GB2122052A - Reducing noise or vibration - Google Patents

Abstract

To reduce the amplitude of noise or vibration produced by a rotating machine or other repetitive source, apparatus is provide comprising transducers (28, 30) positioned at mutually spaced locations for detecting noise or vibration at that location and for providing electrical signals representative thereof. An inducer (24, 26) is provided at each location and is arranged to induce noise or vibration at that location in response to a control signal. A processing arrangement (36, 38, 40) is provided for receiving the electrical signal from each transducer (28, 30) and for modifying the signal in dependence upon a predetermined transfer function which is representative of the mutual interaction between each spaced locations and each of the other spaced locations. This provides a control signal for the inducer (24, 26) and thereby induces a compensating noise or vibration. <IMAGE>

Description

SPECIFICATION Method of and apparatus for reducing noise or vibration This invention relates to the reduction of noise or vibration produced by repetitive sources and more particularly but not solely to the reduction of vibration induced in structures by machinery and the reduction of inherent noise induced by such vibration.

In industrial and marine environments increasing concern is being given to controlling the low-frequency noise and vibration produced by rotating machinery. Conventional techniques of noise reduction (e.g. silencers or anti-vibration mounts) are ideal for reducing high frequency noise but the application of those techniques to low frequencies often results in an unacceptably large, heavy or expensive installation. In recent years there have been some advances in the cancellation of duct-borne noise by means of the introduction of a synthesised signal, similar in form and magnitude but of opposite phase to the noise. Such cancellation obviously requires a knowledge of the noise signal and/or the combined noise and anti-phase signals and various methods have been devised to obtain such knowledge.In many of these methods great care has been devoted to ensuring that the antiphase signal does not interfere with the monitoring of the noise signal.

We have investigated the reduction of structural vibration using a single cancelling force to oppose a single-point vibratory input. Some improvement is possible by such an approach in some circumstances. However, the use of a single point compensation is unsatisfactory in most circumstances because the compensation necessary to nullify the vibration at that point will often induce vibration at other positions in the structure.

Similar problems can arise in the cancellation of gas-borne noise by the provisions of an opposing noise at a single position. The present invention seeks to provide a method and apparatus in which these problems are overcome or at least substantially reduced.

Accordingly to one aspect of the invention there is provided a method of reducing the amplitude or noise or vibration produced by a repetitive source by generating compensating noise or vibration at at least two spaced locations, in which the compensating noise or vibration provided at each location is dependent upon the detected noise or vibration at that location modified by a predetermined transfer function which is representative of the mutual interaction between each of said spaced locations and each of the other of said spaced locations.

By providing a compensating noise or vibration at more than one location and by tailoring the compensation in accordance with the predetermined transfer functions which can be measured for a particular arrangement, a highly effective reduction of noise or vibration results.

In a refinement of the method, information is stored relating to the transfer function occuring at a range of different frequencies and the major frequency components of the detected noise or vibration are separated and modified by the appropriate transfer function for that frequency.

The transfer function may be obtained by interpolation from the transfer function relating to the nearest one of the range of different frequencies.

In a particularly advantageous form of the method detected noise or vibration is repetitively sampled to provide values for processing with the transfer function. The samples may be Fourier transformed to produce a detected signal in the frequency domain prior to modification by the transfer function. The samples may be converted from analogue to digital form prior to Fourier transformation, the transformed digital values being converted back into the time domain by reverse Fourier transformation and then converted into analogue form to provide compensating noise or vibration inducing signals.

In accordance with another aspect of the invention there is provided an apparatus for reducing the noise or vibration produced by a repetitive source, comprising at least two transducers positioned at mutually spaced locations for detecting noise or vibration at that location and for providing electrical signals representative thereof, an inducer for each location arranged to induce noise or vibration at that location in response to a control signal, processing means for receiving the electrical signal from each transducer and for modifying the signal in dependence upon a predetermined transfer function which is representative of the mutual interaction between each of said spaced locations and each of the other said spaced locations to provide the control signals for the inducer and thereby induce a compensating noise or vibration.

The processing means may include a store for containing the predetermined transfer function.

The store may have a capacity for containing information relating to the transfer function occuring art a range of different frequencies and the processing means may comprise frequency identification means for separating major frequency components of the electrical signals from the transducers and for routing the components for modification by the appropriate transfer function for that frequency, and summing means for each transducer combining the modified frequency components thereby to provide the control signal.

In one form the processing means comprises a linear interpolator for determining the appropriate transfer function, for each frequency component of the electrical signals from the transducers, from the transfer function relating to the nearest one of the range of different frequencies.

The processing means may include a sampling device for providing values for processing to provide the control signals. The processing means may further comprise a Fourier transform processor for converting the time domain samples into the frequency domain prior to modification by the transfer function. An analogue to digital converter may be provided for converting the sampled values into digital form prior to processing to provide the control signals. A digital to analogue converter may also be provided for converting the digital control signals into analogue control values.

The processing means may include a store for receiving control signals for driving each inducer to provide the compensating noise or vibration and gating means for repetitively routing the stored control signals to the inducers in synchronism with the detected noise or vibration.

In an apparatus for cancellation of vibration produced by a rotating machine the processing means may comprise a tachogenerator for providing synchronizing signals for actuation of the gating means.

In an apparatus for cancelling vibration the inducers may be electro magnetic shakers which are responsive to the control signals.

In order that the invention and its various other preferred features may be understood more easily, a laboratory model constructed to prove the viability of the principle will be described with reference to the drawings in which: Figure 1 is an elevational view of the model built to simulate a practical vibration situation.

Figure 2 is a schematic block diagram showing the control functions of an apparatus constructed in accordance with the invention in use with the model of Figure 1.

Figure 3 is a schematic diagram used to illustrate the principle of the invention; Figure 4 is a further schematic diagram used to illustrate the principle of the invention.

Referring now to Figure 1 the model comprises a simple beam structure that has most of the important vibrational characteristics of a practical full scale structure. A cross beam 10 is the equivalent of a machinery raft whilst two parallel beams 12 and 14 represent the machine seating.

Two pairs of anti vibration mounts 1 6 and 1 8 support the beam 10 on the beam 12 and 14 represent the shock mounts of a full size system.

The beam structure is isolated from ground vibrations by concrete blocks supported isolators and is supported thereon by anti vibration mounts 20. A shaker 22 of the moving coil electromagnetic type is mounted at the centre of the cross beam 10 and simulates the machinery of a real system. Similar shakers 24 and 26 are mounted on the cross beams 12 and 14 directly below the anti vibration mounts 1 6, 1 8 for providing cancellation vibrations at these points.

Each beam 1 2, 14 is provided with an accelerometer 28, 30 mounted adjacent the antivibration mount and accelerations measured at these points are used as a basis for determination of the cancellation forces. Alternatively, force measuring transducers may be provided to monitor the force through each mount to determine the force which is to be opposed and these may be used to determine opposing forces to be applied.

The rig of Figure 1 is shown coupled with a vibration reducing apparatus in Figure 2. The shaker 22 is driven by a signal generator 32 via a power amplifier 34 and in this way it is assured that the vibration induced is truly repetitive.

The output of the accelerometers 28 and 30 are coupled with a dual channel analogue to digital converter and fast Fourier transformer 36 formed by a Hewlett Packard type HP5420.

Transformed data is fed from here to a computer 38 formed by Hewlett Packard type HP9835A which provides an output to a digital analogue converter 40. This is a specially designed type built to enable the output of a continuous fixed analogue signal in response to a one cycle input of digital information. This enables the computer 38 to calculate the desired control signal and send it to the D/A converter 40. The block 36 is then free to calculate the next control signal leaving D/A converter 40 to continue its output under its own control. The output of the D/A converter 40 is routed to the shakers 24 and 26 via power amplifiers 42 and 44 respectively.

Initially thirty-two time samples per cycle were used, but this number can be altered if experiments suggest it. The HP5420 of block 36 works with five hundred and twelve points, which is sixteen cycles of thirty-two points each. Using a frequency multiplier 46, a tachometer signal from the machine can be stepped up by a factor of thirty-two and used to drive the sampling rate of the HP5420, enabling exactly thirty-two equispaced time samples to be taken per cycle, even if the fundamental frequency of the machine varies (provided it does not vary too rapidly). Because sixteen machine cycles fit exactly into the HP5420's analysis window, rectangular weighting can be used, and no windowing problems arise.Fourier Transforming the sixteen cycles in the HP5420 will result in a spectrum with a bandwidth covering the first sixteen harmonics of the fundamental frequency, oversampled by a factor of sixteen. Because the input signals are genuinely repetitive, their spectra will consist of lines at the harmonics of the fundamental frequency. Hence the HP5420 will display sixteen lines, with fifteen zeros between one line and the next, for each of the two input signals.

This information is transferrd to the computer 38, which discards the irrelevant zeros and preserves only the sixteen -harmonic values. As the Hop 5420 sampling rate is driven externally, there is no information from the HP5420 concerning the absolute fundamental frequency measuredonly that the spectral lines measured are harmonics of the fundamental.

Therefore, at the same time as the HP5420 halts its data acquisition, the computer 38 requests the fundamental frequency from a frequency counter 48 connected to the tachometer signal.

in order to construct, C, the control signals, the sensor signals, S, are multiplied by M-1 at each frequency. M was measured at five hundred and twelve frequencies across the range of interest, but it is unlikely that exactly the correct frequencies were chosen to match with the actual harmonics of the machine, and therefore suitable values are calculated by linear interpolation from the nearest frequencies in -M-1. The matrix multiplication is performed to give C- M-1S at sixteen frequencies.

The two frequency domain representations of the control signals need to be inverse Fourier transformed to obtain two thirty-two point time domain representations. Each signal is sent to the HP5420, and zero packed at the end to make the sixteen points up to two hundred and fifty-six. The FFT gives five hundred and twelve time domain points, over-sampled by a factor of sixteen.

Therefore, in calling back the data to the computer 38 only every sixteenth point is kept; the remainder are discarded.

The two time domain representations of the control signal, each of thirty-two points, are sent to the D/A converter 40, which stores them and cycles through, outputting to the shakers 24, 26, at a rate governed by the frequency multiplier to ensure synchronisation with the machine speed.

Once the structure has settled down to the response from the control shakers, another control cycle can be initiated using the residual vibration as input and adding the new control signal to the old before sending the total control signal to the D/A converter.

It is unlikely that the velocities imposed by the shakers will exactly balance the initial system velocities at the isolators. It can be shown that in order to achieve a reduction of 20 dB in a particular frequency component, the cancelling frequency component must be within 1 dB in amplitude and 60 in phase. This would seem to be quite a stringent requirement, especially near resonances where amplitude and phase change quickly, for the following reasons: (a) Inaccuracy in the measurement of the transfer functions and vibration waveform due to inadequate frequency resolution and rounding errors within the analyser.

(b) Inaccuracies associated with digital approximation to an analogue signal.

(c) Non-linearities in the controlled system and the controlling shakers.

(d) Transverse sensitivity of the accelerometers giving components due to motion in orthogonal directions.

(e) Drift of machine shaft speed.

(f) Changing machine force pattern.

(g) A less than ideal brick wall anti-aliasing filter characteristic.

It is clear, therefore, that it is desirable that some form of feedback adaptibility is included in the cancelling alogrithm. Without this adaptive element the control will be open loop and therefore subject to the well known limitation of poor sensitivity to system parameter variation.

The form that the adaptive part of the algorithm finally takes is probably best determined by experimentation. There are three possible candidates for trial and error adjustment: (a) Magnitude and phase of the system transfer function matrix at the frequencies of interest. This suffers from the disadvantage that the adjustment is made before the system transfer function matrix inversion. This is likely to be a time consuming procedure so that the overall cancellation time constant will be long.

(b) Magnitude and phase adjustment of the harmonic components of the output voltage vector before inverse transforming into the cancelling waveform. Aithough better as it will be carried out after the matrix inversion so that speed is better it suffers from the disadvantage, in common with (a) that it depends on system linearity.

(c) Magnitude adjustment of the reconstituted waveform in Time. This is the most likely answer as it does not depend on system linearity and its response time will be better than (a) and (b).

However, cross coupling between the inputs may make difficult the determination of whether or not the alteration was successful.

In each case, however, the algorithm would have residual power averaging for either discrete frequencies or the total signal and better/worse routines in order to determine whether the action taken was beneficial or not.

An analysis of the principles upon which the present invention as based will now be given in order to aid the understanding:-- A repetitive signal allows any controlling algorithm a complete view of the future, and so the required controlling action can be calculated before the event. Random signals in a system with no time delay between the sensor position and the input of control do not allow this total knowledge of the future and so control must be achieved by attaching shakers in order to alter system parameters such as damping, resonances, etc., rather than "cancelling" the vibration.

Classical Control Theory has been directed towards random signal control, the usual technique being to incorporate in a systemmodel; shakers, amplifiers, etc., and then calculate the parameters these components should have for the whole system to respond in the required way. This method does require the system-machine, machine raft, seating, etc.,-to be modelled. The correct oarameters need to be calculated for each installation seDaratele.

Repetitive signals allow a totally different approach, that of adaptive control. The vibration is measured, control calculated and applied, and the result assessed. The control input is adjusted according to success or failure. This has the advantage of giving theoretically 'perfect' control and is able to be applied to any structure without needing a detailed mathematical model.

The assumption that the disturbing force is repetitive, such as from rotating machinery, allows a control algorithm to measure the vibration during one cycle of the machine, spend some time calculating the required control input, and then apply this control at some later cycle and be guaranteed that the disturbing force is still the same.

The algorithm can then do the same again, measuring the vibration left after the first control operation and calculating a control signal to cancel this, and adding it to the first control signal. By repeated application of the control algorithm the vibration is reduced-even if the control signals are imperfect and full cancellation does not occur on the first pass; as long as the vibration is reduced, subsequent control cycles will lower the residual vibration gradually.

When there is more than one control point, cross-coupling can cause problems. The vibration level at, for example, position 1 is not only affected by the control shaker at 1, but also by the shaker at another position, 2. Depending on the modes excited in the structure at certain frequencies the effect at 1 may be due predominantly to the shaker at 2. This, combined with point impedances which are frequency dependent, requies a more subtle control than simply introducing a 1 800 phase shifted version of the vibration present at a given point.

In the time domain the effects of crosscoupling are hard to analyse. An input at shaker 2 will have a time delay before it reaches measurement position 1 and there will be a 'ringon' dependant upon the structure. If the structure can be considered as linear, working in the frequency domain makes cross-coupling more evident. The response at position 1 is simply the input at 1 multiplied by the point impedance at 1, summed to the input at 2 multiplied by the transfer function from shaker 2 to sensor 1. In symbols writing:

where S is the sensor signal at position i.

where iC is the control signal at position i.

where Mjj is the transfer function from the shaker at j to the sensor at then'S=MC In the absence of any other forces.

Each Mjj is a complex array defining the amplitude and phase change for each frequency point.

in general M will be invertible and hence If a given output signal S is required, then the input C needed to create this is given by C=M-1S Applying this to the algorithm described before, let

be the initial sensor signals due to the machine.

T6 create the opposite vibration with the control input, choose, C1=-M1S0-See Figure 3.

The resulting sensor output S, is the sum of the initial output SO and that due to the control (see Figure 4): S1=S0+MC1 =S,+M(-M-'S,) =SOSO =O In practice, due to various errors in measurement and changes in the disturbing force, S, will not be zero. Hence, a second additional control signal will be applied choosing C2=-M-1S1 giving S2=So+M(Ca+c2) This is repeated as often as required to obtain a control signal that reduces the vibration level the desired extent.

The illustrative embodiment described employs only two shakers to apply vibration cancelling forces at two spaced locations.

The system described can be modified to reduce the computation time required by elimination of inverse transfer function calculations. This is enabled by previously measuring or using previously calculated transfer function data to provide a new cancellation algorithm which is based on reducing the squares of the frequency components of the residual signal using a "maximum gradient" method. The algorithm requires the control frequency spectrum to be multiplied by a constant (between 0 and 1) which is determined experimentally.

Another modification is concerned with one of the major problems of noise cancellation in the frequency domain i.e. the instability of the transfer functions. To overcome this an algorithm has been developed to measure these either before or while the cancellation is in progress. The first update can be made after N cycles, where N is the number of cancellation points. The main drawback is that the new values can be measured only at the fundamental frequency and harmonics of the system. To accommodate any frequency drift these values can be extrapolated to provide the first estimate before a new transfer function can be calculated at the new frequency. The success of this technique depends upon the frequency of the system remaining constant during the N cycles that it takes to calculate the new transfer functions.

The incorporation of either one of the before mentioned modifications is intended to fall within the scope of this invention. The incorporation of both of the before mentioned modifications can result in a significant reduction in the number and required accuracy of measurements made prior to cancellation.

It will be appreciated that the principles of this invention are applicable to arrangements where cancellation is effected at more than two locations.

Although the embodiment described is concerned with cancellation of vibration in structures, the same principles are invovled in cancellation of gas-borne e.g. air-borne noise by providing cancelling sound waves at a plurality of spaced locations. Such arrangements are intended to fall within the scope of this invention and will employ sound generating transducers instead of vibration inducing shakers.

The invention has many practical applications, examples of which are: (i) Royal Navy and Merchant Ship machinery.

(ii) Industrial machinery.

(iii) Scientific Equipment.

(iv) General transport vibration.

In the instances mentioned above the requirement may be to: (i) reduce transfer of vibration from machine to structure, (ii) reduce the motion of the machine/equipment.

(iii) isolate the equipment from ground/structure borne vibration.

Claims (18)

Claims

1. A method of reducing the amplitude of noise or vibration produced by a repetitive source by generating compensating noise or vibration at at least two spaced locations, in which the compensating noise or vibration provided at each location is dependent upon the detected noise or vibration at that location modified by a predetermined transfer function which is representative of the mutual interaction between each of said spaced locations and each of the other of said spaced locations.

2. A method as claimed in claim 1, wherein information is stored relating to the transfer function occuring at a range of different frequencies and the maior frequency components of the detected noise or vibration are separated and modified by the appropriate transfer function for that frequency.

3. A method as claimed in claim 2, wherein the appropriate transfer function is determined by interpolation from the transfer function relating to the nearest one of the range of different frequencies.

4. A method as claimed in claim 1,2 or 3 wherein the detected noise or vibration is repetitively sampled to provide values for processing with the transfer function.

5. A method as claimed in claim 4, wherein the samples are Fourier transformed to produce a detected signal in the frequency domain prior to modification by the transfer function.

6. A method as claimed in claim 5, wherein the samples are converted from analogue to digital form prior to Fourier transformation, the transformed digital values are converted back into the time domain by reverse Fourier transformation and then converted back into analogue form to provide compensating noise or vibration inducing signals.

7. A method of reducing the amplitude of noise or vibration produced by a repetitive source substantially as described herein.

8. An apparatus for reducing the noise or vibration produced by a repetitive source, comprising at least two transducers positioned at mutually spaced locations for detecting noise or vibration at that location and for providing electrical signals representative thereof, an inducer for each location arranged to induce noise or vibration at that location in response to a control signal, processing means for receiving the electrical signal from each transducer and for modifying the signal in dependance upon a predetermined transfer function which is representative of the mutual interaction between each of said spaced locations and each of the other said spaced locations to provide the control signals for the inducer and thereby induce a compensating noise or vibration.

9. Apparatus as claimed in claim 8, wherein the processing means includes a store for containing the predetermined transfer function.

10. Apparatus as claimed in claim 8, wherein the store has a capacity for containing information relating to the transfer function occuring at a range of different frequencies and the processing means comprises frequency identification means for separating major frequency components of the electrical signals from the transducers and for routing the components for modification by the appropriate transfer function for that frequency, and summing means for each transducer combining the modified frequency components thereby to provide the control signal.

11. Apparatus as claimed in claim 10, wherein the processing means comprises an interpolator for determining the appropriate transfer function, for each frequency component of the electrical signals from the transducers, from the transfer function relating to the nearest one of the range of different frequencies.

12. Apparatus as claimed in any one of claims 8 to 11, wherein the processing means includes a sampling device for providing values for processing to provide the control signals.

13. Apparatus as claimed in claim 12, wherein the processing means comprises a Fourier transform processor for converting the time domain samples into the frequency domain prior to modification by the transfer function.

14. Apparatus as claimed in claim 13, wherein the processing means includes an analogue to digital converter for converting the sampled values into digital form prior to processing to provide the control signals.

1 5. Apparatus as claimed in claim 14, wherein the processing means includes a digital to analogue converter for converting the digital control signals into analogue control values.

1 6. Apparatus as claimed in any one of claims 8 to 15, wherein the processing means includes a store for receiving control signals for driving each inducer to provide the compensating noise or vibration and gating means for repetitively routing the stored control signals to the inducers in synchronism with the detected noise or vibration.

1 7. Apparatus as claimed in claim 1 6 for cancellation of noise or vibration produced by a rotating machine, wherein the processing means comprises a tachogenerator for providing synchronising signals for actuation of the gating means.

18. Apparatus as claimed in any one of the preceding claims for cancellation of vibration, wherein the inducers are electro magnetic shakers which are responsive to the control signals.

1 9. An apparatus substantially as described herein with reference to the experimental rig illustrated in the drawings.