Active Noise Control in an Aircraft Cabin

Transcription

1 Active Noise Control in an Aircraft Cabin ipl.-ing. Christian Gerner University of the Federal Armed Forces Hamburg, Mechatronics Holstenhofweg Hamburg, Germany Phone: (+49) (40) Fax: (+49) (40) Prof. r.-ing. elf Sachau University of the Federal Armed Forces Hamburg, Mechatronics Holstenhofweg Hamburg, Germany Phone: (+49) (40) Fax: (+49) (40) ipl.-ing. Harald Breitbach Airbus Germany GmbH, epartment of Acoustics Kreetslag Hamburg, Germany Phone: (+49) (40) Fax: (+49) (40) ABSTRACT In propeller driven aircraft the main source for internal noise are almost tonal disturbances caused by the propeller blades that are passing the fuselage. In a certain four propeller military transport aircraft the maximum sound level in the cabin can reach up to 110 db(a), not taking into account any noise control treatments. Inside the semi closed loadmaster working station (LMWS) the sound level must be reduced down to 86 db(a). It is proposed to reach this goal with an active noise control system. To get the minimum number and the optimum positions for the secondary sources and the error sensors, the method of stepwise reduction is applied on different finite element models and a full-scale test bed in order to validate the. The comparison of the numerical and shows a good conformity at low frequencies. Furthermore, the test bed shows a good system-performance. Therefore, it is possible to get an optimized set of actuator and sensor position with the method of stepwise reduction.

2 INTROUCTION Nowadays, noise is classified as a special form of pollution. Therefore different industrial safety rules have been introduced to reduce noise levels. Furthermore, it is an important selling factor to have a quiet aircraft interior. In propeller driven aircraft the main sources for internal noise are narrow banded almost tonal disturbances caused by the propeller blades that are passing the fuselage in a certain blade pass frequency. One large study on aircraft interior active noise control (ANC) has been the ASANCA project, where a ornier 228 has been equipped with an ANC-system[1], [2]. In a certain four propeller military transport aircraft the maximum sound level in the cabin can reach up to 110 db(a) without taking into account any noise control treatments. Inside the semi closed loadmaster working station (LMWS) the sound level must be reduced down to 86dB(A). Passive treatments are heavy at low frequencies and therefore an active system is proposed for reaching this aim. To get the minimum number and the optimum positions for the secondary sources and the error sensors, evaluations on different FEM-models are performed in the frequency domain using FEMLAB [3]. The results of this evaluation are used to build a full scale test bed in order to validate the and to improve the system performance [4]. NUMERICAL EVALUATION If you have to select K sensors, or secondary actuators, from L possible locations by solving this as a combinatorial problem you need C calculations [5] with: C L! =. (1) K!( L K )! Especially for a large number of possible locations it is very time consuming to solve such a problem because of computational borders. Therefore a new method of selecting actuator and sensor locations is established. In a first step the number of actuators is reduced to N by deselecting L-N actuators with lowest amplitude at the controlled case. After this first step the number of actuators (and sensors) is reduced by selecting the best solution out of the N calculated solutions. This procedure continues with selecting the best solution out of the recalculated remaining N-1 solutions until the desired number of K actuators is reached. It is possible to use this method because of the linear behaviour of the system. This reduces the number of necessary calculations for the reduction from L to K actuators to N = 1+ Z. (2) Z= K+ 1 This method is used for numerical evaluations on a rectangular area - discretized by finite elements - in the frequency domain. Because of the promising results, it is applied on a model that is more suitable to the real situation at the aircraft (Figure 1) with a rectangular wooden box as LMWS (Figure 2) standing in a rectangular lab room as Fuselage. With this model the reduction of actuators was executed to get 8 which is the number of /A channels of the controller- out of 108 locations with a constant number of sensors.

3 Figure 1 FEMLAB Substitute Model of the LMWS and the Fuselage EXPERIMENTAL EVALUATION Test Bed A full-scale test bed (Figure 2) is built in accordance with the FEM-model used for the numerical evaluations. It is equipped with 8 secondary actuators and 11 error sensors. The actuators 200mm HiFi-speakers each powered by one 100W HiFi-amplifier - are positioned like evaluated with the method described above. The error sensors - B&K 4188 microphones - are positioned around the head of the loadmaster sitting inside the LMWS. The primary actuators are two 380mm HiFi-speakers, standing in the middle of the lab room, powered by a 1200W HiFi-amplifier. The system is controlled by a FXLMS-Algorithm with off-line plant modelling [6] running on a dspace-hardware [7].

4 Microphone Secondary Actuator Test Head Figure 2 Test Bed Experimental Results All experiments are conducted at an excitation frequency f that is similar to the first blade pass frequency (BPF). At this frequency the number of actuators is reduced to 4 and the number of sensors to 6 by stepwise selecting the best solution, not from the numerical but from the experimental data, as described above. Higher harmonics of f are neglected by this control algorithm and at the evaluation of the data. The results of the executed experiments with 4 secondary actuators and 6 error sensors are shown at Figure 3, where one can see the significant reduction of the SPL at the loadmasters ears.

5 sound pressure [db] cabin uncontrolled loadmasters left ear uncontrolled loadmasters right ear uncontrolled loadmasters left ear controlled loadmasters right ear controlled Figure 3 Experimental Results COMPARISON OF NUMERICAL AN EXPERIMENTAL ATA The numerical evaluation of the whole LMWS and Fuselage model (Figure 1) is very time consuming. Therefore, numerical and experimental evaluations for the validation of the numerical data are executed with the LMWS closed by a rigid wooden wall (Figure 4),[8]. Figure 4 FEMLAB Model of the LMWS closed by a rigid wooden wall

6 For the stepwise reduction f and the higher harmonics 2 f and 3 f are controlled separately. To get the set of actuator position for the next reduction step the reduction red for all frequencies is calculated from the reduction at the right and left ear red, red by with max( red ) as criteria for the selection of the best set. red = min( red, red ), (3) f The comparison of the experimental data with the numerical data (Figure 5-Figure 7) for the uncontrolled case shows good conformity for f, but larger differences at 2 f and 3 f. For the controlled case (Figure 8-Figure 10) there are differences at f, too. The Reasons for this differences are still under investigation. 120 r l r l sound pressure [db] mic1 mic2 mic3 mic4 mic5 mic6 mic7 mic8 mic9 Figure 5 Experimental and numerical data for f, uncontrolled case 120 sound pressure [db] mic1 mic2 mic3 mic4 mic5 mic6 mic7 mic8 mic9 Figure 6 Experimental and numerical data for 2 f, uncontrolled case

7 sound pressure [db] mic1 mic2 mic3 mic4 mic5 mic6 mic7 mic8 mic9 Figure 7 Experimental and numerical data for 3 f, uncontrolled case sound pressure [db] 90,0 80,0 70,0 60,0 50,0 40,0 30,0 20,0 10,0 0, removed loudspeaker Figure 8 Experimental and numerical data for f, controlled case sound pressure [db] 90,0 80,0 70,0 60,0 50,0 40,0 30,0 20,0 10,0 0, removed loudspeaker Figure 9 Experimental and numerical data for 2 f, controlled case

8 sound pressure [db] 90,0 80,0 70,0 60,0 50,0 40,0 30,0 20,0 10,0 0, removed loudspeaker Figure 10 Experimental and numerical data for 3 f, controlled case Figure 11 shows the significant reduction of the SPL for the controlled case at the numerical data. f based on Figure 11 Numerical results for the controlled and uncontrolled case, SPL [Pa] CONCLUSIONS This paper presented a new algorithm for optimal positioning of actuators and sensors in active noise control applications with a drastically reduced number of iteration steps. The new algorithm has been applied to a test structure, which has been modeled through finite elements to calculated the optimal sensor and actuator locations. Experimental tests have been conducted to verify the calculated positions. The comparison between the numerical and showed inconsistencies, especially at the higher harmonic frequencies. Therefore, it is not recommended to use the proposed algorithm with calculations alone. Because of the high number of possible solutions

9 it is impossible to find the optimal positions only through experiments. For this reason, it is practicable to find estimated position through FEM calculations and do the fine tuning based on the experimental data. Further work needs to be done by applying the stepwise reduction on the structure of a real airplane, including the LMWS. Furthermore, other control algorithms need to be examined for simultaneous reduction of the higher harmonic frequencies. This could reduce the computational power and improve the noise reduction even further. Additionally, other optimization techniques need to be studied (e.g. general reduced gradient algorithm, topology optimization algorithms) and the results need to be compared to the results presented in this paper. Finally, other actuators (e.g. flat panel speakers) should be looked at, to improve the performance of the active noise control. REFERENCES [1] Borchers, I.; et al. Advanced Study for Active Noise Control in Aircraft (ASANCA). In Adaptronic Congress, 1,p Berlin, [2] Borchers, I.; et al. Selected Flight Test ata and Control System Results of the CEC BRITE/EURAM ASANCA Study. In International Congress on Noiese Control Engineering,p Leuven, BE, [3] COMSOL_AB: FEMLAB - User's Guide and Introduction. 2. ed, [4] Gerner, C.; Sachau,.; Breitbach H. Active Noise Control in a Semi- Enclosure within an Aircraft Cabin. In Tenth International Congress on Sound and Vibration, Vol. 7,p Stockholm, [5] Elliott, S.J.: Signal Processing for Active Control. 1. ed. San iego, Calif.: Academic Press, [6] Kuo, S.M.; Morgan,.R.: Active noise control systems : algorithms and SP implementations. New York, NY: Wiley, [7] dspace_gmbh: Controlesk - Experiment Guide For Release 3.4, [8] Santag, S.; Ziese T., Validierung von FE-Berechnungen des LMWS- Testaufbaus durch experimentelle Untersuchungen.Universität der Bundeswehr Hamburg, Fachbereich Maschinenbau, Professur für Mechatronik, Studienarbeit. Hamburg, 2003.