wind velocities. More concretely, we will expand first the multivariate moment generation function hierarchical-ly and derive the multivariate form of generalized regression model of the wind-induced noise on the set of wind velocities. Next, by predicting various statistics of the wind-induced recursively, we evaluate the true acoustic signals dynamically under the wind-induced noise from Bayesian viewpoint. Finally, the experimental confirmation of the proposed method has been confirmed by applying it to the actual data of field measurement. THEORETICAL CONSIDERATIONS The Generalized Regression Model of Windinduced Noise on Wind Velocities It is well-known that there is the averaged relation-ship between a wind-induced noise v k and a wind velocity where n: a priori known integer, and t : known proportional constant. But the wind-induced noise can not be explained only with use of the averaged relationship in Eq.(1). Then, we will consider the deviations from Eq. Then, we will detect and utilise the correlation information embedded under e k as much as possible, by using the information on the wind velocities powers. Here, we consider the m observations of wind velocities u 1k 2 , u 2 k 2 ,L, u mk 2 ( u 1k 2 is the present vel-ocity u 1k 2 = u k 2 ) at the different time or positions. For grasping the whole probability form of the deviations of a windinduced noise without minimum information losses, the joint probability function P(e k , u 1k 2 , u 2 k 2 , L , u mk 2 ) should be considered. By introducing the joint characteristic function associated with the joint probability function on windinduced noise and wind velocities, the hierarchically expanded joint characteristic function can be obtained as follows : The correlation information among the wind noise and the wind velocities can be reflected in each expansion coefficient hierarchically. By using the inverse Laplace transformation of the moment generating function, the statistical prediction of arbitrary moments of a deviation term e k can be derived in the form of function on the observation of wind velocities : That is, the generalized multivariate regression model can be realized for predicting the arbitrary moments of the wind noise with use of the wind velocities. A Wide-Sense Digital Filter for Detecting Acoustic Signals Embedded under Wind-induced noise In the same analytical viewpoints, a dynami-cal signal detection method under the background and the wind noises can be derived. Here, based on the analysis based on the additive property of energy quantities and the physical mechanism, the system and observation equations for the acoustic systems are formulated as: where x k denotes the unknown acoustic signal at a time stage k, and y k denotes the contaminated obser-vation. v k and u k are the background noise and the wind noise. For detecting the acoustic signal x k with use of the observation recursively, the conditioned joint mo-ment generating function M * (q, q 1 ) (=< exp{x k q +y k q 1 } | Y k >) should be considered. Here Y k is the set of past observations {y 1 , y 2 ,L, y k }. Then, by considering Bayes theorem, the unified algorithm of estimating the unknown acoustic signals under these random noises can be obtained : where k i, j * denotes the (i,j)-th predicted joint cummulant of x k and y k . And the coefficients are defined as : Finally, by combining it with the unified prediction dynamical state estimation algorithm in a recursive form can be realized. EXPERIMENTAL CONFIRMATIONS The effectiveness of proposed estimation theory has been experimentally confirmed by applying it to the actual acoustic data in the outdoor field under the wind-induced noises (omitted owing to page-limit). CONCLUSIONS In this paper, for the acoustic system in an outdoor environment, a new derivation of the systematic countermeasure method for the wind noise was proposed with use of the wind velocities. ACKNOWLEDGMENTS We will thank to Mr.M. Kubo, Mr.K.Furukawa and Mr.H.Kondo. 2) T.Katayama:Kalman Filter and Applications (Asakura, 1983 We investigate how the acoustic field created by air inflowing through the side windows of automotives of various sizes depends on the vehicle speed. Extremely large sound pressure levels are found near the frequency of the fundamental mode of the vehicle's Helmholtz resonance (about 20 Hz). Much smaller peaks are also found at the frequencies corresponding to higher harmonics. A positive, tight correlation exists between the peak frequency and the vehicle speed. The peak sound pressure levels increases with the vehicle speed, up to a critical speed, and declines afterwards. In absolute terms, much larger sound pressure levels were found inside a sub-compact car, as compared to a four-door sedan and to a station wagon. INTRODUCTION Interior noise in the low frequency region is usually of minor concern inside contemporary cars. However, the presence of open sunroofs and/or side windows can easily result in very high amplitude, low frequency noise, which strongly interferes with the driver's concentration and wakefulness, causing considerable discomfort and even possibly impairing his ability to provide safe driving [1]. This contribution reports the first results of an ongoing investigation aimed at characterizing the low frequency sound pressure field in vehicle interiors, as a function of vehicle type and speed, as well as of the opening sizes, number and locations. Here we focus on the relation between the peak sound pressure level and frequency and the vehicle speed. METHODS Three vehicles were tested, including one subcompact city car (SC), one sedan (S), and one full size station wagon (SW). Measures were taken positioning a microphone with extended sensitivity to low frequency waves (Brüel & Kjaer type 4193) at a distance of about 10 cm from the right ear of the driver. Experiments were carried out between March and December 2000, with different vehicle speeds, ranging from 80 to 140 Km/h, and a fully open rear right window. Typical test durations were set at around 90 seconds. Narrow band FFT spectra were calculated with bandwidths 0 -100 or 0 -200 Hz and resolutions 0.5 Hz. One third octave band spectra were synthesized from narrow band spectra. Data analysis was performed using the software Kyplot (version 2.0) to calculate best fitting models. RESULTS Spectral shape FIGURE 1. Frequency spectrum of vehicle SC After an initial area of low signal at very low frequencies , the sound pressure level shows a very fast rise, culminating with a sharp peak at frequencies f 1 in the range 16 -22 Hz depending on the various parameters affecting the system. This is followed by an equally sharp decline. Beyond 30 Hz, a more shallow downward trend sets in, which extends to about 80 -100 Hz. Superposed on this "noise floor" are the various peaks due to the higher harmonics, at SESSIONS frequencies n´f 1 . Their amplitudes are much smaller (at least 25 -30 dB) than that of the first peak. The exciting mechanism is provided by the aerodynamical noise due to flow instabilities in the boundary layer between the vehicle interior and the outer flow [2]. These instabilities results in vortex shedding and the convection of discrete vortexes over the opening determines the acoustic excitation of the cavity. Acoustic emission is triggered by the interaction of the vortexes with the downstream edge of the opening, and shows a characteristic frequency where N S is the Strouhal number, v is the vehicle speed and d is the opening streamwise length. The actual emission takes place at a frequency dictated by the acoustical response of the vehicle interior. The latter is very well approximated by a Helmholtz resonator [3]. Under the circumstances found in vehicles, the fundamental proper mode has a frequency where c s is the sound speed, A is the window opening area, and V is the interior volume. For the typical values assumed by these parameters in tested vehicles, f H is of order 20 Hz. Peak frequency and velocity Figure 2. Peak frequencies at various speeds This trend is very robust, as it appears to be present in all vehicles of different shapes and sizes. A power law f µ v a has been found to provide good fitting to data. Results indicate similar trends (a » 0.5) for the SC and the S vehicles. A matched pair t-test shows however that SPL values in the S vehicles are significantly larger (t = 6.54, P < 0.05) than in the SC vehicle. A more shallow slope characterizes the SW vehicle. The same behavior is replicated by the higher harmonics, although they prove at times hard to locate due to unfavourable signal to noise ratios. Finally, the spectrum becomes more and more spread out as velocity increases, with lower resonance peaks and higher high frequency "noise floors", pointing to a rising trend of damping in the system. ACKNOWLEDGEMENTS The assistance of Dr. Renato Gurin during the measurements is gratefully acknowledged. Flow over small steps on the exterior of an airplane causes aerodynamic pressure fluctuations that are up to 30 dB greater than those under a turbulent boundary layer on a surface without pressure gradients. Although of small extent, regions with such excitation may contribute substantially to the interior noise in airplanes, particularly if well-radiating structures like windows are excited. The aim of this investigation is the prediction of interior noise of airplanes resulting from flow over such steps, common on airplane exteriors. This requires derivation of the relations governing the vibration and acoustic radiation of thin-walled structures in the case of excitation by such pressure-fluctuation fields. Existing interior-noise prediction procedures in the framework of statistical-energy analysis are modified to reflect the peculiarities of these fields, which differ from those under a turbulent boundary layer in terms of correlation scales and non-uniformity scale. This paper is limited to the transmission due to non-resonant excitation of the structure. A paper delivered at Inter-noise 2001, The Hague, Netherlands describes the somewhat greater resonant contribution to the radiated field. This work is the result of cooperation between TsAGI, Moscow and Boeing, Seattle. REFERENCES