EP0852793B1 - Hybrid active vibration control method and device, particularly for mechanical and acoustic vibration and the like - Google Patents

Abstract

A filtering device consisting of a nonadaptive feedback filter that generates active attenuation of the vibration on the framework, without generating instability in a first frequency band, and an adaptive feedforward filter is disclosed. Feedforward filtering coefficients are adapted in real time according to an algorithm chosen to minimize the energy of the types of vibration which are picked up by a sensing device as a function of the energy of the types of vibration which are picked up by the sensing device, and of the previously measured transfer function, in the presence of a feedback filtering device and in the absence of a feedforward filtering device, between an actuator device and a first sensing device. The device makes it possible to linearize the feedback attenuation throughout a second frequency band which is wider than the first frequency band, to accelerate the convergence of the minimization algorithm and to enhance the robustness of the feedforward filter.

Description

The present invention relates to active attenuation of vibrations, i.e. the operation which consists in attenuating certain vibrations, by superimposing other vibrations created in phase opposition with the vibrations to be attenuated.

It applies to the active attenuation of all vibrations and more particularly to sound, mechanical vibrations or the like.

We already know the technique of active attenuation by feedback, also called "FEED BACK", which is based on a feedback loop arranged to generate a active attenuation on a frame subject to vibrations (FR 86 03394).

The feedback loop includes an input connected to so-called “close” vibration sensor means, arranged on the framework, and an output connected to actuator means of vibrations placed on the frame, near the means close sensors. The signal measured by the sensor means close is directly injected into the actuator means at through filtering means which correct said signal to try to cancel his energy.

This retroactive technique makes it possible to obtain an attenuation vibration with a certain gain, without causing instability in a processing frequency band. Most of the time, this processing band corresponds to low frequencies, for example in sound vibrations at the frequency band ranging from 0 to 600 Hz.

However, this retroactive technique generates instabilities at high frequencies. It is therefore not totally satisfactory for obtaining vibration attenuation in a wide frequency band.

We also know the technique of active attenuation by anticipation, also called "FEED FORWARD", in which reference vibrations, upstream of the propagation of vibrations, and intended to propagate in the medium to process are detected by so-called "remote" sensor means, then processed by filtering means in order to determine the command to be applied to the actuator means.

The anticipation technique is articulated around means adaptive filtering whose coefficients are adapted in real time according to an algorithm chosen so as to minimize the energy of the vibrations captured by the means close sensors according to the energy of the vibrations of reference received by the remote sensor means.

Such a technique in advance is satisfactory usually to deal with vibrations in a band of narrow frequencies. However, when it comes to mitigating vibrations in a wide frequency band, it generally requires long and costly adaptive filtering.

A solution to shorten the convergence time of the anticipation filtering algorithm is described in the document MIYASAKI et al, 1994, "Consideration about Feed back, Feed forward, Hybrid Control for Active Control of Micro-Vibration ", Second International Conference on Motion on Vibration Control, Yokohama, Aug 30-Sept. 3, 1994. She consists in juxtaposing an active depreciation of retroactive type and advance filtering. Active depreciation by feedback first dampens vibrations at a given frequency, usually the frequency fundamental vibration of the frame. Then the filtering in anticipation performs its attenuation on vibrations thus pre-amortized, which allows the response to be compressed impulse of the secondary paths means actuators / means close sensors and thus shortens the processing of the advance filtering.

However, active feedback depreciation is here performed on a single frequency, which makes this solution inappropriate and ineffective for active treatment on a wide frequency band. This active depreciation is in equivalent to passive amortization since it does not than the fundamental frequency of the framework, which is totally different from active broadband control by retroactive filtering. In addition, this is a simple juxtaposition of early screening and amortization active in which no synergy of techniques is implementation.

We also know active noise attenuation structures (WO95 / 20841) in which the filtering means by anticipation and the feedback filtering means are simply juxtaposed in series to reduce noise. Of plus, the means of advance filtering and the means filtering are both of type adaptive, which prevents real synergy between said means of filtering by anticipation and by feedback.

The present invention provides a solution to these problems.

First, it aims to provide active mitigation of vibrations in a wide frequency band.

Another object of the invention is to provide attenuation active type "hybrid" in which filtering by anticipation is grafted onto feedback filtering or conversely, in order to improve the respective behavior said filtering in advance and by feedback with a resulting attenuation greater than the algebraic sum attenuations of said filterings taken separately.

• une ossature susceptible d'être sujette à des vibrations à atténuer;
• des premiers moyens capteurs de vibrations, disposés sur l'ossature selon une première relation géométrique prédéterminée par rapport à ladite ossature;
• des moyens actionneurs de vibrations, disposés sur l'ossature à proximité des premiers moyens capteurs; et
• des moyens de filtrage comprenant au moins une entrée reliée aux premiers moyens capteurs et une sortie reliée aux moyens actionneurs, les moyens de filtrage étant agencés pour engendrer une atténuation active des vibrations sur l'ossature,
• des seconds moyens capteurs de vibrations, disposés sur l'ossature selon une seconde relation géométrique prédéterminée par rapport à ladite ossature;
• des moyens sommateurs possédant une première entrée, une seconde entrée, et une sortie reliée aux moyens actionneurs.

The present invention relates to an active vibration attenuation device, of the type comprising:

• a frame likely to be subject to vibrations to be attenuated;
• first vibration sensor means, disposed on the frame in a first predetermined geometric relationship with respect to said frame;
• vibration actuator means, disposed on the frame near the first sensor means; and
• filtering means comprising at least one input connected to the first sensor means and one output connected to the actuator means, the filtering means being arranged to generate active attenuation of vibrations on the frame,
• second vibration sensing means, disposed on the frame in a second predetermined geometric relationship with respect to said frame;
• summing means having a first input, a second input, and an output connected to the actuator means.

• des moyens de filtrage par rétroaction de type non adaptatif possédant une entrée reliée aux premiers moyens capteurs et une sortie reliée à la première entrée des moyens sommateurs, et propres à engendrer une atténuation active non adaptative des vibrations sur l'ossature, sans engendrer d'instabilité dans une première bande de fréquences;
• des moyens de mesure propres à mesurer, au préalable, en présence des moyens de filtrage par rétroaction, la fonction de transfert entre les moyens actionneurs et les premiers moyens capteurs;
• des moyens de filtrage par anticipation de type adaptatif comprenant une première entrée reliée aux seconds moyens capteurs, une seconde entrée reliée aux premiers moyens capteurs, et une sortie reliée à la seconde entrée des moyens sommateurs;

les coefficients de filtrage des moyens de filtrage par anticipation étant adaptés en temps réel selon un algorithme choisi pour minimiser l'énergie des vibrations captées par les premiers moyens capteurs en fonction de l'énergie des vibrations captées par les seconds moyens capteurs, et de la fonction de transfert préalablement mesurée;
ce qui permet de linéariser l'atténuation rétroactive dans toute une seconde bande de fréquences plus large que la première bande de fréquences, d'accélérer la convergence de l'algorithme de minimisation, et d'améliorer la robustesse des moyens de filtrage par anticipation.According to a general definition of the invention, the filtering means comprise:

• non-adaptive type feedback filtering means having an input connected to the first sensor means and an output connected to the first input of the summing means, and capable of generating an active non-adaptive attenuation of the vibrations on the framework, without generating instability in a first frequency band;
• measurement means suitable for measuring, beforehand, in the presence of the feedback filtering means, the transfer function between the actuator means and the first sensor means;
• adaptive type anticipation filtering means comprising a first input connected to the second sensor means, a second input connected to the first sensor means, and an output connected to the second input of the summing means;

the filter coefficients of the anticipation filter means being adapted in real time according to an algorithm chosen to minimize the energy of the vibrations picked up by the first sensor means as a function of the energy of the vibrations picked up by the second sensor means, and the transfer function previously measured;
which makes it possible to linearize the retroactive attenuation in a whole second frequency band wider than the first frequency band, to accelerate the convergence of the minimization algorithm, and to improve the robustness of the anticipation filtering means.

According to a first embodiment of the invention, the framework includes at least one cavity defined by one ear and passive mitigation means, the first means sensors and actuator means being housed in said cavity while the second sensor means being arranged out of the cavity.

In another embodiment of the invention, the framework includes a metal beam, a plate, a trellis, seat, ventilation duct or the like.

In practice, the first and second sensor means comprise each at least: a type of sound sensor element microphone, a type acceleration sensor element accelerometer, a displacement sensor element, an element speed sensor, a strain sensor element, a force sensor element or the like.

In a variant, the first sensor means comprise two sensor elements, one being associated with the means of filtering in advance, the other being associated with the means filtering by feedback.

Preferably, the actuator means comprise a source speaker-type sounds, a test body, a jar vibrating or the like.

Advantageously, the filtering means by feedback include a plurality of analog filters and / or active numerics of order greater than or equal to 1, arranged for generate a transfer function to avoid instabilities in the first frequency band in the sense of Nyquist, and the transfer function of the filtering means by feedback is determined so that the phase of said transfer function does not go through the value 0 in the first frequency band.

In practice, the feedback filtering means are infinite impulse response type.

For example, the means of advance filtering are finite impulse response, and the minimization algorithm is of the average least squares type.

According to another aspect of the invention, the device comprises a plurality of first sensor means, and means actuators and the device is articulated around a master / slave multiprocessor structure, each processor slave being responsible for controlling a single actuator means.

The present invention also relates to a method active hybrid attenuation of vibrations, in particular mechanical, sound or similar vibrations used by the device described above.

• la figure 1 est une représentation schématique du dispositif d'atténuation acoustique active selon l'invention;
• la figure 2 représente schématiquement les moyens essentiels et constitutifs du dispositif de la figure 1 selon l'invention;
• la figure 3 sont des courbes illustrant l'atténuation des vibrations sonores en présence/absence de moyens de filtrage par rétroaction, dans une bande de fréquences allant de 0 à 2500 Hz;
• les figures 4 et 5 sont des courbes illustrant l'atténuation des vibrations sonores en présence/absence de moyens de filtrage hybride selon l'invention, dans une bande de fréquences allant de 0 à 2500 Hz;
• les figures 6 et 7 sont des courbes illustrant l'atténuation des vibrations sonores en présence/absence des moyens de filtrage hybride selon l'invention, dans une bande de fréquences allant de 500 Hz à 1500 Hz;
• la figure 8 est une courbe illustrant l'atténuation des vibrations sonores en présence/absence des moyens de filtrage hybride selon l'invention, dans une bande de fréquences allant de 0 à 500 Hz;
• les figures 9, 10A et 10B sont des courbes illustrant l'atténuation des vibrations sonores en présence/absence des filtrage par anticipation;
• la figure 11 représente schématiquement la structure du dispositif d'atténuation multi-voies selon l'invention, dans lequel les filtrages par anticipation et par rétroaction sont numériques;
• la figure 12 représente schématiquement les éléments constitutifs du processeur esclave du dispositif de la figure 11;
• la figure 13 représente schématiquement la structure du dispositif d'atténuation multi-voies selon l'invention, dans lequel le filtrage par rétroaction est analogique; et
• la figure 14 représente schématiquement l'assemblage du filtrage par rétroaction analogique dans le dispositif de la figure 13.

Other characteristics and advantages of the invention will become apparent in the light of the detailed description below and the appended drawings in which:

• Figure 1 is a schematic representation of the active acoustic attenuation device according to the invention;
• FIG. 2 schematically represents the essential and constitutive means of the device of FIG. 1 according to the invention;
• FIG. 3 are curves illustrating the attenuation of sound vibrations in the presence / absence of filtering means by feedback, in a frequency band going from 0 to 2500 Hz;
• Figures 4 and 5 are curves illustrating the attenuation of sound vibrations in the presence / absence of hybrid filtering means according to the invention, in a frequency band ranging from 0 to 2500 Hz;
• Figures 6 and 7 are curves illustrating the attenuation of sound vibrations in the presence / absence of the hybrid filtering means according to the invention, in a frequency band ranging from 500 Hz to 1500 Hz;
• FIG. 8 is a curve illustrating the attenuation of sound vibrations in the presence / absence of the hybrid filtering means according to the invention, in a frequency band going from 0 to 500 Hz;
• Figures 9, 10A and 10B are curves illustrating the attenuation of sound vibrations in the presence / absence of advance filtering;
• FIG. 11 schematically represents the structure of the multi-channel attenuation device according to the invention, in which the filtering by anticipation and by feedback is digital;
• FIG. 12 schematically represents the constituent elements of the slave processor of the device of FIG. 11;
• FIG. 13 schematically represents the structure of the multi-channel attenuation device according to the invention, in which the filtering by feedback is analog; and
• FIG. 14 schematically represents the assembly of the filtering by analog feedback in the device of FIG. 13.

In Figure 1, there is shown a particularly application advantageous, nonlimiting, of the invention, namely attenuation of sound vibrations. In this app, the frame likely to be subject to vibrations at attenuate includes a cavity 2 delimited by an ear 4 and attenuation means 6 of the helmet type 6.

For example, the helmet with retroactive filter is the one sold by the company TECHNOFIRST.

In known manner, this helmet is equipped with a mitigation device active acoustic feedback.

• un microphone 8 disposé dans la cavité 2;
• un haut-parleur 10 disposé dans la cavité 2 à proximité du microphone 8;
• des moyens de pré-amplification 9 comprenant une entrée 7 reliée au microphone 8 et une sortie 11,
• des moyens de filtrage par rétroaction 12 comprenant une entrée 14 reliée à la sortie 11, et une sortie 16; et
• des moyens d'amplification 18 comprenant une entrée 20 reliée à la sortie 16, et une sortie 22 reliée à l'entrée du haut-parleur 10.

In practice, this feedback attenuation device comprises for each ear:

• a microphone 8 disposed in the cavity 2;
• a speaker 10 disposed in the cavity 2 near the microphone 8;
• pre-amplification means 9 comprising an input 7 connected to the microphone 8 and an output 11,
• feedback filtering means 12 comprising an input 14 connected to the output 11, and an output 16; and
• amplification means 18 comprising an input 20 connected to the output 16, and an output 22 connected to the input of the loudspeaker 10.

The pre-amplification means 9, the filtering means by feedback 12 and the amplification means 18 constitute here a feedback loop 30 arranged in a known manner to generate active acoustic attenuation without generate instability in a selected frequency band.

In Figure 1, there is shown near the helmet, a noise source 40 liable to generate vibrations sound for experimental and test purposes.

For example, the frequency band in which the means filtering methods are effective without causing instability in the sense of Nyquist, is of the order of 0 to 600 Hz for sound vibrations (figure 3).

In practice, the filtering means by feedback 12 include a plurality of active analog filters of order greater than or equal to 1, arranged to generate a transfer function to avoid instabilities in the frequency band 0-600 Hz in the sense of Nyquist, and the transfer function of the filtering means 12 is determined so that the phase of said function of transfer does not go through the value 0 in the band 0-600 Hz.

Such feedback filtering is for example described in French Patent 86,03394.

Referring to Figure 3, the helmet allows treatment wide band up to 600 Hz and noise attenuations of around 20 dB. However, a pumping effect appears at from 650 Hz which results in an increase in noise level in relation to the action of the mitigation means passive alone. This phenomenon is well known from the skilled person, and constitutes a non-linearity (degradations performance) compared to the expected results of observing the open loop system.

The Applicant has posed the problem of remedying the disadvantages related to feedback filtering.

The solution according to the invention consists first of all, in use an additional microphone 100 placed at a certain distance from the microphone 8. For example, the microphone additional 100 is arranged on the upper part of the means which allow the two shells of the helmet. Under these conditions, the additional microphone 100 is close to the noise source 40 and thus makes it possible to retrieve useful information to be processed. Of course, this remote microphone can be arranged differently.

Next, according to the invention, summing means 110 are provided at the level of the feedback loop 30. These summing means 110 have a first input 112 connected to the output 16 of the filtering means 12, a second input 114 and an output 116 connected to input 20 of the means amplifiers 18.

Finally, according to the invention, type filtering means by anticipation are grafted to the feedback loop 30 to improve feedback filtering and, more exactly, in order to linearize the active attenuation in the entire frequency band wider than the 0-600 band Hz and thus improve the active attenuation gain in the widened band which can go up to 3000 Hz (figure 4), by total elimination of the mentioned pumping effect above.

In practice, the means of advance filtering 130 include a first input 132 connected to the microphone additional 100, a second input 134 connected to the microphone 8 and an output 136 connected to the second input 114 of the summing means 110.

As will be seen in more detail below, the coefficients filtering the anticipation filtering means 130 are adapted in real time according to an algorithm chosen for minimize the energy of vibrations picked up by the microphone 8, depending on the energy of the vibrations picked up by the microphone 100, in order to linearize the retroactive attenuation in a whole frequency band wider than the band of frequencies processed directly by feedback, speed up convergence of the minimization algorithm, and improve the robustness of the anticipation filtering means.

In practice, the means of advance filtering include type finite impulse response filters adaptive 140. The coefficients of the filters 140 are updated in real time by a minimization algorithm 150. For example, the minimization algorithm is of the type of mean least squares, also called LMS for "LEAST MEAN SQUARES ".

This linearization can be observed at least at two points, especially in the 0-600 Hz band in which the attenuation gain in the band is improved; as well as band level 600-1100 Hz in which the reappearance vibrations related to the pumping of the retroactive filtering is deleted and in which an attenuation appears while this does not exist in the presence of feedback filtering alone (Figures 3 to 8).

This acceleration of the convergence of the algorithm minimization as well as improving the robustness of means of early filtering is observable in comparing the attenuation curves obtained by the means advance filtering alone (Figures 9, 10A and 10B) compared to the attenuation curves obtained by the means hybrid filtering according to the invention (Figures 4 to 8).

It should be noted that the active attenuation of the "hybrid" type obtained according to the invention results from a combination of means of filtering by anticipation and by feedback in which the advance filtering is grafted on the filtering by feedback or vice versa.

This combination of filtering in advance and by feedback according to the invention improves behavior respective of said filterings, with active attenuation resulting greater than the algebraic sum of the attenuations individual said filterings taken separately.

In Figure 2, there is shown in detail the elements constituting the filtering means by feedback as well as the constituent elements of the anticipation filtering means, the latter being combined with the means of filtering by feedback according to the invention.

The anticipation filtering means 130 comprise a first A8 acquisition module associated with sensor means relatives 8 and a second A100 acquisition module associated with remote sensor means 100.

The A8 and A100 acquisition modules are generally alike. However, in some configurations, the acquisition modules can be different. Their constituent elements are identified by the suffix 8 when associated with sensor means 8 and 100 when associated with the remote capture means 100.

• un élément pré-amplificateur d'entrée PE possédant une entrée reliée aux moyens capteurs 8 ou 100 associés et une sortie;
• un filtre de conditionnement FAT spécifique à l'application choisie, de préférence de type anti-recouvrement possédant une entrée reliée à la sortie du pré-amplificateur d'entrée et une sortie; et
• un convertisseur analogique/numérique CAN possédant une entrée reliée à la sortie du filtre de conditionnement et une sortie.

Each acquisition module includes:

• an input pre-amplifier element PE having an input connected to the associated sensor means 8 or 100 and an output;
• a FAT conditioning filter specific to the chosen application, preferably of the anti-overlap type having an input connected to the output of the input preamplifier and an output; and
• an analog / digital ADC converter having an input connected to the output of the conditioning filter and an output.

Each acquisition module is connected to processing means DSP which will ensure in particular the minimization algorithm described above.

In practice, the digital processing means are of the type PSN digital signal processor.

DSP processor includes an E8 input for receiving signals leaving the A8 acquisition module and an E100 input receiving signals from the A100 acquisition module.

DSP processor includes signal output digital for a R rendering module.

This R rendering module includes a digital / analog converter CNAR and a FLR smoothing filter, for example low pass type, the input of which receives the outgoing signal from the digital / analog converter CNAR and whose output is connected to the second input 114 of the summing means 110.

For example, the DSP processor is the one sold by the company TEXAS INSTRUMENT under the reference TMS 320C25.

The operation of the device according to the invention is the next.

In a first initialization step, the means of feedback filtering 12 are put into operation as well that the noise source 40, while the filtering means in advance are put in pause position.

The preamplifiers are regulated on the means of advance filtering PE8 and PE100, to be in full scale of CAN8 analog / digital converters and CAN100.

Then we stop the noise source 40. We then measure the transfer function of the so-called secondary path between the speaker 10 and the so-called control microphone 8 by a method initialization, for example by exciting the actuator means by Diracs type signals, white noise, filtered reference or the like.

Finally, the transfer function is sampled and saved in the memory of the DSP processor. For example, the transfer function is sampled at the frequency of 3000 Hz on a number of 80 points. Advantageously, we settle the gain of amplifier 18 so that the excitation of the speaker 10 produces at the output of the PE8 preamplifier, a signal level close to that set in step previous relating to the dynamic adjustment of the converters.

This transfer function thus previously measured will then use in the calibration phase for adaptation elements of advance filtering.

In operating condition, i.e. during the phase minimization of the hybrid control device (i.e. feedforward combination with feedback or vice versa), the digital processing means acquire periodically, and in real time, the distant noise picked up by remote sensor means 100. They also calculate signal energy, representative of the sum of energies signals delivered by nearby sensor means 8. Then, the anticipation filter means 150 are placed in search of optimal parameters for the best active attenuation. Knowledge of impulse responses previously measured, signals from the means close and remote sensors in real time, allows a minimization algorithm chosen to determine, in time real, the attenuation control signal values active acoustics.

The purpose of convergence here is to minimize the energy of signals delivered by the microphone 8 placed in the cavity to denoise from the helmet.

For example, the minimization algorithm uses the technique mean least squares which is most prevalent in real-time applications.

Alternatively, the minimization algorithm may be a frequency algorithm working on transforms of Fourier of the signals considered.

It should be noted that the impulse response or the speaker / microphone path transfer function from control 8 takes account of filtering by feedback here.

Thus, the instability information related to filtering by feedback are introduced in the impulse response of the filter in advance. Similarly, the information broadband active attenuation related to filtering by feedback appear in the sampled items of the impulse response.

• annulation des amplifications de signaux en haute fréquences liées aux instabilités par rétroaction;
• atténuation du bruit en dehors de la bande de traitement du filtrage par rétroaction (gain jusqu'à 10 dB par rapport au traitement passif des coquilles dans le cas du casque);
• amélioration du traitement dans la totalité de la bande de traitement de filtrage par rétroaction (gain jusqu'à 15 dB par rapport au filtrage par rétroaction seul), ce qui permet de le rendre encore plus linéaire;
• amélioration de la robustesse du système, par exemple aux effets Larsen;
• amélioration des performances par rapport à des filtrages par rétroaction et des filtrages par anticipation utilisés séparément.
• accélération de la convergence de l'algorithme de minimisation,
• amélioration de la robustesse des moyens de filtrage par anticipation (observable en comparant les courbes d'atténuation obtenues par les moyens de filtrage par anticipation seuls (figures 9, 10A et 10B) par rapport aux courbes d'atténuation obtenues par les moyens de filtrage hybride selon l'invention (figures 4 à 8).

Thanks to the participation of this information therefore linked to the instability of the feedback filtering as well as the information related to the active broadband attenuation of this filtering, the anticipation filtering will bring several types of improvements with reference to FIGS. 10B:

• cancellation of amplifications of high frequency signals linked to instabilities by feedback;
• noise attenuation outside the feedback filtering processing band (gain up to 10 dB compared to passive shell processing in the case of headphones);
• improved processing in the entire feedback filtering processing band (gain up to 15 dB compared to feedback filtering alone), which makes it even more linear;
• improved robustness of the system, for example to Larsen effects;
• improved performance compared to feedback filtering and anticipation filtering used separately.
• acceleration of the convergence of the minimization algorithm,
• improvement of the robustness of the anticipation filtering means (observable by comparing the attenuation curves obtained by the anticipation filtering means alone (FIGS. 9, 10A and 10B) compared to the attenuation curves obtained by the hybrid filtering means according to the invention (Figures 4 to 8).

Note that the action of advance filtering does not disturb that of filtering by feedback in the sense where we can stop the minimization during the advance filtering without affecting the performance of the filtering by feedback.

Figures 3 to 10B show the spectral densities of power measured using a fixed microphone in the ear of the experimenter in different configurations. Adverse reactions due to instability of the feedback filtering (rejection up to 8 dB) are eliminated by the action of the advance filtering device (see figures 3, 4 and 5).

Better still, the anticipation device makes it possible to obtain, outside the feed back processing band (o-600 Hz), a gain in attenuation of 2 to 10 dB compared to a passive helmet (figure 6).

Likewise, better anti-noise results have been measured in low frequencies up to +15 dB.

The device described with reference to Figures 1 and 2, uses a single-channel type processing, articulated around the processor TMS 320C25 from TEXAS INSTRUMENT which can execute 10 million instructions per second.

However, it may be too slow when it has to drive a multi-channel device comprising a plurality of sensors 8, remote sensors 100 and actuators 10.

For example, for operation of the device with five microphones 8, five speakers 10, 60 points for impulse responses and 15 filter coefficients by anticipation, the processor can only work at sampling frequencies less than or equal to 1000 Hz.

However, certain experiments require treatment fast, for a number of microphones 8 known as errors and sources of noise reduction 10 greater than five.

In addition, to keep the full efficiency of the algorithm minimization it may be necessary to have good knowledge of the path impulse response secondary. We must therefore save this memory answer with a large number of points. This number of points also determines the number of signal samples reference from remote sensor 100 which must be kept also in internal memory, hence the problem of capacity memory.

The present invention also provides a solution to these problems.

According to the invention, the attenuation device is capable of manage a plurality of channels, for example twenty channels analog input capable of receiving signals from 19 close sensors individualized in 8-1 to 8-19 and a remote sensor 100. The device according to the invention also includes at least sixteen capable output channels to convey signals to sixteen individualized actuators in 10-1 to 10-16.

Such a structure implies the processing of I (number error sensor integer 8) times J (integer number of actuators) impulse responses, one RIJ response for each combination of actuators J and error sensors I.

Very advantageously, the device is articulated around a master / slave multiprocessor structure, each processor slave being responsible for controlling a single actuator means.

Referring to Figure 11, the DSPM master processor does the acquisition of all the signals emanating from the sensors 8 and 100, in particular the so-called remote reference signals as well as the so-called error control signals. He distributes them then to all DSPE slave processors, individualized here in DSPE-1 to DSPE-16.

Each DSPE slave processor calculates the output signal of a single actuator 10.

In practice, the sensors 8 and the remote sensor 100 are connected to the inputs of a BA acquisition block whose output is connected to the DSPM master processor.

This BA acquisition block includes, like the acquisition modules A described with reference to FIG. 2, a preamplifier element PE, preferably a conditioning filter specific to the chosen application FAT and a converter analog / digital CAN.

The conditioning filter can be digital (anti-overlap) or analog (specific).

A laptop type microcomputer can be provided. he is in this case connected to the master processor and is provided with all control software for the entire installation.

The digital assembly is articulated around a processor element PSN digital signal, for example the one sold by the TEXAS INSTRUMENT under the reference TMS 320C50.

Referring to Figure 12, Each slave processor is dedicated to the control of a single actuator. For example, this is the processor associated with the 10-1 actuator and which is in connection with all microphones 8 as well as with the remote microphone 100.

All signals from sensors 8 and 100 are routed through the BA acquisition block and the DSPM master processor to the DSPE-1 slave processor.

The DSPE slave processor generally includes the same elements than those of the single-channel device described with reference in Figure 2. Thus, we find the means of restitution R, the feedback filtering means 12 as well as the advance filtering 130. A summing element 110 receives signals from both at its two inputs filtering to deliver the attenuation signal on its output to the actuator 10-1.

The slave processor includes communication with the DSPM master processor.

In the case of analog feedback filtering (Figure 14), changes are made to the restitution block. An individual summator 100-1, i.e. associated with the slave processor DSPE-1, allows to add the analog signal from 12-1 feedback filtering at analog signal from specific FLR-1 filtering.

• soit au nombre de capteurs 8, lorsque ce nombre est égal au nombre d'actionneurs 10,
• soit au nombre d'actionneurs 10, si ce nombre est inférieur au nombre de capteurs 8,
• soit au nombre de capteurs 8, si celui-ci est inférieur au nombre d'actionneurs 10.

It should be noted that the feedback filtering only makes sense for a pair of transducers comprising an actuator and a sound sensor. Under these conditions, the number of filterings by digital or analog feedback is equal to:

• either the number of sensors 8, when this number is equal to the number of actuators 10,
• either the number of actuators 10, if this number is less than the number of sensors 8,
• or the number of sensors 8, if this is less than the number of actuators 10.

This defines the maximum number of filterings by feedback.

We also define the notion of pairs of transducers 8 and 10, i.e. the treatment routes on which are applied the respective feedback filtering means.

Then, each DSPE slave processor calculates, in parallel advance filtering, feedback filtering which associated with it, in the case of filtering by feedback of digital type.

In the case of analog feedback filtering (Figure 12), there is a network of connections in which plug-in filtering modules by feedback between the couples of transducers 8,10 chosen.

In the description above, we have described an application related to sound vibrations. However, the invention applies to the active attenuation of any vibration.

Thus, the structure subject to vibrations can also be a metal beam, a plate, a trellis, a seat, ventilation duct or the like. In these conditions, the sensor means can be sensor means of sounds, but also of acceleration, constraint, force, displacement, speed or the like. Likewise, the means actuators can be not just a sound actuator such as a speaker, but also a test body, a piezoelectric element, or the like. In addition, close sensor means may include two elements sensors, one being associated with the filtering means by anticipation, the other being associated with the filtering means by feedback.

Claims (17)

1. Active vibration attenuation device of the type comprising:

   a framework likely to be subject to vibration to be attenuated;

   first vibration sensor means (8), arranged on the framework in a first predetermined geometric relationship with respect to said framework;

   vibration actuator means (10), arranged on the framework in proximity to the first sensor means (8) ; and

   filtering means comprising at least one input linked to the first sensor means and an output linked to the actuator means, the filtering means being configured to generate active attenuation of the vibration on the framework;

   second vibration sensor means (100), remote, arranged on the framework in a second predetermined geometric relationship with respect to said framework;

   summation means (110) possessing a first input (112), a second input (114), and an output linked to the actuator means (10);

   characterized in that the filtering means comprise:

   nonadaptive-type feedback filtering means (12) possessing an input linked to the first sensor means (8) and an output (16) linked to the first input (112) of the summation means, and suitable for generating nonadaptive active attenuation of the vibration on the framework, without generating instability in a first frequency band;

   means suitable for measuring, in advance, and in the presence of the feedback filtering means, the transfer function between the actuator means (10) and the first sensor means (8);

   adaptive-type feedforward filtering means (130) comprising a first input (132) linked to the second sensor means (100), a second input (134) linked to the first sensor means (8), and an output linked to the second input (114) of the summation means (110);

      the filtering coefficients (140) of the feedforward filtering means (130) being adapted in real time according to an algorithm chosen to minimize the energy of the vibration which is picked up by the first sensor means (8) as a function of the energy of the vibration which is picked up by the second sensor means (100), and of the thus previously measured transfer function, which makes it possible to linearize the feedback attenuation throughout a second frequency band which is wider than the first frequency band, to accelerate the convergence of the minimization algorithm, and to enhance the robustness of the feedforward filtering means.
2. Device according to claim 1, characterized in that the framework comprises at least one cavity (2) bounded by an ear (4) and passive attenuation means (6), the first sensor means (8) and the actuator means (10) being lodged in said cavity (2), while the second sensor means (100) being arranged outside the cavity.
3. Device according to claim 1, characterized in that the framework comprises a metal-type beam, or a plate, or a trellis, or a seat,or a ventilation duct or the like.
4. Device according to claim 1, characterized in that the first sensor means (8) comprise at least: a sound sensor element of microphone type, an acceleration sensor element of accelerometer type, a movement sensor element, a speed sensor element, a stress sensor element, a force sensor element or the like.
5. Device according to claim 4, characterized in that the first sensor means (8) comprise two sensor elements, one being associated with the feedforward filtering means, the other being associated with the feedback filtering means.
6. Device according to claim 1, characterized in that the second sensor means (100) comprise at least: a sound sensor element of microphone type, an acceleration sensor element of accelerometer type, a movement sensor element, a speed sensor element, a stress sensor element, a force sensor element or the like.
7. Device according to claim 1, characterized in that the actuator means (10) comprise a sound source of loudspeaker type, a test unit, a vibrating platform or the like.
8. Device according to one of the preceding claims, characterized in that the feedback filtering means (12) comprise a plurality of active digital and/or analog filters of order greater than or equal to 1, configured to generate a transfer function making it possible to avoid instability in the first frequency band in the Nyquist sense, and in that the transfer function of the feedback filtering means is determined in such a way that the phase of said transfer function does not pass through the value zero in the first band.
9. Device according to any one of the preceding claims, characterized in that the feedback filtering means (12) are of the infinite impulse-response type.
10. Device according to any one of the preceding claims, characterized in that the feedforward filtering means (130) comprise:

    a first acquisition module (A8) possessing an input linked to the output of the first sensor means (8), and an output;

    a second acquisition module (A100) possessing an input linked to the output of the second sensor means (100), and an output;

    digital processing means possessing a first input linked to the output of the first acquisition module (A8), a second input linked to the output of the second acquisition module (A100), and an output, said digital processing means being suitable for controlling the algorithm for minimizing the energy of the vibration which is picked up by the first sensor means (8) as a function of the energy of the vibration which is picked up by the second sensor means (100); and

    a reproduction module (R) possessing an input linked to the output of the digital processing means and an output linked to the second input (114) of the summation means.
11. Device according to claim 10, characterized in that the first and second acquisition modules (A8 and A100) comprise:

    an input preamplifier element (PE) possessing an input receiving the signals output by the first sensor means (8) or by the second sensor means (100), and an output;

    a conditioning filter (FAT) adapted to the chosen application and possessing an input linked to the output of the input preamplifier element, and an output; and

    an analog/digital converter (CAN) possessing an input linked to the output of the conditioning filter and an output linked to an input of the digital processing means.
12. Device according to claim 10, characterized in that the reproduction module (R) comprises:

    a digital/analog converter (CNA) possessing an input linked to the output of the digital processing means, and an output; and

    a smoothing filter (FLR) possessing an input linked to the output of the digital/analog converter and an output linked to the second input (114) of the summation means (110).
13. Device according to claim 10, characterized in that the feedforward filtering means (130) are finite impulse-response filters and in that the minimization algorithm is of the least mean squares type.
14. Device according to claim 10, characterized in that the digital processing means are of the digital signal processor type.
15. Device according to any one of the preceding claims, characterized in that it comprises a plurality of first sensor means (8-1, 8-2, 8-3, etc.), and of actuator means (10-1, 10-2, etc.) and in that the device is centered around a structure with master/slave multiprocessors, each slave processor (DSPE-1) being tasked with driving a single actuator means (10-1).
16. Method of active vibration attenuation, particularly of acoustic vibration, of the type comprising the following stages:

    a) providing a framework likely to be subject to vibration to be attenuated;

    b) providing first vibration sensor means (8), arranged an the framework in a first predetermined geometric relationship with respect to said framework;

    c) providing vibration actuator means (10), arranged on the framework in proximity to the first sensor means ;

    d) providing filtering means comprising at least one input linked to the first sensor means (8) and an output (16);

    e) configuring the filtering means so as to generate active attenuation of the vibration an the framework;

    f) providing second vibration sensor means (100), remote, arranged on the framework in a second predetermined geometric relationship;

    g) providing summation means (110) possessing a first input (112), a second input (114), and an output linked to the actuator means (10);

    characterized in that it further comprises the
    following stages:

    h) providing nonadaptive-type feedback filtering means (12) possessing an input linked to the first sensor means (8) and an output (16) linked to the first input (112) of the summation means, and configuring the said feedback filtering means so as to generate nonadaptive-type active attenuation of the vibration on the framework, without generating instability in a first frequency band;

    i) measuring, in advance, and in the presence of the feedback filtering means, the transfer function between the actuator means (10) and the first sensor means (8) ;

    j) providing nonadaptive-type feedforward filtering means (130) comprising a first input (132) linked to the second sensor means (100), a second input (134) linked to the first sensor means (8), and an output linked to the second input (114) of the summation means (110) ;

    k) adapting the filtering coefficients (140) of the feedforward filtering means (130) in real time according to an algorithm chosen to minimize the energy of the vibration which is picked up by the first sensor means (8) as a function of the energy of the vibration which is picked up by the second sensor means (100), and the previously measured transfer function;
    which makes it possible to linearize the feedback attenuation throughout a second frequency band which is wider than the first frequency band, to accelerate the convergence of the minimization algorithm, and to enhance the robustness of the feedforward filtering means.
17. Method according to claim 16, characterized in that it is implemented by a device according to any one of claims 1 to 15.

Images