Antoine Moreau

A comparison between measured and predicted fan broadband noise due to rotor-stator interaction

This paper presents a comparison between predictions and measurements of the broad-band noise generated in a low-speed fan rig. It is assumed here that the dominant generation mechanism is due to rotor-stator interaction. In the broadband noise model the stator is modelled as a cascade of flat plate airfoils on which turbulent wakes due to the rotor impinge. The turbulence transverse length scale and intensity of the rotor wake are determined by data from measurement and RANS calculations. The theoretical expression for the acoustic power spectra is shown to be in the form of the product of an aerodynamic response function and the spectrum of the velocity fluctuations normal to the stator blade direction. The response function is obtained by solving numerically an aerodynamic integral equation, while the turbulence velocity spectrum is expressed analytically in a form similar to an isotropic homogeneous turbulence model. The turbulence is also modelled by random velocity fluctuations superimposed on a statistic distribution of the rms velocity. The statistical distribution is composed of a constant term, modelling the background turbulence and a periodic Gaussian-profile to represent the rotor-wake turbulence. Extensive hot wire measurements of both the steady and unsteady wake velocity was measured. Comparisons between these measurements and RANS predictions are also presented.

Open-rotor noise prediction with a RANS informed analytical method

Analytical models are informed by a steady-state RANS calculation with mixing plane in order to predict turbomachinery noise with a short computing time; the method is tested on the example of a contra-rotating open-rotor propulsion system at take-off condition; the results are compared to these of an unsteady RANS calculation with the pressure far field given by the Ffowcs Williams–Hawkings solution for a porous integration surface. Regarding the rotor-alone tones, the sound power level agrees within 0.5 dB between the two methods; for the interaction tones, the discrepancy is around 2 dB reaching up to 10 dB locally on the directivity. So far the sound-generating mechanisms considered by the acoustical models are the steady loading and thickness tonal components and the unsteady periodic loading of the aft rotor due to the interaction with the front-rotor wakes. Inherently, the prediction of interaction noise with the RANS-informed analytical approach lacks accuracy for two main reasons. First, the determination of the unsteady lift relies on a gust response function, the prediction of which ist not commonplace for real loaded blades; the analytical solution for a flat plate is applied in the present study. Second, the wake velocity profiles used as input for the gust response function are mixed out at the interface between the two rotors typically located in the middle of the rotor interval; ideally the values at the aft-rotor leading edge should be used. An evaluation of these two problems shall be undertaken before using the RANS-informed analytical method for aeroacoustic design optimisation.

Similarities of the free-field and in-duct formulations in rotor noise problems

This paper deals with the comparison between open rotor and ducted fan noise. Similar theoretical formulations for the freeeld and in-duct noise problems are derived from the original equation for noise generation given by Goldstein. The form of the nal equations enables to separate the terms related to noise propagation from those representing the acoustic source. Among the similarities shared by the freeeld and in-duct cases, the possibility to de ne a common cut-o ratio for the spinning modes provides some insight in the relation between the position of the sources and the excited modes in a way more general than that already proposed by the authors. With help of this formalism we hope to reach a better understanding of the di erences in noise levels to be expected on contrarotating open rotor (CROR) and contra-rotating turbofan (CRTF) concepts.

Development and application of a new procedure for fan noise prediction

The fan module of the DLR engine noise prediction program PropNoise is introduced in the present paper. By means of analytical models, the program gives a prompt visualization of both aerodynamic and acoustic performances associated with the fan. Single rotating and counter-rotating configurations can be treated. The fan module has two main components: the first is dedicated to the prediction of the steady and unsteady aerodynamic characteristics of the fan, while the second evaluates the noise generation and in-duct propagation. As shown in the paper, this program can be used to calculate performance and acoustic maps for a given fan and allows an acoustic source breakdown at each operating point. Furthermore the acoustic output at the presence of parameter variations provides guidelines for the design-to-noise of fan components. More generally, PropNoise provides helpful insights into the aeroacoustics of fan stages.

Radial Mode Decomposition in the Outlet of a LP Turbine - Estimation of the Relative Importance of Broadband Noise

The sound field in the outlet duct of a high speed low-pressure turbine with three stages was studied to deepen the understanding of its sound generating mechanisms. Special interest was given to the analysis of the two sound field constituents (tones and broadband noise). Three axial sensor arrays were mounted wall-flush downstream of the turbine stage in the annular duct section of the turbine exit. The arrays were positioned at three different azimuthal angles displaced by 120° and traversed azimuthally over 120 degrees in steps of 2 degrees to give a total of 4500 measurement points. Measurements were made at operating conditions from 68% to 93% rotor design speed. Special attention was given to the blade passing frequencies (BPF) of the three turbine rotors. The chosen experimental setup permitted their decomposition into azimuthal and radial modes. With this information, the tonal sound power transmitted upstream as well as downstream could be calculated. The mode analysis results provide a detailed view on the sound interaction processes between the turbine blade and vane rows. Finally, a novel broadband (BB) sound power determination method, previously validated against the ISO 5136 standard method for sound power determination 2 , was applied to the measurement data. The outcome of the BB sound power analysis permitted the comparison of the relative importance of LP turbine tonal and broadband noise.

Relation between source models and acoustic duct modes

The scope of this study is the noise generation by ducted axial fans. The work is chiefly concerned with the impact of source models on the excitation of the induct acoustic modes, and their eect on the total induct transfer function defined as the ratio of the induct to the free-field sound power. In the models, the sources are represented by static point sources (monopoles and dipoles) distributed along the stator vanes. These sources are assumed to be fully uncorrelated for broadband noise. Following Tyler and Sofrin, the sources are correlated and linear phase-shifted for the generation of rotor-stator interaction tones. The results show that the choice between monopole (volume) and dipole (force) sources does have a significant influence on the level of the excited induct modes. Whereas monopoles equally excite modes of positive and negative azimuthal order, the mode pattern generated by dipoles features a strong asymmetry between positive and negative modes. Based on the ray theory, it is shown that induct modes whose phase angle of propagations are perpendicular to the dipole axis are not excited. For both single and uncorrelated sources, the total induct transfer function tends asymptotically to one at high frequencies, which means that the influence of the duct on the total sound radiation can be neglected at high frequencies. This is not the case for correlated sources, as the total induct transfer function is proportional to 1/kR at high frequencies and independent of the nature of the source. This significantly diers from the radiation in free field.

Extrapolation of RANS flow data for imrpoved analytical fan tone prediction

The context of this work is the analytical prediction of fan noise based on flow data extracted from a steady-state RANS (Reynolds Averaged Navier-Stokes) simulation. In turbomachinery RANS simulations the flow field is circumferentially averaged at the interface (so-called mixing plane) between two successive blade rows with different rotation speeds. As a consequence no information about the wake and the potential field is exchanged via the mixing plane. Such information is however mandatory for predicting tonal interaction noise in a fan stage. A method to extrapolate the RANS flow solution beyond the mixing plane was derived by the authors in the past. The present study is dedicated to the application and assessment of that method on two realistic test cases: a contra-rotating openrotor at take-off condition, and the Ultra-High Bypass Ratio Fan designed by DLR at approach conditions. The extrapolation method is coupled to the in-house analytical tool PropNoise. The acoustic results are compared to highly resolved unsteady RANS calculations and for the UHBR fan additionally to experimental and empirical results. It is shown that a substantial increase in prediction accuracy is achieved by extrapolating the RANS data beyond the mixing plane. The results also show that the RANS-informed analytical noise prediction approach delivers very realistic estimations of the sound field.

RANS-informed fan noise prediction: separation and extrapolation of rotor wake and potential field

The context of this work is the analytical prediction of fan noise based on extracted flow data from a RANS (Reynolds Averaged Navier-Stokes) simulation. Due to the low computational cost this method is suitable for multi-disciplinary optimizations. In a RANS simulation with mixing plane the flow field is circumferentially averaged at the interface between two successive blade rows with different rotation speeds. Therefore no information about the wake shape downstream of the mixing plane is available. For the prediction of fan tonal interaction noise the wake characteristics are needed at the leading edge of the downstream blade row. A method is presented, which is able to extrapolate along a streamline the circumferential periodic part of the velocity field downstream of the mixing plane. Furthermore, the method is capable to separate the perturbations into their physical source mechanisms; these are currently the viscous wake and the potential field. As input the method needs the velocity field upstream of the mixing plane. As output it predicts the amplitude and phase of the harmonics as function of the axial distance to the trailing edge. This method was successfully tested for contra-rotating open rotors, contra-rotating fans and rotor-stator fan-stages at different operating conditions.

Measurements compared to analytical prediction of the sound emitted by a high-speed fan stage

Acoustic measurements have been performed to characterize the noise emission of a high-speed low-pressure-ratio fan stage at various operating conditions. In parallel, simulations of different tonal and broadband noise sources have been carried out based on an analytical prediction tool. Due to the presumably unclean inflow conditions of the fan, the experimental results do not allow to conclude clearly about the dominant broadband mechanism and show a strong tone at the first blade passing frequency harmonics despite the cut-off design of the fan. Still, comparisons of the tone levels are good and the analytical models provide indications about the different behaviour of each source as fan speed and fan loading are varied. The aerodynamic excitation pressure is introduced as a helpful quantity to understand the variations of broadband noise. An extension of the broadband excitation model is proposed to account for the interaction of stretched eddies with the rotor.

Comparison between measurements and predictions of fan broadband noise on a low speed fan rig

This paper presents a comparison between measurements and predictions of the fan broadband noise in a low speed fan rig. It is assumed here that the broadband noise is due to the interaction between turbulence generated in the wake of the rotor and the downstream stator vanes. The contribution due to the background turbulence level is also included in the analysis. The stator vanes are assumed to be unloaded flat plates and the wakes are assumed to satisfy two-dimensional plane wake theory. The wake turbulence velocity correlation function is assumed to be periodic in the blade pitch direction (i.e., cyclo-stationary). Aerodynamic input data to the model, both steady and unsteady, is taken directly from hot wire velocity measurements. Comparisons between predicted and measured sound power spectra are in reasonable agreement with spectral shape being particularly well captured by the model. We first review the details of the broadband noise prediction model. Hot wire velocity data will then presented by way of validating the model simplifying assumptions, and to provide the input data to the model. Finally, comparisons between the measured and predicted sound power spectrum will be presented. The work reported here was undertaken as part of the European project, PROBAND.