19 th INTERNATIONAL CONGRESS ON ACOUSTICS MADRID, 2-7 SEPTEMBER 2007

Transcription

1 9 th INTERNATIONAL CONGRESS ON ACOUSTICS MADRID, -7 SEPTEMBER 007 INTERIOR ACTIVE NOISE CONTROL IN TURBOFAN AIRCRAFT: NUMERICAL SIMULATION AND EXPERIMENTAL VALIDATION FOR OPTIMAL ACTUATORS POSITIONING PACS: 3.50.Ki Monaco, Ernesto ; Franco, Francesco ; Iadevaia, Michele ; Lecce, Leonardo Department of Aerospace Engineering Università degli Studi di Napoli Federico II ; Via Claudio, Napoli, I-805, ITALY; ermonaco@unina.it ABSTRACT This paper presents the activities carried on by the authors within the research project named M.E.S.E.M.A. (Magnetoelastic Energy Systems for Even More Electric Aircraft) funded by the European Commission within the th Framework Program and coordinated by the Dipartimento di Ingegneria Aerospaziale of the University of Naples Federico II (DPA). MESEMA is a technology oriented research project which builds upon the success of previous EU projects with devotion to accomplish the objectives of the aeronautics priority through designing, producing and testing innovative transducer systems based on active materials. One of the main targets for the MESEMA Consortium consists in reducing the level of disturbance noise in turbofan aircraft. A noise & vibration control system using magnetostrictive actuators has been designed and developed, with the goal of controlling noise & vibrations in a frequency range between Hz. The environmental noise & vibration excitations is representative of a small/medium turbofan aircraft case. The paper will present numerical and experimental activities carried on within the three years of the project and the achieved results. Final results of the task is represented by 30 actuation/sensing devices and a system controlling external disturbances as well as actuators intrinsic non linearity. THE RESEARCH PROJECT This work presents the activities developed within the Research Project MESEMA funded by the European Commission (Ref.-3). Structural dynamics, acoustics, active materials and active control systems represent the scientific topics of the project with the goal of designing and developing five systems integrating vibration transducers based on active components. Among these a noise & vibration control system using magnetostrictive actuators will be designed, developed and tested, with the goal of controlling noise & vibrations in a frequency range between Hz. The primary noise & vibration excitations will be representative of a small/medium turbofan aircraft case. Final results of the task will be represented by a system made up of about 30 actuation/sensing devices connected to a aeronautical structure performing control of external disturbances as well as of the devices intrinsic non linearity. As experimental test article a fuselage mock-up of the ATR/7 aircrafts family has been chosen available at the acoustic laboratory in the Alenia plant; due to its geometry and overall dimensions it well represents a fuselage section of a generic regional jet (Fig. ). Figure. Mock-up of the ATR aircraft and constraining structure THE TEST ARTICLE: DESCRIPTION AND NUMERICAL MODELLING The experimental test-article (Fig. ), reproducing a real fuselage section, has been used in the untrimmed configuration, i.e. without interior furnishing and seats. The cylinder is closed at the two ends by heavy caps bolted to final flanges, and a proper spring system and fixture connection allow both vibration isolation and free

2 longitudinal deformation of the fuselage section. Figure presents the overall test article dimensions; the mock-up is made of seven frames and six bays. In designing the F.E. model, the authors took in account the typical flexural wavefield behaviour of cylinders (Ref.), including low frequency beam modes, intermediate frequency cylinder modes, Figure. Overall dimension of the fuselage mock-up and high frequency plate modes. In order to calculate the necessary maximum elements dimensions to reach an analysis frequency of about 000 Hz, it has been necessary to determine the number of wavelengths around the circumference (m) and the number of wavelengths along the fuselage (n) correctly. From Ref. it is possible to characterize the maximum value of n and of m in order to avoid spatial aliasing problems within analyses up to 000Hz. According to this preliminary analysis the structural part of the F.E. model is composed by 595 grid points and 5555 elements. The fluid part, representing the cabin cavity, has been modelled still dimensioning the acoustic mesh in order to permit maximum analysis frequencies of the acoustic model up to 000Hz. The fluid part of the model consisted in 05 grid points and 888 elements. THE EXPERIMENTAL ANALYSIS, CORRELATION WITH NUMERICAL MODEL AND ITS UPDATING The experimental tests were aimed to extract modal parameters of both structure and acoustic volume in order to allow the numerical-experimental correlation and the updating of the model. The structural natural frequencies and modes shapes where extracted from the experimental measurements up to 300Hz, but it was not possible to classify all the extracted modes since above 0 Hz the local panels dynamics is dominant over the global mode shapes and the geometric description of the global modes becomes more difficult. Figure 3. Structural F.E. model of the Mock-up 9 th INTERNATIONAL CONGRESS ON ACOUSTICS ICA007MADRID The numerical analysis results have been compared with those coming from experimental tests. The common results of modal testing is a set of modal parameters (resonance frequencies, damping and mode shapes), which characterise the linear dynamics of the structure. Different methodologies for the correlation analysis exist. Within this work the Modal Assurance Criterion (MAC) has been used: it compares all mode shapes in the numerical database with all mode shapes in the experimental database. The following equation is used: { }{, ψ } t { ψ n}{ ψ e} t t { ψ }{ ψ })({ ψ }{ ψ }) MAC( ψ n e ) = () ( n e e n where { ψ } n and { ψ } e are respectively the numerical and experimental eigenvectors (mode shapes). The initial model did not well represent the dynamic behaviour of the real structure for what concerns the natural frequency parameters. A model updating has been carried on by a sensitivity analysis by the mean of sensitivity coefficients. Table presents the natural frequencies values obtained experimentally as well as the corresponding evaluated by the mean of the numerical model once the updating procedure had been carried out. It is possible to notice, at least until 5Hz a good correlation of the natural frequencies values with maximum differences between numerical and experimental ones often below 0 or even 5 per cent. The spatial correlation was less successful due to the presence on the fuselage mock-up of distributed added damping materials (viscoelastic constrained layers) that probably reduced the response peaks of the some structural components at resonance frequencies; the evaluation of the structural damping was anyway beyond the scopes of this work, while the

3 Table. Exp. and num. modal parameters: correlation results Updated F.E. model FEA Hz EMA Hz Diff. % MAC correlation levels were at this point considered acceptable for carrying out the following activities. THE PRIMARY DISTURB FIELD DESCRIPTION As mentioned before the research project was primary oriented to solve the problem of interior noise inside turbofan aircraft; for these vehicles the main noise sources come from (Ref.5): Structural vibrations due to TBL excitation; Structure-borne sound due to vibrations originated by the engines; Local aerodynamic phenomena due to interactions between airflow and structural connections and/or specific arrangements. The first type mainly introduce in the cabin a random like wide-band noise which can be well controlled passively at list above 500Hz; the Figure. Mode shape pair comparison - FEA 7.Hz - EMA.9Hz - MAC 7% second type of disturbances is someway tonal but located at high frequencies (higher then typical turboprops BPFs) where it is possible often to damp vibrations again with passive strategies. Finally the last type of disturbance can introduce on the contrary noise patterns characterised by some hundreds Hz frequency extension with or without a fundamental tone in the middle of the frequency band. The main concern of Alenia Aeronautics was focused on disturbances having some local aerodynamic origins and characterised by a frequency pattern concentrated in the low-medium frequency range where passive control solutions (damping and soundproofing materials) are not really effective. Alenia Aeronautics requirements were to design and test a control system devoted to attenuate a primary noise field originating at the flap channels and measured during in-flight noise measurements carried on their aircraft. The levels of sound pressure into the aircraft have been acquired during the in-flight measurements; they permitted to recognise, beyond the three well known BPFs, a so called bump noise (Fig. 5) due to the aerodynamic effect described above, particularly the channel separating the wing from the flap. It is possible to notice the bump noise pattern in the frequency range between 00 and 00Hz; furthermore the same phenomenon has been observed at higher frequency bandwidths depending from the values of the trim velocity of the aircraft. It was also possible to correlate the bump noise pattern to a similar vibration pattern measured on some structural frames. In the F.E. framework the dynamic load was designed in order to get as analysis results the noise and acceleration spectra measured during in-flight tests within the two frequency ranges Hz and Hz. These two synthetic force spectra (Figures and 7) induces the primary vibration pattern and the interior noise field to be controlled. ACTIVE CONTROL SIMULATION Two control strategies have been simulated: the first one consisting in an Active Structural Acoustic Control (ASAC) aimed to reduce interior noise by controlling the corresponding structural vibrations on the fuselage section; the second one consisting in an Active Noise Control (ANC) aimed to reducing directly interior noise actuating the structural components, but without attempting necessary to reduce vibrations levels (Ref.-7). 9 th INTERNATIONAL CONGRESS ON ACOUSTICS ICA007MADRID 3

4 For what concerning the simulation of the actuators actions on the structure, the initial basic idea has been to focus on inertial actuators able to provide concentrated forces in their application point. As a consequence they have been modelled as single point force acting on the selected nodes of the F.E. model, neglecting their coupling with the dynamics of the overall fuselage structure. It has been chosen to optimize actuation locations and obtain required forces for each one of them contemporarily employing the well known optimization (pseudoinverse) approach proposed in Figure 5. Typical external noise spectrum Ref.7 and based on the minimization of the cost function J (see next equations) in selected Force (N) Frequency (Hz) Figure. The reference force spectrum at low frequency Force (N),8,,, 0,8 0, 0, 0, Frequency (Hz) Figure 7. The reference force spectrum at high frequency control points. J = N n= H w (ω ) = w w () n The previous formula reports the cost function J, where w n represents the response in terms of noise or vibration of the n-th control point. The response vector w is represented by the linear combination of the primary and control fields, that is: w w + R = (3) p F s RF where w p is the vector of the complex response due to the primary field, the S product defines the complex response vector due to the contribution of M secondary forces (R is the complex matrix (NxM) characterizing the responses in the N control points due to the application of M force of unitary amplitude and F S is the control forces complex vector). Finally, if the number of control points (N) is bigger than the number (M) of the force sources, the optimum control force vector FS is reported in the following equation (Ref.7): F S H H = ( R R) R w () p 9 th INTERNATIONAL CONGRESS ON ACOUSTICS ICA007MADRID

5 Employing this approach is possible, then, to evaluate the maximum theoretical response reduction in the selected control points corresponding to an assigned primary field and actuators positions configuration. Furthermore to each configuration it will be associated the complex force spectra required to each actuator in order to reproduce the controlled response level. It is possible to investigate both ASAC and ANC control approaches by considering respectively the responses vectors wp and w corresponding to the structural (acceleration) responses in the first case and to the noise responses in the second case. OPTIMAL CONTROL ACTUATORS PLACEMENT EMPLOYING GENETIC ALGORITHMS In order to select among the many possible sets of control actuator configurations an optimisation activity was required. The used optimisation method is based on genetic algorithms: it is well known that they represent a quite fast, not deterministic approach for selection among many possible solutions of a problem whose effectiveness can be measured by a score (Ref.8). For this analysis actuators potential locations were selected on frames or stiffeners of the two middle bays of the mock-up (Fig.8). The authors developed the genetic algorithm code in MATLAB framework. The score chosen for the scopes of this work has been defined as: Score = T (5) { p} { p} where {p} represents the vector of the controlled interior noise field in the selected control points. The vector {p} is obtained as combination of noise disturb field (primary field) and noise field produced by the control actuators: p = p + P () Figure 8. The potential control actuator locations (two middle bays left and right side) Figure 9. Mean interior noise reduction for the optimal actuators configuration l. f. disturb force field 9 th INTERNATIONAL CONGRESS ON ACOUSTICS ICA007MADRID 5 { } { } [ ]{ } p F S where [P] is the pressure transfer function matrix obtained as mentioned in the previous chapter and {F S } are the optimal control forces evaluated using the pseudo-inverse approach for the selected chromosome. The values of control forces vector {F S } are obtained by employing the Eq. (5). The employed performance parameter (Eq. ()) maintains the same formulation whatever control approach has been chosen, since the final goal remains the reduction of interior noise. Following are presented the results obtained by considering as primary disturb the two reference force spectra presented in Figures and 7 related to the 30 control actuators configuration. It is possible to notice the good predicted performances in terms of noise reduction related to the final control actuator configuration selected by the optimization algorithm up to 00Hz. An important parameter representing one of the main targets that the authors had to meet was the maximum control force value required from the actuators: this parameter represents in fact a key point of the design of the actuators that will be developed within the MESEMA. consortium. Figures 0 and present these values for each one of the 30 control actuators placed in their optimal locations. The force values are comparable to those provided by the already manufactured magnetostrictive actuators. MAGNETOSTRICTIVE CONTROL ACTUATOR CHARACTERISATION AND STRUCTURAL IMPEDANCE CORRELATION In order to characterise the LPA actuator type in terms of force performances once installed on the fuselage section, a measurement campaign has been carried out at the Alenia Aeronauics plant in.

6 Figure 0. Required control force values for the control actuators in their optimal configuration Figure. Optimal actuators placement configuration low frequency disturb force field Laser vibrometer has been employed in order to measure the vibration spectra on the centre flange of the actuator as far as on its seismic masses. It is possible from these simple measurements to calculate the inertial force spectra provided by the actuator to the host structure: from the laboratory characterisation of the devices resulted that the overall inertial force, F(ω), can be obtained scaling the seismic mass acceleration, a(ω), by a factor of 0.5: F ( ω) = a( ω) *0.5 (7) provided that [a]=[m/sec^] and [F]=[Newtons]. Measurements are in line with requirements obtained from the optimization analysis as is possible to notice from Fig. 3. Inertial Force amplitude (Fuselage stringer mounted) 0 8 Force Amplitude [N] 0 8 Figure. Laser tests set-up Frequency [Hz] Figure 3. Actuator inertial force spectra ACKNOWLEDGMENTS The authors wish to acknowledge and thank the MESEMA project consortium members. We also thank the European Community and the European Commission for the financial support of this activity under Contract N AST3-CT References: [] E. Monaco, F. Franco, C. Illibato, A Structural-Acoustic Coupled Model for Designing an ASAC on a Regional Jet Fuselage, Proceedings of Active 00 Conference Williamsburg, Virginia, September 00. [] G. Aurilio, A. Cavallo, L. Lecce, E. Monaco, L. Napolitano, C. Natale, Fuselage Frame Vibration Control Using Magnetostrictive Hybrid Dynamic Vibration Absorbers - Proceedings of Euronoise003 Conference, Napoli - ITALY, May 003. [3] (MESEMA Project web-site) [] L. Cremer and M. Heckl, Structure-Borne Sound - nd Edition, Springer-Verlag (988). [5] J. F. Wilby, Aircraft Interior Noise, Journal of Sound and Vibration, Vol. 90, 99, pp [] C. R. Fuller, S. J. Elliot, P. A. Nelson, Active Control of Vibration, Academic Press, London, (99). [7] P. A. Nelson, S. J. Elliot, Active Control of Sound, Academic Press, London, (993). [8] A. Ovallesco, L. Lecce, A. Concilio, Position and number optimization of actuators and sensors in an active noise control system by genetic algorithms, Proceedings CEAS/AIAA Aeroacoustics Conference 995. [9] C. May, P. Pagliarulo, H. Janocha, Optimization of a Magnetostrictive Auxiliary Mass Damper ACTUATOR 00 Conference, Bremen Germany, June th INTERNATIONAL CONGRESS ON ACOUSTICS ICA007MADRID