Microphone Phased Array to Identify Liftoff Noise Sources in Model ...

JOURNAL OF SPACECRAFT AND ROCKETS Vol. 50, No. 5, September–October 2013

Microphone Phased Array to Identify Liftoff Noise Sources in Model-Scale Tests J. Panda∗ NASA Ames Research Center, Moffett Field, California 94035 and R. Mosher† Aerospace Computing, Inc., Mountain View, California 94043

Downloaded by Jayanta Panda on October 17, 2013 | http://arc.aiaa.org | DOI: 10.2514/1.A32433

DOI: 10.2514/1.A32433 A 70-microphone phased array, protected to withstand the harsh environment of a rocket test facility, was used to identify noise sources in the Ares I Scale Model Acoustics Test. In the unobstructed burn of a single solid rocket motor, the free-flowing plume itself was found to make a long noise source. The scenario changed completely in launch configurations where a 5%-scale model of the Ares I vehicle was tested in static firings. It was found that the impingement by the plume on various regions of the launch pad constituted the primary noise sources. The scenario is very different from current models, which assume that the plume itself is the noise source and do not account for impingement sources. As expected, the addition of water in the trench and the hole for the plume passage attenuated the associated noise sources. Water injection on the top of the pad (“rainbird”) was found to attenuate only the peripheral sources around the primary plume impingement zone. The noise maps suggest that the minimization of impingement by reducing vehicle drift, reducing plume spillage via increasing the size of the hole, and covering-up leakage paths for the sound waves from the trench will attenuate the liftoff acoustics level.

components. Therefore, the lowering of the acoustic level via improvements of the pad configuration, vehicle orientation, optimization of the water injection, and other precautions are beneficial. Every launch vehicle development program pays attention to the aforementioned issues and typically starts with a prediction of levels using the existing literature and the available database from past flights. For the Ares I program, at least two different predictions of the acoustic environment were made. Haynes and Kenny [1] used the procedure described by Eldred and Jones [2] with a small modification of the length of the plume potential core suggested by Varnier [3]. Plotkin et al. [4] used a computer code that seems to carry most of the elements of the Eldred and Jones method [2], with additions of shielding and diffraction of sound waves by the launcher deck, and a different empirical formulation for the total noise radiation [5]. Typically, the level of confidence on these predictions is not very high, and so tests using scaled models inevitably follow. The Space Shuttle program used a 6.4% scaled model [6,7]. A large number of tests and other efforts that preceded the Saturn program are listed in [2,5]. There were a number of tests conducted to identify and alleviate the liftoff levels on Ariane V [8]. There are indications of similar tests for the Japan Aerospace Exploration Agency M-V vehicle [9,10]. Sankaran et al. [11] provide a summary of various launch acoustic efforts. Returning back to the acoustics predictions, the engineering methodology of Eldred–Jones is fundamentally pinned to a model of noise-source distribution along a free plume. The presence of the elaborate launch pad is merely to make the plume bend at the location of the flame deflector (Fig. 1). The shortfall of this model is that it overlooks a significant impingement zone on the flame deflector, which is expected to be a very loud noise source. Recent computational fluid dynamics analysis by Tsutsumi et al. [10] supports this deficiency. Past efforts to compare predicted spectra to those measured in actual flight indicated that improvements could be made by assuming that noise sources lie closer to the nozzle exit plane [7]. To overcome this shortcoming, recent efforts took the path of locating an equivalent single point: the end of the supersonic core. Varnier [3] states, “ : : : the sound power peak location is related to the supersonic length of the flow, which appears to be the adequate reference length for a future jet noise model.” Sutherland [5] writes, “The model applies the widely accepted concept that the dominant sound source for supersonic jet flow is close to, and downstream of, the supersonic tip.” The model of Haynes and Kenny [1] truncated the potential core at the deflector and applied Sutherland’s suggestion

Nomenclature b d G R S St U w θ λ

= = = = = = = = = =

beamformed output aperture of the array cross-spectral matrix radial position of a microphone from array center weight applied to individual microphones Strouhal frequency plume velocity at nozzle exit steering vector angular position from the image center wavelength

Subscripts D f i, j m R

= = = = =

nozzle exit diameter frequency index for interrogation grid microphone index Rayleigh resolution

I.

Introduction

E

VERY part of a launch vehicle and launch pad as well as a large number of components used for the ground operations are subjected to the high acoustic loads generated during liftoff. The acoustic load is a major contributor to the vibroacoustics environment to which every component has to be designed, tested, and certified. Even a small reduction of a few decibels of the acoustic level translates into a sizable reduction of cost and weight, and it reduces the risk of failure during the qualification testing of a very large number of

Presented as Paper 2012-1171 at the 50th AIAA Aerospace Sciences Meeting, Nashville, Tennessee, 9–12 January 2012; received 31 May 2012; revision received 9 January 2013; accepted for publication 11 January 2013; published online 29 July 2013. This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. Copies of this paper may be made for personal or internal use, on condition that the copier pay the $10.00 per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include the code 1533-6794/ 13 and $10.00 in correspondence with the CCC. *Aerospace Engineer, Experimental Aero-Physics Branch. Associate Fellow AIAA. † Associate Engineer, Experimental Aero-Physics Branch. Member AIAA. 1002

1003


PANDA AND MOSHER

Fig. 1

Eldred–Jones [2] model of noise source distribution during liftoff.

The goal of the present work is twofold: first, to demonstrate the suitability of a microphone phased array for rocket vehicle applications, and second, to identify noise sources in selected firings of the Ares I Scale Model Acoustic Test (ASMAT). Correct identification of sources not only improves the predictive ability but helps with a quieter design of the launch pad and optimization of the water injection. Typical liftoff tests use a very large number of single microphones; however, such microphones are incapable of identifying the sources. The ASMAT test was a part of the Ares I program, which had been replaced by the Space Launch System (SLS) program. NASA decided to continue with the test program because the insights gained would be valuable for the SLS and various commercial programs. There were altogether 17 tests (and one motor-only test) conducted, all of which were static firings (i.e., the model was held fixed). However, the model positions were varied to represent vehicle locations at different time instances of liftoff. The original goal of ASMAT was to measure the acoustic and ignition overpressure (IOP) levels on the vehicle, service tower, and launch pad via individual microphone sensors. The present paper will not show data from these sensors. Instead, attention will be confined to the results obtained from the deployment of the phased array instrumentation. The first use of a microphone phased array to launch acoustics was by Gély et al. [8]. They used a 24-microphone ring array in a smallscale test of the Ariane 5 launch pad to identify sources responsible for the excessive low-frequency noise in the payload fairing. No rocket boosters were used for this test; instead, heated compressed air was used to simulate the rocket plumes. The beamformed maps correctly identified the sources, which led to various modifications such as extending some flue trenches and optimization of the water injection. The present application extends the array use from scaled simulations using heated air to direct measurement of solid-rocket plumes in launch-pad configurations.

II.

Phased Array and the Test Facility

A. Array Hardware Fig. 2 level.

Haynes and Kenny [1] modifications for predicting Ares I liftoff

(Fig. 2). The data presented in this paper, however, indicate that the elusive source is neither the supersonic tip of a free plume nor the end of the potential core but the impingement regions on the pad.

The basic array plate containing the 70-element condenser microphone array (Fig. 3) was designed for earlier wind-tunnel tests [12]. For the present application, 1∕4 in: G.R.A.S. condenser microphones, with protection grids removed, were flush mounted on the array plate. The microphones were connected to G.R.A.S. preamplifiers and then to dual-channel amplifiers via 164-ft-long

Fig. 3 a) Microphone pattern on the array plate, and b) point spread function for the two indicated frequencies for a point source located at the origin and z 240 in: away from the plate.

1004

PANDA AND MOSHER


Fig. 4 Phased array during a solid motor firing in a horizontal test stand (photo credit: NASA MSFC imaging team).

weather-protected cables. The present use in an outdoor rocket test facility required additional instrumentation and protection from the harsh environmental conditions. The array plate was modified to include a visible bandwidth video camera at the center with two dynamic pressure transducers (Kulite) to measure impulse loading from the IOP pulse. The video camera was used to directly image the physical noise sources. The aperture of the video camera lens was remotely adjusted via a servo-controlled actuator that allowed for adjustment for the variable daylight conditions and plume glare. A weather-resistant case was built around the plate to protect the sensitive equipment from rain, snow, wind, and other elements. The case was continuously purged using dry missile-grade air. Additionally, a large plastic bag was placed over the whole weather resistant case; the bag was removed before tests. All of these preventive measures were found to be effective through subsequent rain tests and ultimately during prolonged exposure at the test site when several rain, snow, and thunderstorms passed over the array hardware. After the initial testing, calibration, and software validation at NASA Ames Research Center (ARC), the entire phased-array hardware was shipped to test stand TS116 at NASA Marshall Space Flight Center (MSFC). There, the setup was rebuilt in two parts; the array box with all microphones and other sensors were mounted close to the model, and the electronic components (computer, dataacquisition system, amplifiers, etc.) were placed indoors in an airconditioned basement away from the test stand. The long cable bundle was passed through split conduit and other measures to protect from the environmental elements. The entire array setup was subjected to a noninterference criterion, where the operation of the phased array had to be completely independent and separate from the test objects. A phased array identifies the locations and levels of sources of sound waves propagating toward its direction. Ideally, the array (or arrays) would

be located near the model components most sensitive to launch acoustic loads, such as the payload fairing. The selected location was the best compromise that identified sources radiating both upward toward the payload and laterally toward the tower, and that also satisfied the noninterference condition. For the horizontal, motor-only test, the array with the protection chamber was placed 15 ft away from the nozzle centerline (Fig. 4). The array bottom edge was elevated 2 ft above the ground surface, and the array itself was rotated about its y axis by 30 deg, so that the camera had a clear view of the entire plume. For all other ASMAT tests involving the model and the launch pad, the array chamber was mounted on a tall tower (Fig. 5) about 15 ft away from the test stand and 16 ft above ground. The tower was secured via outriggers bolted to the concrete floor. Additionally, tether cables were used to increase torsion rigidity of the assembly. The tower with the array chamber was erected typically on the morning of each test and taken down after the test. Before erecting the tower, the array protection chamber was opened and each microphone was inspected and calibrated. A 72 channel, 16 bit VXI system was used for simultaneous acquisition of the microphone signals. The personal computer interfacing with the data-acquisition system was accessed over a local network from the nearby control room. The same PC was used for capturing video signal from the array camera, which was connected via a FireWire extender. The data acquisition was started manually at T minus 10 s and lasted for 30 s. Each microphone channel was simultaneously sampled at 49,152 samples per second for a usable bandwidth of 20 kHz, corresponding to a full-scale bandwidth of approximately 1 kHz. The actual motor burn lasted for about 6 s, out of which the first 2 s was the full-power burn. Analysis of signals from the Kulite transducers showed that the array plate was subjected to an IOP wave with a pressure rise of ∼0.6 psi. The average acoustics level varied from test to test in the range of 135 to 150 dB. The array plate also experienced water spray and impingement from fine debris. Nonetheless, the setup performed remarkably well over the year and a half of test duration. All together, about five sensors were lost. The most problematic part was the VXI data-acquisition system, which was found to be sensitive to temperature fluctuations; various components had to be changed frequently. Some smaller problems due to vibration and variable light condition were fixed using better mounts and a camera aperture control system described earlier. B. Array Software

For the present paper, a conventional beamforming procedure was used. The region seen in the photographic image from the array camera was divided into a grid mesh. The steering vectors wj from each of the grid points were established. The microphone data were used to calculate the cross-spectral matrix G, followed by an application of the classical beamforming equation [13,14]: bjj f

w†j SGST wj P m sm 2

(1)

The † symbol represents a complex conjugate and transpose operation, while the superscript T represents only transpose. The diagonal elements of G were deleted to improve the signal-to-noise ratio. The row matrix S containing weight applied to individual microphones [14] was found to be useful in improving spatial resolution. A linearly increasing weight Sm based on the radial location Rm of the microphone was used: Sm S1 1 − W 1 Rm ∕R0

Fig. 5 Setup of the microphone phased array at the ASMAT test stand. ASMAT 3: model at hold-down position (photo credit: NASA MSFC imaging team).

(2)

The constant W 1 was 0.1; R0 represented the radius of the outermost microphone. It is worth mentioning here that the noise source results presented in this paper are subjected to the inherent assumptions of the conventional beamforming process: that the complex noise sources can be modeled as sum of many monopoles, and sound propagation is strictly linear. The beamforming schemes were implemented in the Matlab platform. A quad-core personal computer was used along with the

1005

PANDA AND MOSHER

Matlab Distributed Computing Toolbox. Once the cross-spectral matrix was calculated, the rest of the beamforming operation was executed fairly quickly. To facilitate direct identification of the noise sources, the interrogation grid for beamforming was created over a photograph of the region of interest. The photograph, captured from the video recording, required a correction for barrel distortion due to the fish-eye lens used with the camera. To select the number of grid points, the Rayleigh criterion was applied to the highest frequency of interest. According to this criterion, for a circular aperture of diameter d, the angular resolution in radians θR is directly related to the wavelength λ of interest:


sin θR 1.22

λ d

(3)

For the present estimate, the maximum separation between any pair of microphones was assumed to be d. This equation shows that, for a given array size, the angular resolution improves with frequency. Conversely, the highest frequency of interest dictates resolution of the interrogation grid. In this case, a uniformly spaced grid of 1/10 the array resolution at the highest frequency of interest was used. The conventional beamform calculations were applied on individual narrow bands of the cross-spectral matrix. Finally, attempts were made to apply some of the advanced beamforming algorithms to improve spatial resolution at the low frequency and to reduce the side lobes at high frequencies. Typically, such procedures were found to work well for compact noise sources. However, different sets of difficulties arose for different algorithms when attempts were made to use with the rocket test data. The test stand created a reverberant environment, and the noise sources were distributed; both of these seemed to create various problems in the attempt to deconvolve the point spread function. More work is in progress to achieve this goal. For this paper, only conventional beamform data are presented.

booster flowed through the LM and MLP exhaust holes to one side of the flame deflector and finally flowed out of the trench. It should be noted that, just like the shuttle launch pad, there is a gap between the bottom of the MLP and the top of the trench. The effectiveness of a variety of different water-suppression systems was evaluated during the ASMAT series. There were two sets of water systems: one specifically targeting the ignition overpressure that appeared as soon as the solid motor was ignited, and a second system to reduce the overall acoustic levels. The second system is relevant for the present paper. Similar to the Space Shuttle launch pad, there were three parts to the acoustic suppression system. The first part is the “hole water”, water injection through the hole in the Mobile Launch Platform; the second part is the “trench water”, water injection inside on the deflector and inside of the flame trench; and the third part is the “rainbird”, water injection on the top of the MLP. Besides the horizontal firing, there were 17 vertical tests. The array was used in six vertical tests, plus the horizontal motor-only test (Table 1). The first vertical test, ASMAT 3, represented a hold-down condition when the vehicle was on the pad during the ignition of the first-stage booster. For the rest of the ASMAT tests, the model was elevated to replicate different vertical positions of the vehicle. A critical element of the liftoff process is the sidewise drift that accompanies vehicle elevation. ASMAT replicated the maximum expected drift of Ares I for most of the test conditions. In Table 1, the elevation and drift values are nondimensionalized by the nozzle exit diameter D. In ASMAT 17, a situation was tested where the vehicle elevation involved no drift, such that the plume mostly passed through the hole in the MLP deck. Additionally, all water-injection systems were turned off. The hole and the trench water were turned on in ASMAT 11, and all three systems (hole, trench, and on-deck) were used in ASMAT 12. Note that, in ASMAT 11 and ASMAT 12, the model was moved sidewise (drifted). Therefore, a straightforward comparison with the ASMAT 17 was not possible. There were several cases tested to verify the effectiveness of the on-deck water injection; the phased array could be used with only one of them (ASMAT 12).

C. Ares I Scale Model Acoustic Test Facility

ASMAT was conducted at test stand TS116 at NASA Marshall Space Flight Center. It used a uniformly scaled (5%) launch pad, the Mobile Launch Platform (MLP), a service tower, and a simplified model of the Ares I vehicle (Fig. 5). The launch pad was similar to that used for the Space Shuttle, with two trenches and a center flame deflector. The single plume of the Ares I solid rocket booster impinged on one side of the deflector. The MLP configuration was somewhat different from that of the Space Shuttle, with one exhaust hole and the addition of a launch mount (LM) between the hole and the solid rocket booster nozzle exit. Among other purposes, the LM was expected to confine the plume within the exhaust hole. Part of the way through the testing program, it was concluded that the launch mount did not help the acoustic environment and reduced the effectiveness of the on-deck water-injection system. This led to the removal of the launch mount. The service tower stood between the Ares I vehicle and the exhaust plume flowing through the trench; the intention was to provide some blockage to the acoustic path. This idea was perhaps derived from the noise source map of Eldred and Jones (Fig. 1) [2]. The inline vehicle was modeled by a long cylinder with a variable diameter corresponding to the different zones of the full-scale vehicle. A solid rocket motor with ∼10; 000 lbf thrust was used to model the firststage booster. At the hold-down condition, the plume from the Table 1

III.

Results and Discussions

A. Validation of the Array Performance

The initial system check, calibration, and software validation were performed in the ARC acoustics laboratory; a detailed description can found in [15]. Additional in situ validations were performed before every test at the ASMAT test stand. An example of this validation work is shown in Fig. 6, where a single speaker is kept at a corner of the launch pad. The peak 10 dB range of the beamformed source strength is shown in each figure. The color scales show sound pressure levels over a 48-Hz-wide band centered at the indicated frequencies. The background photograph was captured by the video camera at the center of the array box. Recall that the array was about 16 ft above ground with a ∼45 deg angle. An examination of Fig. 6 shows the expected trend associated with conventional beamforming. Because of the inherent point spread function of the array (Fig. 3b), the spot size was very large at the lowest frequency, while at the highest frequency (10 kHz), side lobes appeared, which give the appearance of pseudosources. In general, smearing of the source is progressively reduced and the number of the pseudosources increases with frequency. There were a large number of reflective surfaces around the model. Besides the concrete floor and the metallic pad, there were large steel blast curtains behind the model. The presence of these reflective surfaces created images of the noise source and

Part of the ASMAT test matrix where phased array was used

Test Name Elevation H∕D Drift Y∕D Trench H2 O Hole H2 O Horizontal N/A N/A N/A N/A ASMAT 3 0 0 Off Off ASMAT 4 3.9 0.61 On On ASMAT 7 7.9 0.9 On On ASMAT 11 8.6 0.9 On On ASMAT 12 8.6 0.9 On On ASMAT 17 8.6 0 Off Off

On-deck H2 O N/A Off Off Off Off On Off

Comment — — With LM With LM With LM No LM No LM No LM


1006

PANDA AND MOSHER

Fig. 6 Validation of the phased array operation via identification of a single speaker source.

increased the population of pseudosources (side lobes). The “sweet range” was found to be between 2 and 6 kHz, (100 to 300 Hz fullscale), where the array resolution was reasonably good and the pseudosources were mostly lower than 10 dB. Most of the ASMAT test data presented in this report are in this frequency range. Using plume diameter D and exhaust velocity U for Strouhal frequency calculations, f 2500 and 5000 Hz correspond to St fD∕U of 0.2 and 0.4, respectively. Strouhal numbers are used for the rest of the figures in this report. Finally, there is a need to discuss uncertainties in the absolute magnitude of the noise sources. Note that the color scale next to each figure of the beamformed map shows the magnitude of the noise sources as measured at the array location. The uncertainties in these numbers arise due to three reasons: the inherent convolution of the true noise source with the array point spread function, the short

duration of the motor burn, and a loss of coherence among pairs of microphones. The first reason becomes clear from Fig. 3b. Even in an ideal situation, the summation process applied during beamforming smears the low-frequency noise sources and brings about side lobes that make the appearance of pseudosources. The steady part of the motor burn typically lasted for 2.2s (Fig. 7), which was used for most of the beamformed results. The small duration of the sample implied fewer available cycles at the lower frequencies and increased side lobes at the high frequencies. This is the second reason for uncertainty. The third reason is the inherent nature of the distributed noise sources. Data gathered from ASMAT tests showed a progressive loss in coherence for sensor pairs with longer separation distances. Such loss was also found to increase with frequency (Fig. 7b). Therefore, the averaging process of the beamforming technique led to source strengths that are lower than the levels seen in

Fig. 7 a) Time trace of pressure fluctuations from one of the microphones in the horizontal motor test, and b) coherence spectra between indicated pairs of microphones.

PANDA AND MOSHER

1007

all microphones. Such a comparison for ASMAT 3 below shows differences of 2, 7, 12 and 16 dB for frequencies of, respectively, 1, 2.5, 5, and 10 kHz. Nevertheless, it is expected that the uncertainties are similar between any two configurations. Therefore, the beamformed maps should be used for a comparative study among various ASMAT configurations.


B. Horizontal Rocket-Motor-Only Test

Fig. 8 a) Plume of a solid-rocket motor fired horizontally over a concrete pad, and b) beamformed noise source distribution.

the autospectra. No attempts were made to compensate for this loss. Also, an accurate estimation of the uncertainty is difficult. A somewhat crude estimation may be made by comparing the peak level in the beamformed map to that seen in the average spectra from

The setup used in the horizontal test was discussed earlier with Fig. 4. Unlike the photograph of that figure, the camera at the center of the array plate captured an image with a better definition of the internal structure of the plume (Fig. 8a). This frame, from the very start of the burn, shows a clear periodic shock pattern present in the underexpanded plume. The glow in the plume is due to afterburning. The smoke is believed to be mostly made of aluminum oxide powder. The three vertical lines are the poles used for holding microphones for a separate suite of instrumentation. Conventional beamform results are shown in Fig. 8b. The color scales represent source strength in a 96-Hz-wide frequency band centered at the indicated frequencies. The first observation to make is that there exist two distributed noise sources; one is right along the plume, and the other is a reflection on the concrete test floor. The hard concrete pad acted as a mirror to the plume sources. It is obvious that any attempt to make free-field acoustic measurements in a horizontal rocket stand needs to pay attention to this reflection. For some frequency bands where the interference between the two sources is constructive, the amplitude will increase, while destructive interference will decrease the net level at other frequencies. Another interesting observation is the very long spatial extent (greater than 30 diameters) of the noise sources that extends to the end of the visible afterburning core of the plume. The quasi-periodic shock pattern of the plume is found to create periodic modulation of the noise source. An earlier conference publication [15] provides beamform maps at other frequencies (1000, 2500, and 10,000 Hz). Such maps show that the peak in the source distribution moves slightly closer to the nozzle exit with an increase in frequency;

Fig. 9 Beamformed noise sources at four indicated frequencies. ASMAT 3: model is at hold-down position, without water injection, and launch mount is in place.

1008

PANDA AND MOSHER

however, the downstream extent of the sources remains almost independent of frequency.


C. Ares I Scale Model Acoustic Test 3

Figure 9 shows the distribution of the noise sources at four different frequencies for a hold-down condition (i.e., at zero elevation). No water was used for this test. A photograph of this test showing the locations of the hot plume and the position of the array is shown earlier in Fig. 5. The photograph that serves as the background of Fig. 9 was collected before the test by the camera at the center of the phased array. As expected, the source distribution at the lowest frequency, St 0.08 (960 Hz), is excessively smeared due to the low resolution of the array, while at the highest frequency, St 0.8 (9984 Hz), the multitudes of side lobes are artifacts of the beamforming process. The best balance between the two extremes, the “sweet spot”, are the maps at St 0.2 (2500 Hz) and 0.4 (5000 Hz). The color scales show sound pressure levels over a 48-Hzwide band centered at the indicated frequencies. In general, Fig. 9 identifies two strong sources: 1) the gap between the MLP and the trench, and 2) the open part of the flame trench on the right side of the service tower. Additional video footage taken during the test (not shown here) verified the flow of very hot gas through the trench in the absence of any trench water. This hot plume and its impingement on the deflector are believed to be the origin of the above two sources. There was no leakage of hot plume out of the gap between the MLP and the trench. Therefore, it is believed that the first of the aforementioned sources originated inside the trench and then emanated through the gap between the MLP and the trench. Note that this gap existed on all four sides; therefore, the vehicle model was expected to be exposed to sound radiation from all four sides. A phased array detects only those sources radiating toward the array surface; that is why the beamformed plots of Fig. 9 identify only the front side of the MLP as the source. Closing this gap between the MLP and the trench should reduce the acoustic level experienced by the vehicle at the hold-down condition. An important observation can be made by comparing the noisesource model of Eldred–Jones (Fig. 1) and the actual, measured source distribution of Fig. 9. The rocket plume shot out of the right-hand side exhaust of the trench. The Eldred–Jones model assumes the entire plume length to be the noise source, while the beamformed map shows that the peak 10 dB range of the source is confined within the trench. The presence of the second source at the gap between the MLP and the pad also could not be envisioned from the Eldred–Jones formulation. The differences may be more fundamental in nature. Flow-induced noise is generated via decay of turbulent eddies. The manifestation of this fundamental physics is very different for a plume flowing without obstruction and a plume impinging on a surface. In freeflowing plumes, turbulence decay is very slow; therefore, noise generation occurs over a long, distributed region (Fig. 8) [16], which is measured in tens of nozzle diameters or in multiples of the length of the potential core. Eldred and Jones [2] essentially modeled this distribution as a slowly varying function. On the other hand, when a plume impinges on a solid surface, a rapid change in turbulence occurs within a short distance in the vicinity of the impingement zone. This leads to a very compact noise source. Figure 9 shows that, in a launch configuration, the primary noise sources are compact and mostly located at the impingement zones. The differences between the Eldred–Jones model and the actual source distribution became more distinct in the subsequent tests. D. Ares I Scale Model Acoustic Test 4

A photograph from this test is shown in Fig. 10. For this test, the model was lifted by 3.9D above the launch mount, thereby exposing a part of the hot plume above the launch pad. More importantly, the model was also drifted toward the service tower by about 0.61D. This allowed part of the plume to spill out of the hole and impinge on the LM, the MLP, and the base of the service tower. The liftoff trajectory for the Ares I vehicle dictated this drift; the model-scale static test merely replicated this feature. The other significant part of the test configuration is the water injection. A large volume of water was

Fig. 10 Sparks fly off the plume impingement point on the LM and MLP in ASMAT 4: model lifted by 3.9D above the Mobile Launch Platform.

injected inside the trench, inside the MLP hole, and on the top of the flame deflector via ducts built on the pad. The water flow was initiated before motor ignition. Figure 11 shows the noise source locations identified at three different frequencies. Unfortunately, the video obtained from the array camera was underexposed. Nevertheless, the outline of the trench and the MLP are visible. The bottom part of the model is visible as the white strip at the top center of each photo; as marked, the nozzle exit lies just below this strip. An examination of the source maps shows that the primary noise source extended from one side of the launch mount to the base of the tower. This was the region where the plume from the rocket motor impinged on the MLP platform. The beamformed source map shows that it is the impingement of the plume that is responsible for most of the noise generation. Note that a large part of the plume passed through the trench; however, unlike ASMAT 3, water injection inside the MLP hole and the trench weakened the source so much that it was not visible in the top 10 dB range shown. Finally, the noise map has no resemblance to the Eldred–Jones model [2] of Fig. 1, which completely misses the impingement zones as liftoff noise sources. E. Ares I Scale Model Acoustic Test 17

Once again, in ASMAT 17, the sound-suppression systems were turned off (completely dry), and the model was held at 8.6D above the MLP platform. Additionally, there was no sidewise drift, so that the maximum possible part of the plume was passing through the MLP hole. Figure 12 is a frame from a video camera that captured a side view of the setup. This figure shows the bare part of the plume (seen through the support tower) above the MLP as well as the exhausting hot gases through the trench. The beamformed maps of Fig. 13 show a U shape. A closer examination reveals that the noise sources are at three regions that are connected: 1) like ASMAT 3, the gap between the MLP and the trench; 2) like ASMAT 3, the open part of the flame trench on the right side of the service tower; and 3) unlike ASMAT 3, the region around the MLP hole. The absence of hole and trench water brought the first two sources back to prominence. The third region is due to the plume impingement around the hole of the MLP deck. The hole was designed to be fairly small to minimize IOP reflection. In spite of no sidewise drift of the model, the edges of the plume impinged on the rim of the hole and created the loudest noise source. Note that the background photo used to superimpose beamformed map in Fig. 13 was from a video taken before the motor ignition. The ladder, man lift, amplifiers, etc., were removed before the test.


PANDA AND MOSHER

1009

Fig. 11 Beamformed noise-sources at the indicated frequencies measured in ASMAT 4. F. Ares I Scale Model Acoustic Test 11

The hole and the trench water system were turned back on in ASMAT 11. The model was kept at the same elevation but was drifted per trajectory information. Figure 14 is a front view of the test with the

Fig. 12 Side view of the model and the launch pad in ASMAT 17 (no water; elevation is 8.6D; drift is 0; photo credit: NASA MSFC imaging team).

phased array setup in the foreground. The beamformed maps of Fig. 15 verify the effectiveness of the water system; the trench and the gap around the MLP are no longer in the source maps. However, compared to ASMAT 17, a larger part of the plume impinged on the deck, resulting in an increase of the peak source strength (see color bar). In other words, compared to ASMAT 17, the spatial extent of the sources is much smaller, yet the peak level is higher.

Fig. 13 Beamformed noise sources from ASMAT 17 at two indicated frequencies.

1010

PANDA AND MOSHER


Fig. 16 Photograph from ASMAT 12: rainbird (on-deck) water injection. G.

Fig. 14 Photograph of ASMAT 11 (drift is 0.9D; elevation is 8.6D; hole and trench water; photo credit: NASA MSFC imaging team).

Fig. 15 Beamformed noise source maps at indicated frequencies from ASMAT 11.

Ares I Scale Model Acoustic Test 12

All three parts of the water-suppression system (hole water, trench water, and rainbird) were turned on in ASMAT 12. The vehicle elevation and drift were the same as in ASMAT 11. A very large flow of water was used for this test. Unlike a real liftoff situation, the rainbird system was turned on before the motor ignition. Thus, there was a pool of water accumulated on the top of the MLP when the motor was ignited. Figure 16 shows the plunging of the plume in this pool, resulting in a large splash of water. A comparison of the microphone time traces between this and ASMAT 11 (Fig. 15) show observable attenuation of the acoustic level. The blown-up time traces show that the “N waves” in the Mach wave radiation were particularly affected; the high positive peaks of the sound waves show good attenuation from the rainbird water. Figure 17c also indicates a time-dependent attenuation. Just after ignition, when the plume plunged into the accumulated pool, there was a large attenuation; however, the levels increased significantly after a fraction of a second, when the accumulated pool evaporated and the plume made a hard impingement on the MLP deck. This is evidenced by the beamformed maps shown in Fig. 18, which identify compact noise sources around the impingement point. The peak source strength in Fig. 18 is only a couple of decibels lower than that of the no-rainbird case (Fig. 15). A comparison between the two figures shows that a benefit of the rainbird system is to reduce the size of the distributed source. This may be interpreted as an attenuation of the peripheral

Fig. 17 Time traces from an absolute pressure transducer on the array plate: a, b) ASMAT 11, and c, d) ASMAT 12.

1011


PANDA AND MOSHER

Fig. 18 Beamformed noise source maps at the indicated frequencies from ASMAT 12 (drift is 0.9D; elevation is 8.6D; hole, trench, and on-deck water).

sources caused by the deflected part of the plume re-impinging on the service tower. Discussions so far have been centered on the location and strength of the noise sources as seen by the phased array. The actual acoustic levels are due to an integration of radiation from all such sources. A comparison of the acoustic spectra from the five ASMAT tests discussed previously is shown in Fig. 19. Each spectrum in this figure is an average from all 70 microphones present in the array. It is difficult to make a direct comparison because more than one parameter was changed between most pairs. Nonetheless, a comparison between the no-LM cases (ASMAT 17, 12, and 11) shows that the water-suppression system produced suppression in almost all frequency ranges. Compared to ASMAT 12, the addition of the rainbird water in ASMAT 11 produced a 3.4 dB reduction of the

Fig. 19 Comparison of average spectra from the indicated tests. The dB values in the legend are differences in OASPL from the reference case.

overall level at the array location. A comprehensive discussion of the effectiveness of the water-suppression system on the model and the service tower will come out in separate reports.

IV.

Conclusions

A phased array composed of seventy 0.25-in. condenser microphones, was used in Ares I Scale Model Acoustic Tests at NASA Marshall Space Flight Center. Many precautions were taken to protect the hardware from the harsh rocket test environment. The first test involved a motor-only burn. For the next six tests, the array was set up in the vicinity of a 5% scale model of the Ares I vehicle in a launch configuration. The tests involved static firing at different model configurations representing various stages of liftoff. The microphone phased array provided unprecedented insights into the noise sources during liftoff of a rocket vehicle. The following is a list of significant findings: 1) The beamformed noise maps verified that, in the unobstructed plume of the motor-only test, noise sources were distributed over a very long distance along the plume. The top 10 dB range was found to spread over 30 nozzle diameters. This length was only weakly dependent on frequency. The sources were also modulated by the shock cells that were present in the plume. This distribution of the nonimpinging jet corresponds to the slow variation that is the basis of the prediction by Eldred and others. 2) In the launch configurations, on the other hand, the primary noise sources became remarkably more compact. Instead of the plume itself, the impingement zones on various locations of the launch pad were found to be the primary sources. As the vehicle was elevated, a part of the plume spilled outside the hole of the launch platform, creating an impingement zone and, in turn, the loudest noise source. Lateral drift of the vehicle further strengthened plume impingement. Compact noise distributions associated with plume impingement are not adequately represented by the acoustic source models of Eldred and others. 3) The beamform results obtained from different configurations indicate that the reduction in liftoff acoustics is tied to minimization of plume impingement on the launch pad. This can be attained via


1012

PANDA AND MOSHER

minimization of vehicle drift in the early part of liftoff, removing extraneous components such as the launch mount, and possibly increasing the size of the hole on the Mobile Launch Platform (MLP). 4) When no water was used for attenuation, it was observed that the uncovered part of the trench and the gap around the MLP platform were major noise sources. 5) The noise sources inside the trench were found to be significantly attenuated via water injection in the MLP hole, on the top of the deflector, and inside the trench. 6) Results obtained from one test involving injection of the ondeck water (“rainbird”) shed light into the cause of noise reduction. The rainbird was found to attenuate noise sources only around the periphery of the primary plume impingement zone. The beamformed maps demonstrated that the motor plume displaced water quickly and brought back the impingement source; however, the spatial extent of the source was reduced. 7) It was found that the current liftoff model (NASA SP-8072 [2]) fails to account for the impingement noise sources. There is a need to update this model. The 40-in.-diam. array was found to be most effective in resolving noise sources above 2 kHz, which was satisfactory for the present model-scale test. For application in a full-scale launch, where frequencies of interest are below 2 kHz, the array either needs to be placed close to the pad or needs to be larger in size. Use of advanced beamforming algorithms provided limited success but needed to be investigated further. For the present test, the array was placed away from the model. Because the noise sources are directional in nature, placement of the array closer to the model will make the noise source maps more relevant for the vehicle development program. Finally, the plume of steam was found to obscure the path of the motor plume; an infrared camera would have improved the plume visibility.

Acknowledgments This work was supported by NASA Engineering Safety Center technology development project TI-09-00597 with Roberto Garcia of NASA Marshall Space Flight Center (MSFC) as the technical monitor. We greatly acknowledge the constant help from a large number of engineers and technicians led by Dennis Strickland of NASA MSFC in setting up the array. We would also like to thank Allison Lee of NASA MSFC for help with the logistics of the operation and Janice Houston of Jacobs Engineering for coordinating the test operations.

References [1] Haynes, J., and Kenny, J. R., “Modifications to the NASA SP-8072 Distributed Source Method II for Ares I Lift-Off Environment Predictions,” 15th AIAA/CEAS Aeroacoustics Conference, AIAA Paper 2009-3160, 2009.

[2] Eldred, K. M., and Jones, G. W., Jr., “Acoustic Load Generated by the Propulsion System,” NASA SP-8072, 1971. [3] Varnier, J., “Experimental Study and Simulation of Rocket Engine Free Jet Noise,” AIAA Journal, Vol. 39, No. 10, 2001, pp. 1851–1859. doi:10.2514/2.1199 [4] Plotkin, K. J., Sutherland, L. C., and Vu, B., “Lift-Off Acoustics Predictions for the Ares I Launch Pad,” 15th AIAA/CEAS Aeroacoustics Conference, AIAA Paper 2009-3163, 2009. [5] Sutherland, L. C., “Progress and Problems in Rocket Noise Prediction for Ground Facilities,” 15th Aeroacoustics Conference, AIAA Paper 1993-4383, 1993. [6] Dougherty, N. S., Nesman, T. E., and Guest, S. H., “Shuttle SRB Ignition Overpressure: Model Suppression Test Program and Flight Results,” Proceedings of the JANNAF 13th Plume Technology Meeting, Defense Technical Information Center, Laurel, MD, April 1982. [7] Dougherty, N.S., and Guest, S. H., “A Correlation of Scale Model and Flight Aeroacoustic Data for the Space Shuttle Program,” 9th AIAA Aeroacoustics Conference, AIAA Paper 1984-2351, 1984. [8] Gély, D., Elias, G., Bresson, C., Foulon, H., and Radulovic, S., “Reduction of Supersonic Jet Noise—Application to the Ariane 5 Launch Vehicle,” 6th AIAA/CEAS Aeracoustics Conference, AIAA Paper 2000-2026, 2000. [9] Fukuda, K., Tsutsumi, S., Fujii, K., Ui, K., Ishii, T., Oimuna, H., Kazawa, J., and Minesugi, K., “Acoustic Measurement and Prediction of Solid Rockets in Static Firing Tests,” 15th AIAA/CEAS Aeroacoustics Conference, AIAA Paper 2009-3368, 2009. [10] Tsutsumi, S., Kato, S., Fukuda, K., Takaki, R., and Ui, K., “Effect of Deflector Shape on Acoustic Field of Launch Vehicle at Lift-Off,” 47th AIAA Aerospace Sciences Meeting, AIAA Paper 2009-328, 2009. [11] Sankaran, S., Ignatius, J. K., Ramkumar, R., Satyanarayana, T. N. V., Chakravarthy, S. R., and Panchapakesan, N. R., “Suppression of High Mach Number Rocket Jet Noise by Water Injection,” Journal of Spacecraft and Rockets, Vol. 46, No. 6, 2009, pp. 1164–1170. doi:10.2514/1.43421 [12] Burnside, N. J., Jaeger, S. M., Reinero, B. R., Horne, W. C., and Soderman, P. T., “Array Design and Performance for a Large Scale Airframe Noise Study,” 8th AIAA/CEAS Aeroacoustics Conference and Exhibit, AIAA Paper 2002-2576, 2002. [13] Underbrink, J. R., “Aeroacoustics Phased Array Testing in Low Speed Wind Tunnels,” Aeroacoustics Measurements, edited by Mueller, Thomas J., Springer, Berlin, 2001, pp. 98–217. [14] Brooks, T. F., and Humphreys, W. M. Jr., “A Deconvolution Approach for the Mapping of Acoustic Sources (DAMAS) Determined from Phased Microphone Arrays,” 10th AIAA/CEAS Aeroacoustics Conference, AIAA Paper 2004-2954, 2004. [15] Panda, J., and Mosher, R., “Use of a Microphone Phased Array to Determine Noise Sources in a Rocket Plume,” 49th Aerospace Sciences Meeting, AIAA Paper 2011-974, 2011. [16] Panda, J., and Seasholtz, R. G., “Experimental Investigation of Density Fluctuations in High-Speed Jets and Correlation with Generated Noise,” Journal of Fluid Mechanics, Vol. 450, Jan. 2002, pp. 97–130. doi:10.1017/S002211200100622X

M. Costello Associate Editor