Experimental and numerical studies of the aeroacoustics of axisymmetric supersonic inlets

Experimental and numerical studies of the aeroacoustics of axisymmetric supersonic inlets

Journal of Sound and Vibration (1995) 184(5), 853–870 EXPERIMENTAL AND NUMERICAL STUDIES OF THE AEROACOUSTICS OF AXISYMMETRIC SUPERSONIC INLETS K. P...

1008KB Sizes 1 Downloads 12 Views

Journal of Sound and Vibration (1995) 184(5), 853–870

EXPERIMENTAL AND NUMERICAL STUDIES OF THE AEROACOUSTICS OF AXISYMMETRIC SUPERSONIC INLETS K. P. D, Z. Y  W. F. N Department of Mechanical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061-0238, U.S.A. (Received 1 February 1994, and in final form 6 June 1994) A series of experiments were conducted at an outdoor facility to evaluate the aerodynamic and acoustic performance of a supersonic inlet with a modified auxiliary door geometry. A 1/14 scale model of an axisymmetric, mixed-compression, supersonic inlet designed for civil transportation (P-inlet) was used in conjunction with a 10·4 cm (4·1 in) turbofan engine simulator, to test a new auxiliary door geometry designed to reduce engine fan noise radiated to the forward sector. The flow distortion at the fan face was reduced by modifying the auxiliary inlet doors. The new door geometry uses door passages with increased circumferential span to improve the distribution of the flow entering through the doors. To provide a basis for comparison, a baseline inlet with an auxiliary door geometry representative of the original designs was also tested. The results show that the new door geometry is successful in reducing circumferential distortion of the flow Mach number near the fan face by 30% compared to the baseline configuration. In addition, far field radiation of the blade passing frequency tone and overall noise is reduced by an average of 4 dB (SPL) in the forward sector (0° to 110° from the inlet axis). A 3-D, viscous numerical simulation of the baseline door configuration reveals a large region of flow separation downstream of the auxiliary door. This flow separation leads to a significant increase in circumferential distortion at the fan face. 7 1995 Academic Press Limited

1. INTRODUCTION

Recently there has been renewed research interest in the development of a new-generation supersonic cruise aircraft for commercial transportation. Current supersonic passenger aircraft provide reduced flight times compared with subsonic transports, but also significantly increase airport community noise levels; the latter has resulted in severely restricted airport access. In order for a high speed civil transport to be commercially successful, its acoustic impact on airport communities must be minimized. Aircraft takeoff and landing approach represent the two flight conditions when the communities surrounding an airport are most adversely affected by excessive noise. Although jet noise is expected to be the predominant noise source for Supersonic Transport (SST) aircraft, previous analyses by Trefny and Wasserbauer [1] indicate that ‘‘forward propagated engine fan noise is a significant component during takeoff and landing approach’’. Many of the properties of the forward propagated fan noise from an SST aircraft are determined by the design of the engine inlet. Unlike conventional subsonic inlets, supersonic inlets incorporate many complex, variable features in order to satisfy aerodynamic requirements at a range of flight speeds. The effects of these variable inlet features on the acoustic behavior of the inlet are not fully known. 853 0022–460X/95/300853 + 18 $12.00/0

7 1995 Academic Press Limited

854

K. P. DETWILER ET AL.

Auxiliary inlet doors, which are required at low flight speeds to compensate for the small capture area of supersonic inlets, have been shown to adversely affect acoustic performance. Woodward et al. [2] tested the aeroacoustic performance of a 1/3 scale, supersonic ‘‘P-inlet’’ and concluded that open auxiliary doors significantly increased the fundamental tone of the fan noise. The increased fan noise was suspected to be the result of increased flow distortion at the fan face generated by the auxiliary doors. Nuckolls and Ng [3] developed a modified auxiliary inlet door geometry designed to reduce the radiated fan noise from a supersonic inlet by reducing the circumferential distortion of the flow field near the fan face. Tests were conducted with a 1/14 scale supersonic inlet coupled to a 10·4 cm (4·1 in) diameter turbofan engine simulator. The geometry of this inlet is identical to that of the P-inlet tested by Woodward et al. [2]. The turbofan simulator was operated at 60% design speed to simulate landing approach conditions. When compared to the aeroacoustic performance of the same inlet fitted with the original, baseline geometry doors, the modified doors reduced circumferential flow distortion by a factor of two and lowered forward radiated fan noise by 6 dB. Nuckolls and Ng [3] also demonstrated similarities in the noise radiation behavior between the 1/14 scale supersonic inlet and the larger inlet tested by Woodward et al. [2], and concluded that, despite the difference in the Reynolds number, the small scale inlet and engine simulator could be used to investigate acoustic trends and inlet noise mechanisms. The present study focused on evaluating the aeroacoustic performance of the same modified auxiliary door geometry developed by Nuckolls and Ng [3], but at simulated aircraft takeoff conditions. During the landing approach the aircraft’s engines were typically throttled to a lower speed and the fan noise was dominated by the blade-passing frequency tone and its harmonics. For aircraft takeoff, the engines were operated near design speed and the fan noise was comprised of a series of distinct tones, called combination tones, in addition to the blade-passing tones. The objective of this experiment was to determine how the modified auxiliary doors performed in comparison to the baseline configuration, with the higher level of inlet airflow and the combination-tone fan noise representative of aircraft takeoff conditions. The second objective of this experiment was to provide a better understanding of the flow physics of the original P-inlet through the use of computational fluid dynamics (CFD). A 3-D viscous, compressible code (PARC [4]) was used to calculate the flow in the P-inlet fitted with the original baseline geometry doors at conditions similar to that of aircraft takeoff. The calculation revealed interesting flow features that could be explored for future effort in designing a supersonic inlet that may minimize flow distortion at the fan face.

2. THE EXPERIMENT

2.1.    The turbofan engine simulator, a Tech Development Model 460, is similar to that used in previous aerocaustic test programs to simulate a Pratt and Whittney JT9D turbofan engine [5]. The simulator features a single-stage fan section, consisting of 18 fan blades and 26 stator guide vanes, powered by an air turbine. Due to limitations on the compressed air supply at the test facility, the simulator was tested at 70 000 rpm, or 88% design speed (88 PNC). The test speed provided a supersonic blade tip velocity, and therefore generated a fan noise spectrum comprised of combination tones.

 

855

Figure 1. The auxiliary door geometries (baseline versus modified): (a) rear view, A–A; (b) side view.

2.2.    The test inlet used in this experiment was based on an axisymmetric, mixed compression supersonic inlet, typically referred to as the ‘‘P-inlet’’. Designed for transonic flight conditions and an aircraft cruise speed of Mach 2·65, the P-inlet incorporated several variable features including: a translating centerbody to permit variations in inlet geometry, boundary layer bleed systems to minimize shock – boundary layer interactions, and auxiliary inlet doors to increase the inlet capture area at low flight speeds. The current test inlet, shown in Figure 1, is essentially a 1/14 scale model of the P-inlet, and incorporated the aforementioned variable features except for the cowl and centerbody bleed systems. Because both the aerodynamic results of Trefny et al. [1] showed the bleed systems to provide negligible boundary layer reduction at low flight speeds and the concerns of shock – boundary layer interaction do not apply to low speed flight, the cowl and centerbody bleed systems were not incorporated into the design of the current test inlet. In the test inlet, the centerbody assembly was supported by four equally spaced struts located near the entrance of the fan (Figure 1). For all tests, the centerbody was placed in the fully retracted position. The cross-section of the support struts was designed to minimize the strut wake shed into the fan. Although the wakes shed by the struts are recognized as a source of fan noise generation, the centerbody support struts are an integral part of the mechanical design of the P-inlet and therefore were included in the test inlet. 2.3.    The supersonic test inlet had four auxiliary doors equally spaced around the inlet circumference. The auxiliary inlet door geometries evaluated in this experiment are compared in Figure 1. The auxiliary doors designated as ‘‘baseline’’ had the same geometry as the fully open auxiliary doors of the 1/5 scale P-inlet model developed and tested by NASA Lewis [2]. The modified auxiliary door design featured an increased circumferential span (83° versus 50° for the baseline geometry); an effort to make the auxiliary door provide a more circumferentially uniform air distribution to the fan face. Photographs of the two auxiliary door designs are given in Nuckolls [3]. 2.4.   Measurements of total pressure recovery and flow distortion were used to compare the overall aerodynamic performance of the baseline and modified inlet doors and to facilitate the interpretation of the acoustic results. Total pressure and Mach number measurements

856

K. P. DETWILER ET AL.

Figure 2. The aerodynamic instrumentation of the supersonic test inlet and simulator.

were made with conventional probes at the cowl lip, inlet throat, fan face and fan exit stations shown in Figure 2. A 1·6 mm (1/16 in) pitot-static probe was used for the Mach number measurements. For the total pressure measurements at the fan face and fan exit stations, a 1·6 mm (1/16 in) diameter Kiel probe was chosen for its reduced sensitivity to flow angularity. In a previous experiment conducted by Nuckolls [3], the flow upstream of the auxiliary doors was found to be axisymmetric. Therefore, measurements for the cowl lip and throat stations were taken at only one circumferential angle. The flow field in the region of the fan face was non-axisymmetric due to the influence of the flow from the auxiliary doors. The Mach number and total pressure distributions at the fan face station were resolved with the 35 point probe grid shown in Figure 2. The measurement grid spanned from the center of the support strut at 0° to the center of the auxiliary door (45°). For the measurement at 0°, it was necessary to remove the support strut at that location (the acoustic measurements were taken with all four struts in place). More details of the experimental set-up are available in Detwiler [6]. 2.5.   The acoustic measurements were conducted at an outdoor facility. The tests were static, meaning the engine simulator pulled air into the inlet from the surrounding atmosphere. Although an inlet control device (ICD) is typically used during static acoustic tests of aircraft engines to reduce the effect of atmospheric turbulence, it was not practical to use one with the supersonic inlet due to the presence of the auxiliary doors (an ICD would not cover the auxiliary doors). Instead, a system of data averaging was used to minimize the effect of random atmospheric turbulence on the acoustic results. In addition, the simulator was mounted in a test stand 122 cm (48 in) above the ground to reduce the possibility of exciting a ground vortex within the inlet flow field. Bruel and Kjaer model 4136 condenser microphones with 0·64 cm (0·25 in) diameter diaphragms were used to measure the acoustic field. The microphones were positioned 122 cm above the ground (the height of the inlet centerline) and placed along a circular arc centered at the inlet face, as shown in Figure 3. A radius of 122 cm (48 in) was selected for the microphone location arc to provide acoustic measurements in the far field (i.e., KL w 1, where K is the acoustic wavenumber of the blade passing frequency (BPF) and L is the distance between the inlet and microphones). The 12 microphone measurement points were located at 10° increments, from 0° to 110°, from the inlet centerline axis. The microphone signals were analyzed on a Bruel and Kjaer model 2034 dual-channel spectrum analyzer. The spectrum analyzer performed narrow bandwidth FFT (fast Fourier

 

857

Figure 3. The microphone layout, plan view: W, microphone location.

transform) conversions of the acoustic data. The upper frequency limit of the spectrum was set at 25·6 kHz, providing a spectrum bandwidth of 32 Hz. The upper frequency limit of the spectrum was sufficient to measure the fundamental BPF tone of the simulator (approximately 21 kHz for the test speed of 88 PNC), and was below the roll-off point for the microphones. The FFT results from the spectrum analyzer were used to record the BPF tone level and to investigate other fan-related tones. To compensate for the effects of random atmospheric turbulence, the analyzer was configured to calculate the linear average of ten consecutive noise spectra. Ten consecutive values of average BPF tone level were then recorded at each microphone position. In addition to the Fourier analysis, the root mean square (r.m.s.) of the microphone signal was also measured using a voltmeter to provide the overall sound pressure level (OASPL) of the acoustic signal. A low-pass filter was used during measurements of the overall sound pressure level to remove noise signals above 30 kHz. The OASPL, which represented the integration of the noise spectrum over the frequency range, was useful because it included the noise contributions of the combination tones as well as the BPF tones (the OASPL also included the noise contributions of the broadband and jet noise, but these were of little significance when added to the BPF and combination tones). 3. NUMERICAL SIMULATION

The 3-D, viscous, compressible PARC code [4] was used to calculate the flow field of the P-inlet fitted with the original, baseline auxiliary doors. The PARC code uses the Beam-Warming approximate factorization algorithm to solve the Reynolds’-averaged Navier–Stokes equation in conservation law form. The turbulent calculations use the Baldwin–Lomax [7] algebraic turbulent model. In this calculation, the boundary layers were assumed to be turbulent throughout the flow field. Due to symmetry, it was only necessary to model a quarter of the inlet. A mesh size of 101 × 36 × 36 was used in the calculation. Denser packing was used near the solid walls and also near the vicinity of the auxiliary door in order to resolve the sharp flow gradients. An independent grid study was performed to investigate the effort of grid spacing and grid size. Results showed that the current mesh used is near the optimal. The calculations were performed on a Cray Y-MP computer, requiring 6·5 Mb of memory. Convergence was achieved in about 1000 iterations and 3100 s of CPU time.

858

K. P. DETWILER ET AL.

The boundary conditions used in the calculations were as follows: uniform stagnation pressure and temperature at the inflow boundary, and a no-slip condition at all solid walls. Static pressure was specified at the outflow boundary so that the average throat Mach number from the calculation matched that of the experiment. The outer boundary between the freestream and the cowl lip was specified with the stagnation pressure and temperature such that air could also be drawn into the inlet at this boundary. The auxiliary door was modeled through the use of the boundary conditions. At the location corresponding to the position of the auxiliary door, the boundary condition was specified so that flow could be drawn into the inlet at a fixed angle corresponding to the inclined angle of the door. In this calculation, no attempt was made to model the detailed geometry of the cowl lip. Likewise, the four support struts inside the inlet were not modeled. It should be emphasized that the purpose of the calculation was to provide a qualitative analysis of the flow field and to supplement flow data that would otherwise be very difficult to obtain in the experiement. 4. AERODYNAMIC RESULTS

4.1.   The modified auxiliary door geometry was designed to provide more circumferentially uniform flow to the fan. Circumferential flow variations were of particular concern because they caused the blade loading to vary as the fan blades rotated, and unsteady blade loading increased the noise generation of the fan [8]. In contrast, radial variations of the flow parameters did not cause fluctuating blade loadings and had little effect on noise generation. For this reason, discussions of inlet distortion in this paper will focus on circumferential gradients of the flow parameters. The measured total pressure distributions at the fan face stations of the baseline and modified inlets are presented as contour plots in Figure 4. The plots represent a 90° portion of the annular inlet passage (between centerbody support struts) at the fan face. The symmetry of the auxiliary door construction enabled the extension of the 45° probe measurement grid to the 90° contour plots shown. The approximate locations of auxiliary inlet doors are indicated on the plots. The contour for the baseline inlet shows 98–99% total pressure recovery at all measurement points, except near the outer wall in the vicinity of the auxiliary door centerline. This low total pressure region is caused by a separation of the flow entering through the auxiliary door (confirmed by numerical simulation to be presented later). The region of separated flow creates a steep circumferential gradient of total pressure along the outer wall in front of the fan. This flow distortion is expected to increase the generation of fan noise. The total pressure contour for the modified inlet has a more uniform distribution when compared to the baseline inlet results. The lower total pressure region near the inner wall (due to the presence of the boundary layer) is quite evident in Figure 4. Near the outer wall, at where the auxiliary door is located, the modified case shows a low total pressure region. However, the gradient of the total pressure in this region is much less severe than the corresponding region in the baseline case. An area-average of the fan face total pressure values for both inlets showed a 2% reduction in total pressure recovery (from 97% to 95%) for the modified inlet. The axial Mach number contours, shown in Figure 5, support the observations from the total pressure distribution plots. The axial Mach number contour for the baseline inlet showed a region of low velocity near the outer wall, in the vicinity of the center of the auxiliary door. The low velocity in this region is, again, due to flow separation at the

 

Figure 4. The total pressure distribution at the fan face, 88 PNC: (a) baseline; (b) modified.

859

Figure 5. The Mach number distribution at the fan face, 88 PNC: (a) baseline; (b) modified.

downstream edge of the baseline auxiliary doors. The Mach number in the outer region of the fan changed from 0·5 to 0·2 within an angular distance of 22·5°. A gradient of axial velocity of this magnitude is expected to cause significant fluctuations in the blade loading, and therefore increased fan noise generation. The Mach number contour for the modified inlet does not show as much circumferential variation near the outer wall as the baseline case. In the outer region of the modified inlet, the axial Mach number changed by less than 0·1 throughout the 90° angular position of the contour plot. This small gradient of axial Mach number indicates that the increased span of the modified doors has improved the circumferential distribution of the flow through the auxiliary doors. Quantified values of circumferential distortion are useful for comparing the performance of the two inlet configurations. Values of circumferential distortion of the axial Mach number, given by the formula (Mmax − Mmin )/Mave , were calculated for each of the seven measured radial positions at the fan face station. These values were then averaged by area-weighting to yield the overall circumferential distortion values shown in Table 1. Compared to the baseline configuration, the modified auxiliary doors reduced circumferential distortion of the axial Mach number by a factor of 1·3. Also shown in Table 1 are the circumferential distortion results from Nuckolls [3] for the same inlets tested at a fan speed of 60 PNC. The current reduction in circumferential distortion is consistent with the

860

K. P. DETWILER ET AL.

T 1 Circumferential distortion of Mach number at the fan face Inlet configuration Baseline inlet Modified inlet Reduction factor

Circumferential distortion (88 PNC)

Circumferential distortion (60 PNC)

0·31 0·23 1·3

0·22 0·11 2·0

results of the lower fan speed, although the reduction factor has decreased for the present test. Nuckolls [3] demonstrated that a reduction in circumferential distortion at the fan face lowered the level of forward radiated fan noise from a supersonic inlet. The modified inlet was therefore expected to provide a reduction in fan noise radiation at the test speed of 88 PNC. 4.2.      The results from the aerodynamic measurements show that the area averaged fan pressure ratio is almost identical between the baseline and modified door case (fan stagnation pressure ratio = 1·38). Since the fan rpm and pressure ratio are the same, the two inlet configurations must have the same overall fan mass flow rate. The significance of this is that all of the following acoustic comparisons are made under similar fan operating conditions.

Figure 6. The effect of the simulator speed on radiated noise, baseline inlet: *, combination tone; BPF, blade passing frequency; 4, fourth combination tone. (a) 60% corrected design speed (PNC); (b) 88% corrected design speed (PNC).

 

861

5. ACOUSTIC RESULTS

5.1. -   Narrow-band spectra are useful for illustrating the characteristics of the radiated engine noise. Each of the noise spectra presented in this section represents the linear average of the frequency spectra of ten successive data sets; the averages were calculated to compensate for random fluctuations in the levels of the fan noise tones. The effect of simulator fan speed on the noise signal radiated from the test inlet is illustrated in Figure 6; the samples shown are for the baseline inlet configuration at the 20° microphone position. At the lower fan speed, 60 PNC, the blade passing frequency tone (BPF) is seen to dominate the spectrum (Figure 6(a)). The onset of combination tone noise is evidenced by the smaller tones located at integer multiples of the fan rotational frequency, 833 Hz. Although the fan tip speeds are subsonic at 60 PNC, regions of high axial air velocity at the fan face (generated by the auxiliary doors) produce supersonic

Figure 7. Radiated noise spectra at 88 PNC, baseline inlet: *, combination tone; BPF, blade passing frequency; 4, fourth combination tone. (a) 20° microphone position; (b) 60° microphone position; (c) 110° microphone position.

862

K. P. DETWILER ET AL.

relative blade tip speeds, which lead to the generation of the combination tones. The sample noise spectra for the simulator speed of 88 PNC (Figure 6(b)) show an expected increase in the level of the BPF tone as well as the abundance and magnitude of the combination tones; at 88 PNC, the noise spectrum contains combination tones at all multiples of the shaft rotational frequency. Sample spectra of the noise radiated to the 20°, 60° and 110° microphone positions from the baseline inlet with the simulator running at 88 PNC are presented in Figure 7. These figures illustrate the changes in the radiated noise at different microphone positions. It is important to note the increase in the BPF tone level as the angular position of the microphone is increased. This trend, observed with both inlet configurations, will be further substantiated by the BPF tone radiation results of the next section. The higher level of the BPF noise signal in the rearward sector of the measurement field is most likely due to the radiation of noise from the fan exit and auxiliary doors. The shape of the broadband noise (i.e., those frequencies between the distinct fan tones) also changed at the aft microphone positions. As the angular position of the microphone increased from 20° to 110°, the broadband noise increased by roughly 4 dB over most of the frequency range. For frequencies below 1000 Hz, the broadband noise at the 110° location was approximately 15 dB greater than the corresponding noise level at 20°. This low frequency noise present in the aft sector was believed to be generated by aerodynamic turbulence, or mixing noise, from the jet plumes of the fan and drive turbine. 5.2.    This section presents the key results of the acoustic testing. Figure 8 is a directivity plot on which is shown the average level of the BPF tone at each microphone position for both inlet configurations. Each BPF tone value represents the average of ten consecutive BPF readings from frequency spectra recorded with the signal analyzer. Uncertainty bands are included in the figure at two angles; the uncertainty bands show a calculation of one standard deviation, from the mean, of the ten measurements taken at that angular position. The modified inlet provided an average reduction in the BPF tone of 4 dB over the range of microphone positions. An investigation into the radiation patterns of the two inlets is presented later in the discussion section.

Figure 8. The directivity of the BPF tone: w, baseline; W, modified.

 

863

Figure 9. The directivity of the overall noise: w, baseline; W, modified.

The overall noise radiation patterns of the two inlets are shown in Figure 9. The sound pressure level of the overall noise (OASPL) included the noise contributions of the combination tones and broadband noise along with that of the BPF tone. The uncertainty bands shown in Figure 9 represent the range of the microphone signal voltage recorded during measurements at that angular position. The modified inlet provides an average reduction in the radiated overall noise of 3 dB (OASPL) over the range of microphone positions. The reduction in overall noise for the modified inlet reflected both the lower levels of the combination tones and the reduction in the BPF tone. The BPF tone and overall noise radiation plots for both inlet configurations showed a trend of rising noise levels in the aft sector (90–110°), reflecting the contribution of the sound radiated from the fan exit. The noise radiated from the fan exit is expected to be of greater magnitude than that radiated from the inlet mouth, due to the considerably shorter duct length between the fan and the fan exit (essentially the fan shroud only). A shorter duct length resulted in less attenuation of the noise as it propagated to the duct opening. The modified inlet showed an improvement in acoustic performance over the baseline inlet in the aft sector, indicating a reduction in the noise radiated from the fan exit for the modified inlet. The increase in the noise levels (both BPF tone and overall) at the rearward sector of the measurement field may also be due to the fact that the microphone center is chosen at the mouth of the inlet. As such, the distance between the microphone location at 110° and the center of the actual noise generator (fan and drive turbine) is shorter by approximately 25% than at 0°. The choice of the mouth of the inlet as the microphone center is consistent with the larger scale test by NASA (Woodward et al. [2]), such that comparison can be made between the two tests. It is also worth noticing that this experimental set-up has multiple ‘‘point’’ sources distributed axially—it is therefore impossible to place the microphone center at the noise center, because there are multiple noise centers. Perhaps most important is the fact that the conclusions of this study, where a comparison is made between the baseline and modified inlet configurations, are independent of the choice of the microphone center. The acoustic results presented thus far represent the noise radiated from the inlets to a single measurement plane. The radiation field surrounding the inlet is not expected

864

K. P. DETWILER ET AL.

to be completely axisymmetric, however, due to the potential for noise radiation through the auxiliary doors. For this reason, additional tests were conducted to verify the acoustic performance of the inlets at circumferential angles other than the primary test configuration. For the primary test configuration, the microphone positions were arranged in a plane (parallel to the ground) that bisected the auxiliary inlet doors. To measure the noise radiation at different circumferential angles, the inlet was rotated relative to the microphone measurement plane. The additional overall noise measurements were made at circumferential angles of 22.5° and 45° from the primary inlet orientation. The acoustic results from the additional measurement planes show a reduction in OASPL noise radiation for the modified inlet, similar to that which occurred in the primary test orientation, although the magnitude of the reduction was slightly less in the aft sector of the additional planes [6]. Thus, it is expected that the modified inlet will provide a reduction in radiated noise at all points in the forward radiation field. 6. NUMERICAL RESULTS

In order to establish the credibility of the numerical simulation, the total pressure contour from the experiment was compared with the calculation at the fan face station. The results are presented in Figure 10, which clearly shows that, qualitatively, the two are in good agreement. Note that the size of the region of non-uniformity created due to the air

Figure 10. A comparison of the total pressure ratios at the fan face: (a) experiment; (b) CFD.

 

865

Figure 11. The Mach number contour at a longitudinal cut through the center of the door.

injection from the auxiliary door appears to be larger in the calculation. This may perhaps be due to the relatively crude model used to model the door region in the calculation. Further refinement in this can be incorporated in future effort. The Mach number contour along a longitudinal cut of the inlet through the center of the auxiliary door is shown in Figure 11. The flow accelerates from the entrance of the inlet to the throat region and then decelerates between the throat region and the auxiliary door. This deceleration caused an adverse pressure gradient and the resulting thickening of the boundary layers, at both the centerbody and the outer wall, is quite evident in Figure 11. This figure also reveals some interesting interactions between the auxiliary air flow and the primary airflow. Immediately upstream of the door, the casing boundary layer at the outer wall is thickened considerably, and is eventually mixed in with the high momentum fluid from the injection flow. The auxiliary air flow momentum is so strong that the jet penetrates all the way to the centerbody, and in that process also mixes with the centerbody boundary layer. Perhaps most important is the significant flow separation at the outer wall immediately downstream of the door. This should not come as a surprise; as the fluid is drawn in, it cannot turn aburptly around the corner at the downstream edge of the door. This separation bubble grows as the flow is convected to the fan face. This flow feature is also evident in Figure 12, in which Mach number contours for several cross-sectional planes along the inlet are presented. In Figure 12, plane 1 is upstream of the door, and it clearly shows the buildup of the outer wall and the centerbody boundary layers. At plane 2, which is located at the upstream edge of the door, the thickening of the outer wall boundary layer due to the effect of the flow injection is very noticeable. Planes 3 and 4 are about one-third of the way down the door and at the downstream edge of the door, respectively. At these two planes, the flow distortion due to the air jet through the auxiliary door is quite evident. Downstream of the door at plane 5, the separation bubble, as also shown in Figure 11, is clearly visible. This region of separated flow grows bigger in size at plane 6, which corresponds to the fan face.

Figure 12. The Mach number contour—a cross-sectional cut.

866

K. P. DETWILER ET AL.

Figure 13. The secondary flow kinetic energy contour—a cross-sectional cut.

The CFD data were also processed to investigate the variation of the circumferential distortions of Mach number for planes 1–6, in a way similar to what was done on the experimental data. Results show that the circumferential distortions starts with a small value at plane 1, and then increases gradually through planes 2 and 3. However, it reaches a local minimum at plane 4, before the value starts rising sharply again through planes 5 and 6. This conclusion is almost self-evident in Figure 12. At planes 5 and 6, the separation bubble at the outer wall has severely distorted the flow. As a result, this separation bubble has led to a significant increase in the circumferential distortion factor, especially when compared to that at plane 4. This has an important implication: in order to minimize circumferential distortion (for the purpose of reducing noise generation), the fan should be located as close to the downstream edge of the door as possible. This, in effect, is to minimize the region of interaction between the separated flow at the outer wall and the fan. One remaining question has yet to be answered regarding Figure 12: What causes the circumferential distortion factor to reach a local minimum at plane 4? It is hypothesized that the interaction of the outer wall boundary layer with the injected flow through the auxiliary door created a pair of vortices [9]. It is conjectured that the presence of the vortex causes more mixing and helps in spreading out the auxiliary airflow, which then leads to a minimum circumferential distortion factor at plane 4. The above hypothesis is supported by Figure 13, which shows the contour of the kinetic energy associated with the cross-plane velocities, i.e., (v 2 + w 2)/2, also commonly referred to as the secondary flow kinetic energy. As shown in Figure 13, at plane 3, the presence of the two vortices near the edge of the door is quite evident. Downstream of the door at planes 5 and 6, where the separation bubble is located, the vortices are seen to migrate outward in the circumferential direction, while they still remain fairly concentrated near the outer wall. 7. DISCUSSION

The lower BPF tone radiation of the modified inlet compared to the baseline inlet, as illustrated in Figure 8, can be attributed to two factors: the reduction in circumferential distortion and the higher throat Mach number of the modified inlet. The reduction in circumferential distortion, achieved through the improved flow distribution of the modified doors, has reduced the noise generation of the fan. The details of the Mach number and mass flow distributions in the two inlets are presented in Appendix A. It is shown that the throat Mach number of the modified inlet is near the choked condition, whereas the throat Mach number of the baseline inlet is only about 0·56. The higher throat Mach number of the modified inlet has attenuated the fan noise as it propagated forward through the

 

867

inlet. With the present far field noise radiation data alone, it is not possible to separate the effects of these two factors. However, when the same auxiliary door configurations were tested at a lower simulator speed (60 PNC), Nuckolls and Ng [3] demonstrated a 6 dB average reduction in BPF tone radiation with the modified inlet. This reduction could only be attributed to the decrease in circumferential flow distortion at the fan face (the noise reduction due to the choking effect was minimal, since the throat Mach numbers for both inlets at 60 PNC were below 0·5). Hence, it is postulated here that the effect of the reduced distortion at the fan face of the modified inlet provides a considerable portion of the tone noise reduction in the current test. The higher Mach number in the throat of the modified inlet is due to the reduction in auxiliary door flow compared to the baseline configuration. To simplify the acoustic comparison between the inlets, it would be desirable for both inlets to have the same throat Mach number, thus eliminating concerns of the choking effect influencing the acoustic comparison. 7.1.       In order to gain insight into the BPF tone radiation patterns of the two inlet configurations, a study of the modal content of the fan noise was conducted. As introduced by Tyler and Sofrin [10], the BPF noise signal generated by the fan can be represented as a superposition of rotating pressure distributions, called modes, propagating along the inlet passage. Through a kinematic relationship involving the fan blades and flow disturbances, it is possible to determine the circumferential order of the modes generated in the inlet; the circumferential order, (m) = B 2 kV, where B is the number of fan blades, k is the spatial distortion harmonic number spanning all positive and negative integers, and V is the number of stationary disturbances that interact with the fan blades. In the current supersonic test inlets, there are three apparent sources of distortion near the fan: the row of stator blades behind the fan (of order V = 26), the four centerbody support struts (of order V = 4), and the auxiliary inlet doors (of order V = 4). Although it is possible for each of the disturbance sources to generate an infinite combination of modal patterns, there is a finite set of modes that can propagate through the inlet duct; the amplitudes of most of the modes will decay rapidly in the duct, while certain modes will propagate through the inlet unattenuated. The criteria for determining whether a particular mode is propagating depends on the frequency of the signal, the radial and circumferential order of the mode, and the radius and hub/tip ratio of the duct. Tyler and Sofrin [10] have demonstrated that for a particular mode order and hub/tip ratio, there is a minimum (or critical) Mach number at which the tips of a pressure pattern must rotate in order for the mode to propagate through the duct unattenuated. The circumferential mode orders that can be generated by the fan which can propagate through the inlets are presented in Table 2. Two locations along the inlet passage were evaluated based on the criteria for the modes to propagate: the fan face (hub/tip = 0·38), and the inlet throat (hub/tip = 0·75). For the lower circumferential-order modes; the actual tip Mach numbers are well above the critical Mach number and several radial mode orders above the fundamental are expected to propagate. For the circumferential mode m = 14, the actual tip Mach number is just above the critical Mach number, so that all radial modes above the fundamental will be nonpropagating (the critical tip Mach number increases with increasing radial order). Each of the modal patterns that propagate to the inlet mouth can be expected to radiate to the far field to form a radiation pattern comprised of lobe-shaped amplitude variations. Homicz and Lordi [11] developed a simplified method for determining the approximate angular location of the primary lobe of a radiation pattern from an unflanged circular duct.

868

K. P. DETWILER ET AL.

T 2 Propagating modes in the supersonic inlet Fan face Fan face Throat Throat Radiation Circumferential (hub/tip) = 0·38, (hub/tip) = 0·38, (hub/tip = 0·75), (hub/tip = 0·75), angle mode order, m actual Mtip critical Mtip actual Mtip critical Mtip (degrees) 2 6 8 10 14

10·0 3·3 2·5 2·0 1·4

1·4 1·3 1·2 1·2 1·2

9·1 3·0 2·3 1·8 1·3

1·2 1·1 1·1 1·1 1·1

8 21 27 34 51

Although the inlet evaluated here is an annular duct, numerical evaluations by Homicz and Lordi have shown that, for hub/tip ratios less than 0·5, the radiation pattern can be approximated by circular duct predictions. The approximate angular positions of the primary lobes of the modal patterns that can radiate from the supersonic inlets are given in Table 2. As shown in Table 2, there are five different circumferential mode orders that can radiate from the supersonic inlet. In addition, for the lower circumferential order modes, there can be several additional radial orders (i.e., n = 1, 2, 3, . . . ) which could radiate from the inlet. The possible superposition of so many radiation patterns within the forward sector prevents the selection of a single noise source (the stators, support struts or auxiliary inlet doors) as the primary noise source in the supersonic inlet. For a simulator speed of 60 PNC, Nuckolls and Ng [3] were able to identify evidence of the primary lobe of the m = 8 mode in the radiation patterns of the modified and baseline inlets. The higher simulator speed of the current test, however, increased the frequency of the noise signal and resulted in a large increase in the number of propagating modes in the inlets; the radiation patterns of the inlets are comprised of several mode patterns and it is not possible to identify a single primary mode. Only one modal pattern from the rotor–stator interaction noise (the m = 14 mode) can radiate from the inlets, however, so it is not expected that rotor–stator interaction provides a significant noise contribution to the forward sector. From the calculated directivity angles shown in Table 2, it is important to note that the primary lobes of all the modes which are propagating in the inlet are confined to radiate within the 0–60° sector. In other words, the mode radiation directivity analysis indicates that the noise emitted from the opening of the mouth of the supersonic inlet will not radiate significantly to the 60–110° measurement sector. A preliminary experimental investigation of this calculated radiation behavior was made using an acoustic baffle (a 48 × 48 in wooden panel faced on both sides with 1·5 in acoustic foam) to isolate the noise radiated from the mouth of the inlet [6]. Essentially, the baffle was placed around the inlet and positioned to prevent noise from the auxiliary doors and fan exit from radiating to the 0–90° sector. The results showed that the noise radiated to the 90° microphone position was approximately 10 dB lower than the noise radiated to the 0°, 30° and 60° microphone positions (only four microphone positions were measured). From the results of the modal analysis and this preliminary experimental investigation using the baffle, it is concluded that the portion of the measured radiation field (Figure 8) between 60° and 110° is comprised primarily of noise radiated from the auxiliary doors and fan exits of the test inlets. Within this rearward angular sector, the reduction in noise radiation with the modified inlet cannot be attributed to the choking effect in the modified inlet throat, because the noise radiated from the auxiliary doors and fan exit does not propagate through the inlet throat. Thus, it is plausible that the noise radiated from the

 

869

auxiliary doors and fan exit of the modified inlet has been reduced primarily by the reduction in circumferential flow distortion at the fan face. 8. CONCLUSIONS

One of the principal objectives of this research was to evaluate the aerodynamic and acoustic performance of a supersonic inlet with modified auxiliary inlet doors under simulated aircraft takeoff conditions. A series of experiments was conducted with a small scale model of an axisymmetric, mixed-compression, supersonic inlet connected to a 10·4 cm (4·1 in) diameter turbofan simulator. The simulator was tested at near design speed in an attempt to match the simulator noise source to that of an aircraft engine during takeoff. Two auxiliary inlet door geometries were evaluated: a baseline geometry representative of current door designs, and a modified geometry designed to reduce inlet flow distortion and fan noise radiation. Steady state aerodynamic measurements of the inlet flow field were made along with far field acoustic measurements of the fan noise. Through improved distribution of the auxiliary door airflow, the modified door geometry provided a factor of 1·3 reduction in circumferential distortion of the axial Mach number at the fan face, compared to the baseline door configuration. The modified inlet also lowered the BPF tone and overall noise by an average of 4 dB (SPL) in the forward and aft sectors (0–110° from the inlet centerline axis). In the forward sector between 0° and 60°, the noise reduction achieved with the modified door geometry was attributed to the reduction in circumferential flow distortion at the fan and a higher flow Mach number in the throat of the modified inlet. As a compromise for the distortion and acoustic improvements, the average total pressure recovery at the fan face of the modified inlet is reduced by approximately 2% compared to the baseline configuration. Computational fluid dynamics was also used to provide a qualitative analysis of the internal flow field of the baseline inlet. Results show the existence of a large separation bubble behind the downstream edge of the auxiliary door. This leads to a significant increase in circumferential distortion at the fan face. The presence of a pair of vortices, due to the interaction of the outer wall casing boundary layer with the injected airflow through the auxiliary door, can lead to better mixing of the auxiliary air flow through the door with the primary airflow from the mouth of the inlet. ACKNOWLEDGMENTS

The authors would like to thank Mr Abhijit Pande for obtaining some of the experimental data. We are also indebted to Mr Dick Woodward of the NASA Lewis Research Center, and Mr Bob Golub of the NASA Langley Research Center for many helpful discussions. REFERENCES 1. C. J. T and J. W. W 1986 NASA TP-2557. Low-speed performance of an axisymmetric, mixed-compression, supersonic inlet with auxiliary inlet. 2. R. P. W, F. W. G and J. G. L 1984 American Institute of Aeronautics and Astronautics Journal of Aircraft 21, 665–672. Low flight speed fan noise from a supersonic inlet. 3. W. N and W. F. N 1993 ASME Paper 93-GT-279; to appear in Transactions of the American Society of Mechanical Engineers, Journal of Engineering for Gas Turbine and Power. Fan noise reduction from a supersonic inlet during simulated aircraft approach.

870

K. P. DETWILER ET AL.

4. B. K. C and J. R. S 1989 Arnold Engineering Development Center, AEDC-TR89-15. PARC code: theory and usage. 5. D. A. T 1990 AIAA Paper 90-3951. Rotor wake/stator interaction noise-predictions versus data. 6. K. D 1993 Master’s Thesis, Virginia Polytechnical Institute and State University, Blackburg, Virginia. Reduced fan noise radiation from a supersonic inlet. 7. B. S. B and H. L 1978 AIAA-78-257. Thin layer approximation and algebraic model for separated turbulent flows. 8. B. D. M 1975 Journal of Sound and Vibration 40, 497–512. Axial fan noise caused by inlet flow distortion. 9. F. M. W 1991 Viscous Fluid Flow. New York: McGraw-Hill. 10. J. M. T and T. G. S 1962 Society of Automotive Engineers Transactions 70, 309–332. Axial flow compressor noise studies. 11. G. F. H and J. A. L 1975 Journal of Sound and Vibration 41, 283–290. A note on the radiative directivity patterns of duct acoustic modes. 12. W. H. B and N. P 1984 NASA CR-172390. Low speed performance and acoustic tests of an axisymmetric supersonic inlet—phase one tests with auxiliary doors closed.

APPENDIX A: MACH NUMBER AND MASS FLOW DISTRIBUTIONS IN THE INLETS

In addition to affecting the generation of fan noise, properties of the inlet flow field are instrumental in determining how the fan noise will propagate forward to the inlet openings. High flow Mach numbers in the inlet increase the time required for a noise signal to propagate through the inlet, and thereby dissipate more of the acoustic energy; this process of noise attenuation is referred to as the ‘‘choking effect’’, and is expected to occur for flow Mach numbers greater than 0·5 [2]. The throat Mach numbers for the modified and baseline inlets are shown in Table A1. At the test speed of 88 PNC, the throat of the modified inlet is choked while the baseline inlet has a throat Mach number of only 0·56. The choked throat of the modified inlet is expected to attenuate the noise propagating to the front of the inlet, while the throat of the baseline inlet, with a Mach number only slightly greater than 0·5, is not expected to significantly reduce forward propagating fan noise. The inlet mass flow distributions and average auxiliary door Mach numbers, also shown in Table A1, were estimated from the aerodynamic data. The auxiliary door mass flow rate was calculated as the mass flow through the fan (obtained from the fan performance map and the fan face total pressure data) minus the mass flow through the inlet throat; an approximate auxiliary door Mach number was then calculated using isentropic relations. The auxiliary door mass flow rate with the modified door geometry is approximately half the flow rate achieved with the baseline door geometry. The lower flow rate of the modified doors produces a corresponding increase in the flow rate through the throat of the modified inlet, resulting in the higher throat Mach number discussed previously. Although the auxiliary door Mach number is increased by 10% with the modified geometry (from 0·36 to 0·39), it is not expected that the choking effect will affect the propagation of noise through either door configuration (since both Mach numbers are below 0·5). T A1 Inlet mass flow distribution and Mach number at 88 PNC Inlet Modified Baseline

Mach number ZXXXXCXXXXV Throat Doors 1·00 0·56

0·39 0·36

Mass flow distribution (%) ZXXXXXCXXXXXV Throat Doors 77 53

23 47