EP0555266B1

EP0555266B1 - Active vibration control system

Info

Publication number: EP0555266B1
Application number: EP91918376A
Authority: EP
Inventors: Colin Fraser Ross; Graham Paul Eatwell
Original assignee: Noise Cancellation Technologies Inc
Current assignee: Noise Cancellation Technologies Inc
Priority date: 1990-10-29
Filing date: 1991-10-22
Publication date: 1998-09-02
Anticipated expiration: 2011-10-22
Also published as: EP0555266A1; ATE170654T1; DE69130112T2; WO1992008223A1; DK0555266T3; JPH06502257A; ES2121790T3; CA2095033A1; AU660423B2; DE69130112D1; AU8739391A; CA2095033C; GB9023459D0

Abstract

An active vibration control system in which an input sensor (2) generates first signals representative of a primary vibration field, actuators (1) are driven by second signals to produce a secondary vibration field, monitoring sensors (3) producing third signals are positioned in a first region (4) differing from a second region where vibration is to be controlled, a controller (5) is responsive to the first signals to generate second signals which reduce vibration in the second region, and the controller is adaptive with reference to the first and third signals to maintain reduced vibration in the second region.

Description

Field of Invention

The invention relates to active control systems for vibration which reduce vibration in a region where it is undesirable or impossible to mount residual sensors in order to monitor the reduced level of vibration.

Background to the invention

The term vibration, when used in this document, should be taken to include sound and other small amplitude linear disturbances.

The use of simple active techniques to control sound or vibration is well known, and the technology is progressing fairly rapidly. The first work that was reported, related to the control of sound in a duct (Leug US2043416). The system described used an 'upstream' microphone to act as an input sensor. This input sensor generated a signal related to the amplitude and phase of the sound wave progressing down the duct (the primary sound wave). The signal from the input sensor was used in a feedforward controller to generate another signal which was used to drive the loudspeaker positioned further down the duct. (Here up and down refer to the direction of propagation of the primary sound wave.) The secondary sound field generated by the loudspeaker destructively interfered with the primary sound field to create a quieter region downstream of the loudspeaker. The transfer function of the feedforward controller (sometimes referred to as the transfer characteristic, or frequency response, or impulse response), which relates the output signal of the controller to its input signal, was adjusted during a set-up phase and then left fixed. Since Lueg there were similar systems described which accomplished the same task but used more modern technology.

Lawson-Tancred (GB1492963) described a system for a duct but he also began to recognize that the input sensor would also respond to the cancelling field (secondary field) generated by the loudspeaker (or actuator). This made the controller more difficult to set-up as the degree of attenuation was potentially very sensitive to small changes in the feedforward transfer function. Consequently, small changes in the acoustics of the system being controlled (here a duct) or changes in the feedforward transfer function of the controller due to the 'drift' in analogue components resulted in poor sound reduction performance. One feature of his invention was to incorporate a passive attenuator between the loudspeaker and the input sensor. Swinbanks (GB1456018) made further improvements by using an input sensor which only responded to and an actuator which only generated sound in one direction, thus acoustically decoupling the input sensor and the actuator. Chaplin (GB1555760) recognized that this 'acoustic feedback' could be dealt with by electronic subtraction once digital technology was available with its more constant characteristic. This move to digital technology and the use of signal subtraction was a key aspect in the development of the technology.

Later Swinbanks (GB215480 and GB2054999) devised a scheme for adjusting the feedforward transfer function of the controller in response to the output from a second microphone positioned in the region where quiet was required (a residual sensor). The initial work envisaged manual adjustment of the feedforward transfer function, but recognized that the process could be automated. In 1981 Ross, one of the current inventors, (GB2097629/US4480333) showed how the transfer function of the feedforward part of the controller could be automatically adjusted using information from a residual sensor in the region where the sound attenuation was required. The term controller now being used to include the feedforward path and the mechanism which adjusts the transfer function of that feedforward path. One key aspect of that invention was the recognition of the fact that the feedforward transfer function was at its optimum when the correlation between the input sensor and the residual sensor was at a minimum. The automatic procedure minimized the correlation, and thus described a self-calibrating system which could be used to reduce the sound in a region around the residual sensor. When a system of this type is used in a duct, for example, because the primary field is propagating in one direction the region where quiet is produced is everywhere downstream of the actuator, and the residual sensor is at a point which represents the acoustic excitation in that whole region due to waves emanating upstream in the duct. Thus even though the region is large, and goes beyond the locality of the residual sensor, the residual sensor is in the region and the controller is functioning by minimizing the sound at the residual sensor.

In 1983 Chaplin (PCT/GB82/00299 or GB2107960) described the workings of a controller which was for controlling 'periodic' sounds made by rotating or repetitive sources. This controller used a tachometer as an input sensor. An input sensor of this type is only responsive to the phase of the (primary) vibration field as it is not responsive to the amplitude of the vibration. The system required sensors (microphones or accelerometers) in the region where quiet was required, and the operation of the system was to minimize the residual sensor signal's correlation with the input sensor signal. There have been further developments of such systems which envisage using multiple actuators to produce the secondary vibration field and multiple residual sensors to measure the combination of the primary and secondary vibration fields in the region where quiet is required. Nelson and Elliott (GB2149614) described a system of this type in which a few algorithms are used to reduce the sound in a cabin. The system minimizes the sum of the squares of the amplitude of the residual sensor signals.

All systems of this type will give the best vibration attenuation at or near to the residual sensors as they are designed to reduce the sound at those points. One fundamental problem has been that when an active vibration control system is installed in a passenger vehicle, for example, it is not convenient to position the residual sensors in the region where the sound reduction is required. Typically it is only acceptable to mount the microphones in the roof trim and yet the region in which the sound reduction is required is the region where the passengers' ears move. In the past, part of the skill in designing an active control system for such a cabin was in positioning the microphones so that when the sound was reduced at the microphones it also tended to be reduced in the region required. There have been many papers published which describe an internal sound field in terms of its 'acoustic modes' and how to position the microphones and loudspeakers best to reduce the modal amplitudes and thus the sound in the cabin generally (Nelson, Curtis, Elliott & Bullmore, Active minimization of harmonic enclosed sound fields, Journal of Sound and Vibration, 117(1),1-13). However, all of these papers have envisaged using control systems which minimize the sum of the squares of the amplitudes of the residual sensor signals.

Further publications by Nelson & Elliott (PCT/GB87/00706 and a paper which predates the application: Elliot and Sothers, A multichannel adaptive algorithm for the active control of start-up transients, EUROMECH 213, September 1986, Marseille) describe more control system algorithms for the minimization of the residual sensor signals. However, since the desire has been to reduce the sound in the cabin generally the publication suggests that the algorithms described achieve this, and do not specifically mention that they are limited to minimizing the sum of the squares of the residual sensor signals and that any attenuation of the primary sound field more generally is due to the careful positioning of the actuators and residual sensors.

A system for controlling the vibration in a helicopter structure generally is described by King (US4819182 and GB2160840) which uses multiple (12) residual sensors, however, the system again envisages that by reducing the vibration at the accelerometers the vibration of the whole structure will be reduced, but it does not describe a way of achieving this.

An earlier patent application by the current inventors (PCT/GB89/00964) discusses the provision of an active control system inside a cabin which achieves attenuation of the sound at passengers' head positions.

Finally, U.S. Patent No. 4,689,821 discloses a system in which one or more input sensors generate first signals related to the phase or amplitude of a primary vibration field produced by a propeller/turbofan, an actuator produces a secondary vibration field, monitoring sensors positioned in a first region produce third signals responsively to excitation by the primary and secondary vibration fields, and a controller including a feedforward path which responds to the first signals to produce second signals for reducing vibration in a second region excited by the primary and secondary vibration fields tends to be reduced. The controller adapts the transfer function of the feedforward path with reference to the first and third signals, but no account is taken of relationships between vibration fields as between the first and second regions or between these fields and the output of the controller.

Summary of the Invention

According to the present invention an active vibration control system comprises:

at least one input sensor which generates first signals related to the phase and/or the amplitude of a primary vibration field,

a plurality of actuators driven by second signals which produce a secondary vibration field,

a plurality of monitoring sensors positioned in a first region where they are excited by the combination of the said primary and secondary vibration fields and which produce third signals,

a controller including a feedforward path which is responsive to the said first signals and generates the said second signals so that the vibration in a second region, which is excited by the said primary and secondary vibration fields, tends to be reduced,

characterised in that the second region is different from the first region and is devoid of monitoring sensors, and in that the controller, by making use of a relationship determined by calibration between the primary vibration fields in the first and second regions, adapts the transfer function of the said feedforward path with reference to the said first and third signals, so that the vibration in the second region is maintained at a reduced level.

The relationship is preferably determined, at least in part, from a measured relationship between the third signals and fourth signals obtained from calibration sensors positioned in the second region only during calibration.

In any of the foregoing the controller's transfer function adaptation may also make use of the relationship between the secondary vibration fields in the first region and the controller output.

The controller's transfer function adaptation may also make use of the relationship between the secondary vibration field in the second region and the controller output.

In any of the foregoing the controller's transfer function adaptation may also make use of the relationship between the secondary vibration field in the first and second regions.

In any of the foregoing the controller's transfer function adaptation may be made so as to minimize the weighted sum of products of calculated signals which represent the expected vibration in the second region.

The controller's transfer function adaptation may be made so as to minimize the weighted sum of products of calculated signals which represent the expected vibration in the second region and the weighted sum of products of the transfer function elements.

The controller's transfer function adaptation may be made so as to minimize the weighted sum of products of the third signals where the weighted sum is chosen so that the vibration in the second region is reduced.

The controller may be calibrated using calibration sensors positioned in the second region which produce the said fourth signals representative of the vibration in the said second region.

The relationships A, B, C and D (as hereinafter defined) may be stored as required in the controller.

The relationships A, C or D (as hereinafter defined) may be adjusted in order to take account of changes.

The first region may be close to and/or bounded by the surface of a cabin interior to be occupied by at least one person, the second region being that region within the said cabin interior which will normally be occupied by the person's head.

The actuators may be positioned close to or on the interior surface of the said cabin interior, or the actuators may be incorporated in the structure surrounding the said cabin.

As applied to a motor vehicle such as a car, the primary vibration field will be that produced by the engine or tyres of the vehicle, the cabin interior being the passenger compartment of the vehicle and the actuators being incorporated in the engine or suspension mounting systems of the vehicle.

The first region may be close to and/or bounded by the surface of a body and the second region may be the surrounding space, and the actuators may be positioned on or close to the surface of the said body, or incorporated into the structure of the body.

The first and second regions may partly overlap.

The invention will now be described by way of example with reference to the accompanying drawings, in which Figure 1 shows the layout of prior art systems; and Figure 2 shows an embodiment of the invention.

In the prior art system shown in Figure 1 the residual sensors, 8, are positioned in the region where quiet is desired, 9. The input sensor, 2, is a single tachometer. The actuators, 1, are loudspeakers. The feedforward part of the controller, 7, has its transfer function adjusted in response to the signals from the residual sensor signals.

The present invention provides an active vibration control system which reduces the vibration in a (second) region without the need to position residual sensors in that region. The invention does require sensors (monitoring sensors) but these can be outside the second region, where the vibration is to be reduced. To recognize the fact that these monitoring sensors are not required to be in the second region they are described as being in a first region. When the system is operating, and the vibration in the second region is reduced (by the tendency of the primary and secondary fields to cancel each other), the vibration at the monitoring sensors (ie in the first region), which will also be a combination of the primary and secondary fields, will not necessarily have a lower amplitude than the primary field alone. The third signals (from the monitoring sensors) are used, in conjunction with the first (input sensor) signals, to adapt the transfer function of the feedforward part of the controller in order to maintain the attenuation in the second region.

In one embodiment of the invention this operation can be thought of as calculating the amplitude of the vibration in the second region from the values of the third signals and the first signals and adjusting the feedforward transfer function to minimize this calculated value.

It can thus be seen that the present invention differs from previously disclosed systems in that the additional sensors (in this case called monitoring sensors to distinguish them from the 'residual' sensors of previous systems) are not necessarily in the quiet region and thus the adaption of the feedforward transfer function of the controller of the present invention does not minimize the sum of the squares of the amplitude of these additional sensor signals.

One way in which the present invention operates to achieve this is now described with reference to Figure 2. The active vibration control system has actuators, 1, driven by second signals which produce a secondary vibration field, input sensors, 2, which generate first signals related to the phase and/or amplitude of the primary vibration field, monitoring sensors, 3, positioned in a first region, 4, which produce third signals, and a controller, 5, which generates the second signals in response to the first and third signals. The primary and secondary fields meet in a second region, 6, and the transfer function of the part of the controller between the first signals and the second signals (the feedforward path), 7, is adjusted in response to the first and third signals in order to maintain a reduction in the vibration in the second region.

The functioning of the controller will now be described in the frequency domain. However, this should not be considered as limiting the invention to systems which operate in the frequency domain. There are many equivalent time-domain systems (and other frequency-domain systems) which use follow the following formulation and in which those skilled in the art will see the correspondence. Examples of the correspondence between different types of time and frequency domain algorithms are given in Widrow & Stearns (Adaptive Signal Processing, Prentice-Hall, USA, 1985), and in Nelson & Elliott (PCT/GB87/00706).

During a calibration phase, before control is required, calibration sensors are mounted in the second region in order to give information about the acoustics of the system under control and the statistics of the signals. These calibration sensors produce fourth signals. The calibration sensors can be removed after the calibration proceedure has been finished.

The signals from the input sensors (the first signals) are represented by u (where u is a vector of i complex numbers for a single frequency, with each element corresponding to one of the i first signals). The n second signals (which are used to drive the n actuators) are represented by an n-long vector x. The third signals (from the monitoring sensors) due to the primary field alone are represented by an r-long vector Y _m, and P _m when the primary and secondary fields are present. The fourth signals (from the calibration sensors) due to the primary field alone are represented by a t-long vector Y _c, and P _c when the primary and secondary fields are present.

The transfer function of that part of the controller between the first and second signals is represented by a complex matrix Q (which has n rows corresponding to the n second signals, and i columns corresponding to the i first signals). Thus: x = Qu.

The transfer function relating the second signals to the third signals is represented by the complex matrix A (which has r rows and n columns) and the transfer function relating the second signals to the fourth signals is represented by the complex matrix D (which has t rows and n columns). Thus: P c = Y c + DQu, and P m = Y m + AQu.

It is a primary aim of the system to minimize the vibration field in the second region. This is achieved by positioning the calibration sensors in the second region at a set of points which are representative of the vibration characteristics of the said second region. The output of the calibration sensors is described above as the fourth signals. In one embodiment of the invention the active vibration control system operates by calculating the expected value of the fourth signals from the third signals and the first signals and adjusts the feedforward transfer function to minimize the weighted sum of the squares of these calculated signals.

During part of an initial calibration phase the relationship between the parts of the fourth and third signals that are correlated with the input signals and which are due to the primary field only can be measured, and can be expressed in the following way: {Y c u*}{uu*}-1 = C{Y m u*}{uu*}-1 + e.

Where C is a complex matrix with t rows and r columns which is chosen to minimize the norm of e, and {..} is the expectation operator.

Thus when the control system is operating the expected value of the fourth signals is approximated by = C ( P m - AQu ).

If the fourth signals were available and the aim of the system were to minimize the weighted sum of the products of these signals, P _c*GP _c (where G is a full, t-by-t, complex, Hermitian matrix), then the optimum feedforward transfer function Q_opt is: Qopt = - (D*GD)-1D*G{Y c u*}{uu*}-1.

If, additionally, the variance of the second signals were to be minimized so that the function to minimize was P _c*GP _c + ax*Hx (where H is a complex Hermitian matrix) then the optimum feedforward transfer function would be: Qopt = - (D*GD+aH)-1D*G{Y c u*}{uu*}-1.

During the calibration phase no attempt is made to minimize the vibration in the second region but the following quantities can be measured:

D (the relationship between the fourth signals and the second signals) this is either measured with the primary field off by supplying suitable second signals during this first part of the calibration, x _t, to each actuator, which excite a secondary vibration field, and recording the resulting fourth signal, p _c, in which case the signal to noise ratio is good, or, alternatively, with the primary field on when the signal to noise ratio can be improved by increasing the level of the second signal during this first part of the calibration. The estimate of D is {p c x t*}{x t x t*}-1.

A (the relationship between the third signals and the second signals) this is either measured with the primary field off by supplying other suitable second signals during this second part of the calibration, x _r, to each actuator, which excite a secondary vibration field, and recording the resulting third signal, p _m, in which case the signal to noise ratio is good, or, alternatively, with the primary field on when the signal to noise ratio can be improved by increasing the level of the calibration second signals during this second part of the calibration. The estimate of A is {p m x r*}{x r x r*}-1.

C (the relationship between the part of the fourth and third signals which is correlated with the input signals and which is due to the primary field alone varying over the whole range of expected primary fields) this is measured during the third part of the calibration by allowing the primary field to vary over the whole range of expected fields and recording Y _c, Y _m, u during the variation. The estimate of C is then calculated as {Y c u*}{uu*}-1{uu*}-1{uY m*}[{Y m u*}{uu*}-1{uu*}-1 {uY m*}]-1.

When the system is first switched on, the controller's feedforward transfer function can be set equal to the value given in equation (6) or (7). However, once the system has been operating for some time the vibrational characteristics may have changed and it will be desirable to adapt the controller's feedforward transfer function in order to compensate for the changes. This now requires the calculated value of the fourth signal to be used as the calibration sensors are no longer available.

The new value of the optimum feedforward transfer function is: Qopt' = - (D*GD+aH)-1D*G{C(P m - AQopt u)u*}{uu*}-1.

It may also be updated in a gradient descent algorithm of the form: Qopt' = (1-f)Qopt - cD*G{CP m u* + (D-CA)Qopt uu*}{uu*}-1. where c is the step size and f is a small factor to allow the size of the second signals to be limited whilst minimising the vibration in the second region.

During operation of the controller it is desirable not only to adapt the feedforward transfer function but also to adjust the stored values of any parameters on which the adaption process depends and which may have changed. It is possible to re-estimate the value of A, on line, directly, using 'system identification', such as described by Chaplin (GB2107960). It is more difficult, on the other hand, to maintain an accurate value of either C or D.

Modification of D. The transfer function D represents the signal path from the second signals, through the actuators, through the vibration bearing medium, to the second region. The transfer function A represents the signal path from the second signals, through the actuators, through the vibration bearing medium, to the first region, and finally through the monitoring microphones. It will thus be seen that some elements of the two transfer functions are similar and so it may be appropriate to model the transfer function D as the product BA, where B is chosen so that D=BA. Changes in A can then be used to update D by assuming that B remains constant.

Choice of C. The initial choice of C, which satisfies equation (4), is given by equation (10). However, it is likely that the matrix [{Y _m u*}{uu*}^-1{uu*}^-1{uY _m*}] which covers the range of expected primary vibration fields will be ill-conditioned and thus computation of equation (10) will be inaccurate. This indicates that there is some flexibility in the choice of C for excitations which do not fall in the sub-space [{Y _m u*}{uu*}^-1{uu*}^-1{uY _m*}]. When the primary field falls outside that sub-space it is desirable that the performance of the system is maintained. This could be, for example, that the system minimizes the weighted sum of the products of the monitoring sensor signals (third signals) for fields outside the class, and this can be accomplished by ensuring that the matrix C tends to be such that D-CA=0 (or C tends to be equal to B) for the class of excitations outside the known sub-space. In this situation the system operates to minimise the sound in the second region by minimising a weighted sum of products of the third signals, the weights being selected so that the vibration in the second region is minimised.

An alternative method of finding C would extend the range of primary vibration fields measured during the third part of the calibration by using additional sources, driven by additional test signals, and close to the positions of the real sources. This would ensure that the system would continue to minimise the weighted sum of products of the expected fourth signals for all classes of expected primary fields and those created by the additional test signals which would be chosen to represent likely departures of the primary field from its normal state.

The subject invention is particularly useful for controlling sound in a cabin or passenger compartment of a passenger vehicle. In that case the monitoring sensors could be placed in or close to the trim of the cabin walls, or in or close to the seats and the second region would be the region in the cabin which would be occupied by the heads of the passengers. The actuators for producing the secondary vibration field could be loudspeakers positioned in or close to the trim of the cabin or in or close to the seats. Alternatively the actuators could form part of the structure of the cabin walls. For example, they could be electro or magneto strictive materials attached to the cabin walls which cause the cabin walls to vibrate. A further alternative form of actuators could be positioned in or close to the suspension bushings or engine mounts or any other part of the vibration path from a source of sound in a vehicle cabin to the cabin interior. In this case the monitoring sensors may be better positioned close to the actuators and not in the trim as first described.

The invention will also be seen to be useful in controlling the noise radiated from an object when it is difficult or impractical to position residual sensors in the far field. This would be the case in controlling the radiated noise from a transformer casing, a ship, or the tyre of a road vehicle. In any of these cases the monitoring sensors would likely be positioned close enough to the surface of the radiating body to be in the near field of both the actuators and the body itself. Consequently, whilst the primary and secondary fields will combine they will not combine in a way to cancel each other when the cancellation is at its optimum in the far field. The present invention is thus particularly useful in controlling noise from such objects.

It is understood that the invention is not limited to situations where the first and second regions do not overlap.

Claims

An active vibration control system comprising:

one or more input sensors (2) which generate first signals related to the phase and/or the amplitude of a primary vibration field,

a plurality of actuators (1) drivenby second signals which produce a secondary vibration field,

a plurality of monitoring sensors (3) positioned in a first region (4) where they are excited by the combination of the said primary and secondary vibration fields and which produce third signals,

a controller (5) including a feedforward path which is responsive to the said first signals and generates the said second signals so that the vibration in a second region (6), which is excited by the said primary and secondary vibration fields, tends to be reduced,

characterized in that said second region (6) is different from said first region (4) and is devoid of monitoring sensors, and in that the controller (5), by making use of a relationship determined by calibration between the primary vibration fields in the first and second regions (4, 6), adapts the transfer function of the said feedforward path with reference to the said first and third signals, so that the vibration in the second region is maintained at a reduced level.
A system according to claim 1, in which the relationship is at least in part determined from a measured relationship between the third signals and fourth signals obtained from calibration sensors positioned in the second region (6) only during calibration.
A system according to claim 1 or claim 2, in which the adaptation of the controller's transfer function also makes use of the relationship between the secondary vibration field in the first region (4) and the output of the controller (5).
A system according to claim 1 or claim 2 or claim 3, in which the adaptation of the controller's transfer function also makes use of the relationship between the secondary vibration field in the second region (6) and the output of the controller (5).
A system according to any of claims 1 to 4, in which the adaptation of the controller's transfer function also makes use of the relationship between the secondary vibration field in the first and second regions (4, 6).
A system according to any of the preceding claims, in which adaptation of the controller's transfer function is made so as to minimize the weighted sum of products of calculated signals which represent the expected vibration in the second region (6).
A system according to any of claims 1 to 6, in which the adaptation of the controller's transfer function is made so as to minimize the weighted sum of products of calculated signals which represent the expected vibration in the second region (6) and the weighted sum of products of the transfer function elements.
A system according to any of claims 1 to 6, in which the adaptation of the controller's transfer function is made so as to minimize the weichted sum of products of the third signals where the weighted sum is chosen so that the vibration in the second region (6) is reduced.
A system according to claim 2 or any claim appendant thereto, in which the controller (5) is calibrated using calibration sensors positioned in the second region (6) which produce the said fourth signals representative of the vibration in the said second region.
A system according to any of the preceding claims, in which any of the relationships A, B, C or D as hereinbefore defined are stored in the controller (5).
A system according to any of the preceding claims, in which any of the relationships A, C or D as hereinbefore defined are adjusted in order to take account of changes.
A system according to any of the preceding claims, in which the first region (4) is close to and/or bounded by the surface of a cabin interior and the second region (6) is that part of the said cabin interior where the occupants' heads tend to be.
A system according to claim 12, in which the said actuators (1) are positioned close to or on the interior surface of the said cabin interior.
A system according to claim 12, in which the said actuators (1) are incorporated in the structure surrounding the said cabin.
A system according to claim 12, in which the primary vibraration field is produced by the engine or tyres of a car, the said cabin interior is in the said car and the said actuators (1) are incorporated in the engine or suspension mounting systems.
A system according to any of claims 1 to 11, in which the first region (4) is close to and/or bounded by the surface of a body and the second region (6) is the surrounding space.
A system according to claim 16, in which the said actuators (1) are positioned on or close to the surface of the said body.
A system according to claim 16, in which the said actuators (1) are incorporated in the structure of the body.
A system according to any of the preceding claims, in which the first and second regions (4, 6) partly overlap.