Development and optimization of a hybrid passive/active liner for flow duct applications 1
INTRODUCTION Design of an acoustic liner effective throughout the entire frequency range inherent in aeronautic applications, that is the fan noise propagating in the engine inlet: BPF and first harmonics or Buzz Saw Noise Acoustic liner State of art Passive treatments: e.g. SDOF, 2DOF Middle and high frequencies Narrowband attenuation (resonance) Purely active absorbers: p = 0 Low frequency components attenuation 2
THE HYBRID CELL CONCEPT Hybrid active/passive technology combining passive properties of absorbent materials and active control To realize a control of the wall impedance in such a way to ensure an optimal noise reduction throughout a large frequency domain Low frequencies Active field High frequencies Passive field Broadband equivalent of a λ/4 resonant absorber p = 0 Resistive layer Actuator Rigid wall v = 0 d Behaves as a classical passive absorber, mainly depending on d Advantages Versus passive solutions: increase frequency bandwidth to low frequencies Versus purely active solutions: active system separated from the flow Extension of the liner 3
THE MATISSE EXPERIMENTAL TEST BENCH Simple geometry: hybrid cell optimization process applied to Matisse set-up Acoustic primary source Instrumentation ducts Anechoic outlet Silent flow generation system Test region absorbent treatment 1 2 3 Pressure measurements 3.2 m 4 0.066 m Plane wave analysis domain: 700-2500 Hz Flow velocities up to 50 m/s Performance indexes: Transmission Loss (TL) 4
DESIGN AND OPTIMIZATION PROCESS Sellen et al., 9th AIAA/CEAS Aeroacoustic conference, Hilton Head, 2003, AIAA- 2003-3186 1. Determination of the optimal impedance Hilbrunner et al., 9th AIAA/CEAS Aeroacoustic conference, Hilton Head, 2003, AIAA-2003-3187 2. PASSIVE PART OPTIMIZATION 3. ACTIVE PART OPTIMIZATION Selection of the most suited porous layer according to the compromise: Significant noise reduction Achievable hybrid absorber THEORETICAL STUDY Geometry of the cell Actuator characteristics and control microphone selection Controller design Surface impedance measurements of different porous configurations NORMAL INCIDENCE MEASUREMENTS Pressure cancellation 4. Experimental validation on the Matisse facility (flow duct under grazing acoustic incidence): performance assessment Mazeaud et al., 10th AIAA/CEAS Aeroacoustic conference, Manchester, 2004, AIAA-2004-2852 5
1. DETERMINATION OF THE OPTIMAL IMPEDANCE Results Optimal impedance for different flow velocities Frequency dependence of real and imaginary parts Negative decreasing reactance Flow dependence: optimal resistance and reactance decreases as flow velocity increases 6
1. DETERMINATION OF THE OPTIMAL IMPEDANCE Results Sensibility study Objective: to define a tolerance range for the future realization step of targeted resistance and reactance Insertion loss parameter 800 Hz 2500 Hz 4000 Hz Frequency dependence of optimal impedance Optimal attenuation zone narrow, large noise reduction loss outside optimal region (especially at low frequencies) Sometimes two optimal areas appear: with or without flow 7
2. PASSIVE PART OPTIMIZATION Existing materials Different materials were tested (wire meshes, rockwool, ) The best compromise Wire mesh WM2 : σ e = 0.3 Z 0 20 mm air cavity Pressure cancellation Remarks Resistance: quite good, slightly low when frequency increases Resistance: quite good, slightly low when frequency increases Reactance: strongly negative Reactance: almost zero 8
2. PASSIVE PART OPTIMIZATION Hybrid functioning Active mode Pressure cancellation WM2 : 0.3 Z 0 PASSIVE Passive mode 10 mm air cavity ACTIVE 15 mm air cavity 20 mm air cavity Optimization of the hybrid functioning Determination of a commutation frequency (1800 Hz) between active and passive modes Depending on the authorized size of the complete system, from specifications 9
2. PASSIVE PART OPTIMIZATION Increasing attenuation levels Insertion loss simulation: MATISSE duct wall Treatment length Number of walls covered Mixed resistive layer 10 db 10 db 10 db WM1 WM2-WM1 WM2 Remark Increasing treatment length increase attenuation especially in low frequency range Two symmetrical walls covered increase attenuation over almost the whole frequency bandwidth WM2 WM1 WM2 WM1 10
3. ACTIVE PART OPTIMIZATION Back cavity Wire mesh Error sensor PZT actuator Cell size Collaboration with Metravib 55 mm Homogeneity of the pressure Bets position for the sensor : at the center 11
3. ACTIVE PART OPTIMIZATION Selection of the most suited type of controller Turbojet inlets covering extension of the liner surface MIMO system Analog filters : non adaptive Digital filters Feedforward structures Upstream reference insufficiently correlated with the sound to cancel Excessive memory and calculations requirements for real-time applications with huge number of cells Adaptive feedback cell by cell (IMC-MDFXLMS algorithm) 12
3.I SINGLE-CHANNEL STRUCTURE IMC-FXLMS block diagram Adaptive digital controller with the Filtered-x LMS algorithm applied to the IMC architecture (Internal Model Control, Elliott 95) Remarks Perfect secondary-path model the feedback contribution is removed the system acts as a feedforward controller Performance necessarily connected to the predictability of the perturbation d(n) 13
3.I. SINGLE-CHANNEL STRUCTURE Simulation results Performance of a two-tone in noise control 0.8 and 1.8 khz tones (Sampling frequency 10 khz) S/N = 15 db 20 taps Control ON at 0.2 s Magnitude (db) Frequency (Hz) Time (s) Strong attenuations Fast convergence (< tenth a second) Permanent stability 14
3.2. MULTI-CHANNEL STRUCTURE Object of the IMC: Estimation of the primary noise at the error sensor For MIMO systems, all the secondary contributions have to be taken into account Memory costs and computation loads become limiting factors for realtime applications Objective: development of a multi-channel algorithm based on a parallel functioning cell by cell Idea: Only the self and main feedback produced by the cell is reduced Cross-contributions then seen as part of the signal to minimize 15
3.2 MULTI-CHANNEL STRUCTURE Advantage Drawback Cells independent from the algorithm point of view Acoustic coupling remains due to a biased estimation of the primary noise Stability problem Simulation results for 4 hybrid cells 1.2 khz tone 2 taps, control ON at 0.2 s Instabilities Magnitude (db) Time (s) Frequency (Hz) Stability assured by means of a parallel bandpass filtering around each fixed and known tone of the primary noise 16
3.2. OPTIMIZATION OF THE MULTI-CHANNEL STRUCTURE Multi-tone ANC based on self adaptive band pass filtering of the reference Simulation results for 4 hybrid cells Bidirectional swept sine 1.5 khz & 2 khz tones SNR = 10 db Hybrid behavior: tone over 1.8 khz are not concerned by the ANC Control of evolving signals: fast convergence with few taps PASSIVE mode Without control 20 taps Control ON at 0 s With control ACTIVE mode Magnitude (db) Time (s) Frequency (Hz) Magnitude (db) Time (s) 17 17 Frequency (Hz)
4. EXPERIMENTAL VALIDATION Algorithm implementation system: Simulink Compilation to the floating-point DSP: Matlab/Real-Time Workshop Monitoring & acquisition systems: dspace ControlDesk & I-deas Experimental results for 4 hybrid cells Hybrid behaviour ANC of evolving signals 1 khz & 1.5 khz tones 20 m.s -1 flow 8 taps, control ON at 4 s Low number of control filters coefficients 10 Hz.s -1 unidirectional linear sweep 40 m.s -1 flow 8 taps, control ON at 20 s Magnitude (db) Frequency (Hz) Time (s) Magnitude (db) Frequency (Hz) Time (s) 18 18
4. EXPERIMENTAL VALIDATION Active cell functioning WM 2 Flow velocity dependence 2 active cells / 4 active cells mean flow 20 m/s 10 db Remarks Flow dependence essentially at low frequencies Attenuation decreases as flow velocity increases High attenuation with 2 active cells Importance of the treatment length 19
4. EXPERIMENTAL VALIDATION Hybrid cell functioning Comparison between predictions and measurements for v = 50 m.s -1 ACTIVE PASSIVE TL (db) experimental behaviour as predicted commutation frequency : 1800 Hz high attenuation Frequency (Hz) Passive prediction + Passive measurements Active prediction * Active measurements 20
CONCLUDING REMARKS Conclusion Development of a self-contained hybrid cells thanks to The IMC-MDFXLMS algorithm An adaptive bandpass filtering based on a multi-tone detection system Experimental validation of the theoretical predicted results Fast convergence and excellent stability Low number of taps for the control filters Up to 20 db at low frequencies and 15 db at higher frequencies Current investigation Broaden the frequency range of control to narrowband noise Test the hybrid liners on a more realistic test bench More hybrid cells (~ 50 cells) Higher flow velocities (~ M=0.3) 21