Pieter Sijtsma

CLEAN Based on Spatial Source Coherence

To obtain higher resolution acoustic source plots from microphone array measurements, deconvolution techniques are becoming increasingly popular. Deconvolution algorithms aim at identifying Point Spread Functions (PSF)in source plots, and may therefore fall short when actual beam patterns of measured noise sources are not similar to synthetically obtained PSFs. To overcome this, a new version of the classical deconvolution method CLEAN is proposed here: CLEAN-SC. By this new method, which is based on spatial source coherence, side lobes can be removed of actually measured beam patterns. Essentially, CLEAN-SC iteratively removes the part of the source plot which is spatially coherent with the peak source. A feature of CLEAN-SC is its ability to extract absolute sound power levels from the source plots. The merits of CLEAN-SC were demonstrated using array measurements of airframe noise on a scale model of the Airbus A340 in the 8 ×6 m2 closed test section of DNW-LLF.

Quantification of airframe noise using microphone arrays in open and closed wind tunnels

The reliability of the phased array technique for quantifying airframe noise was assessed in the DNW-LLF open and closed wind tunnel sections. Acoustic measurements were performed on wing noise from a 1:10.6 scaled Airbus A340 model, using a 4-m diameter out-of-flow microphone array in the open jet and a 1-m diameter wall array in the closed test section. Apart from the differences in array resolution, the source characteristics are generally similar for both test sections. The open jet results show that, although the absolute integrated array level can be too low due to e.g. coherence loss, the relative sound levels determined with the array (i.e. differences between configurations) are accurate within 0.5 dB. The difference between the absolute array levels in the closed test section and the farfield levels in the open jet is smaller than 3 dB for all frequencies. The relative array levels in the closed test section agree with the open jet within 1 dB, provided that the flow conditions - and therefore the noise source characteristics -are similar.

Using phased array beamforming to identify broadband noise sources in a turbofan engine

Fan broadband noise is a major component of the total noise emitted by turbofan engines, especially at lower shaft speeds. It is generated in the rotor/stator region, but the exact origin is not always known. This article discusses the application of phased array beamforming techniques for a better understanding of the source mechanisms of fan broadband noise. The Conventional Beamforming technique was applied, as well as the deconvolution technique CLEAN-SC and the beamforming technique ROSI for rotating sources. Beamforming was applied to acoustic data measured by two circular microphone arrays that were mounted in the intake and in the bypass of a Rolls-Royce fan rig. These arrays are normally used for the detection of azimuthal modes. The merits of beamforming are discussed by considering a number of typical low shaft speed cases. Using the intake array, in one of the cases forward radiating broadband noise sources were found that were coherent over a large area. These could have been due to a rotor instability. In an other case, the forward radiating broadband noise seemed to have its origin at the stator vanes. This could be made plausible with the help of the intake mode detection results. Using the bypass array, stator bound noise sources were found that seemed to be distributed along the span of the vanes. In other words, tip sources seemed to be of minor importance for aft radiating fan broadband noise. The bypass mode detection results seemed to indicate that the spanwise distributed noise sources were located at the trailing edges of the stator vanes.

Feasibility of In-Duct Beamforming

*This paper discusses the possibility of source loca tion by phased array beamforming on fan and stator of turbofan engines using an intake wall-mounted microphone array. To demonstrate the feasibility, beamforming techniques were applied to existing measured data of a circular microphone array in the intake of a f an rig. This array is normally used for azimuthal mode detection. From the measured data set a test case was selected with low engine speed, and with an intake liner between the fan and the array. Beamforming methods with both stationary and rotating focus are applied , and the contributions of tonal noise and broadband noise were separated. Thus, tonal and broadband noise sources can be identified on both fan and stator. The free-field Green’s func tion is used for the definition of the steering vectors, thereby ignoring duct wall reflec tions. By a simulation study it is shown that the presence of an intake liner is a necessary condition for obtaining meaningful beamforming results. Further simulations showed that significantly improved beamforming results can be obtained with an in-duct “cage” arra y consisting of a number of parallel microphone rings.

Source Location by Phased Array Measurements in Closed Wind Tunnel Test Sections

The feasibility of high frequency phased array measurements on aircraft scale models in a closed wind tunnel test section was investigated. For that purpose, 100 microphones were built in a 0.6x0.5 m’ plate. which was installed in a floor panel of the 8x6 m2 test section of the Large Low-speed Facility of the German Dutch Wind tunnel (DNW-LLF). F or the microphone positions a sparse array design was used that minimises side lobes in the beamforming process. To suppress boundary layer noise, the array could optionally be covered with a 0.5 cm thick layer of acoustic foam and a 5% open perforated plate. To assess the effect of wall reflections, tests without wind were performed with a loudspeaker at several positions in the tunnel section. Furthermore, wind tunnel tests were carried out on an Airbus transport aircraft model. It is shown that location of acoustic sources is indeed possible for frequencies between 2 and 30 kHz, but their levels may differ from those measured in an anechoic environment. For the lower frequencies, application of the layer of foam and the perforated plate is beneficial. Finally, it is shown that filtering out the most dominant source can extend the array potential.

CORRECTIONS FOR MIRROR SOURCES IN PHASED ARRAY PROCESSING TECHNIQUES

When an aero-acoustic source is close to a reflecting wall, results from conventional phased array beamforming techniques are disturbed by the nearby mirror source, which is coherent with the original source. The recalculated source spectrum tends to deviate from the spectrum of the same source obtained in an anechoic environment. A periodic modulation of the spectrum occurs, which is most prominent at low frequencies. In this paper, a number of non-conventional beamforming techniques to correct for this spectral modulation is investigated. First, a technique is discussed which adds the mirror source to the transmission model. It was found that this technique is not very suitable because of its lack of robustness. Then, a more robust beamforming technique is proposed that minimises the influence of a given mirror source. By this technique, much better results were found. Nevertheless, at low frequencies the method still suffers from lack of robustness. Finally, a modification to this minimisation technique is proposed which preserves the robustness. Using this “controlled minimisation” technique, the best agreement was found between the recalculated spectra of a source close to a wall and the same source in anechoic conditions. The beamforming techniques were applied to array measurements on a calibration source in the DNW Low Speed Wind Tunnel LST.

Mode detection with an optimised array in a model turbofan engine intake at varying shaft speeds

Modal measurement techniques in engine intakes have been used previously to analyse the generated fan noise. A proven method is to use a wall-mounted array of Kulite transducers and operate the (model) turbofan under constant shaft speeds. A drawback of this method is the large number of (expensive) microphones and acquisition channels needed to obtain complete m-mode spectra at high engine orders. Furthermore, to get a full scan of the m-mode spectra as a function of shaft speed, many measurements are required. The issue of the large number of microphones was addressed by using a sparse array instead of an equidistant array. An array optimisation technique, similar to a technique used for the design of phased microphone arrays for sound source localisation, was used to define such a sparse intake array. This array consists of 100 Kulites and is able to determine without aliasing the modal spectrum from m = -79 to m = -1-79, which is appropriate to determine the modal content up to 3 BPF of a modem turbofan. This array was tested in a Rolls-Royce model fan rig at Ansty as a part of the RESOUND project. A new digital data-acquisition system made it possible to simultaneously and continuously record the Kulite pressure data as the engine speed was varied continuously from idle to maximum speed or vice versa, with each acceleration/deceleration lasting for a period of 9 minutes. Time histories of the Kulites were processed giving power spectra of the engine orders, which revealed the rotor locked tonal components. For each rotor revolution, a Discrete Fourier Transform was applied and, after averaging over a number of revolutions, the m-mode spectra were determined. In this way, a full modal scan with respect to shaft speed in a very limited testing time was obtained. Paper presented at the 7th AIAA/CEAS Aeroacoustics Conference, 28-30 May 2001, Maastricht, The Netherlands.

Acoustic Array Measurements of a 1:10.6 Scaled Airbus A340 Model

Acoustic array measurements were carried out on a 1:10.6 scaled Airbus A340 model. Tests were done in both the open jet and a closed test section of the DNW-LLF wind tunnel. The purpose of these measurements was to investigate several noise reduction concepts for high-lift devices (slats and flaps) in landing configuration. The possibilities and limitations of arrays for the determination of quantitative results are discussed. Besides the identification of dominant noise source regions with conventional beamforming, local source spectra were determined using a power integration method. For the open jet results, the integration method was applied with and without the ’diagonal removal’ (DR) technique, in which the main diagonal of the cross-power matrix is discarded. It is shown that application of DR results in meaningful local spectra, whereas without DR the results are obscured by the influence of the main diagonal. On the other hand, by comparing integrated spectra with absolute sound levels on farfield microphones, it is shown that the application of DR, in combination with coherence loss, results in significantly reduced absolute levels. However, while the absolute sound levels can be too low, level differences between configurations can be accurately determined under certain conditions. From the closed test section array results, dominant source regions are identified as a function of frequency and angle-ofattack. These results are quantified by application of the integration method to several source areas on the wing. The local and overall effect of noise reduction devices is assessed. The effect of coherence loss on the absolute levels is investigated by varying the effective array size.

Experimental Validation of Lifting Surface Theory for Rotor-Stator Interaction Noise Generation

An important element in the National Aerospace Laboratory NLR prediction model for rotor-stator interaction noise is the unsteady lifting surface theory for a rotating blade row. To validate this lifting surface theory, a wind-tunnel experiment has been carried out using a through-flow nacelle with a hub, a 16-bladed rotor, and an 18-vane stator. Two mechanisms of sound generation were recognized. As usual in rotor-stator configurations, sound was produced by interaction of rotor viscous wakes with the downstream stator vanes. In addition, due to the small gap between rotor and stator, sound was generated by interaction of the stator vanes, with flow distortions caused by the thickness of the rotor blades and vice versa. Trip wires were attached to blades and vanes to simulate high-Reynolds-number flow. Unsteady pressures were measured on a stator vane, on a rotor blade, and in the inlet. Experiment and theory are briefly described, and results from both are compared. For most of the measured cases, the agreement is good.

In-Duct and Far-Field Mode Detection Techniques

This paper treats 4 different techniques for azimut hal mode detection by circular microphone arrays inside a flow duct, and in the fa r field around a free jet. In-duct and far- field applications are considered separately becaus e of the different background noise properties. The mode detection techniques were applied to experimental data from tests in the NLR small anechoic wind tunnel facility KAT. These tests were carried out using a convergent wind tunnel nozzle of circular cross-sec tion with diameter varying from 80 cm in the upstream part to 20 cm in the downstream part. At the nozzle exhaust a free jet was generated at Mach numbers up to 0.7. In the upstream wide part of the nozzle acoustic modes were generated by the mode synthesizer of EADS-IW. Tonal sound was generated by a circular phased array of 30 actuators, which prov ided azimuthal modes in the range of m = - - - -14 to m = +14. Inside the nozzle, near the exhaust, azimuthal mode spectra were determined with a circular array of 64 flush-mounted transducers. In the far-field, around the jet, mode spectra were measured with a 2 m diameter traversable circular array of 92 microphones. It is argued that the best mode detect ion results are obtained with a technique based on cross-correlation with a reference signal, or with a technique based on principal component analysis. The mode measurements showed the ability of the mode synthesizer to generate prescribed modes, which was one of the main objectives of the tests.