VIBRO-ELECTRIC CONDITION MONITORING

Abstract

Apparatus (10) for monitoring the condition of an item of electrical equipment (1) whilst in operation comprises a vibration sensor (11) and an electrical sensor (12) operable to detect a characteristic operational electrical signal of the equipment (1). The output of the vibration sensor (11) and the electrical sensor (12) is supplied to a spectrum generator (13) and then to a processing unit (14) operable to process the respective frequency spectrums to generate a frequency response function. Once a frequency response function is generated, the processing unit (14) is operable to compare the generated frequency response function to a model frequency response function. This allows any variations between the generated frequency response function and the model frequency response function to be identified. This could be indicative of a fault and could provide an identification of the nature of the fault.

Claims

1. A method for monitoring the condition of electrical equipment, the method comprising the steps of: detecting vibration of the equipment; obtaining a frequency spectrum of the detected vibration; detecting a characteristic operational electrical signal of the equipment; obtaining a frequency spectrum of the detected characteristic operational electrical signal; processing the respective frequency spectrums to generate a frequency response function and comparing the generated frequency response function to a model frequency response function so as to identify any variations between the generated frequency response function and the model frequency response function.

2. A method as claimed in claim 1 wherein the method involves detection of more than one characteristic operational electrical signal.

3. A method as claimed in claim 1 wherein more than one attribute of the characteristic operational electrical signal is measured, and the method includes generation of separate frequency response functions for each attribute and comparison of the separate generated frequency response functions for these attributes.

4. A method as claimed in claim 1 wherein the method involves detection of vibration of the equipment at multiple points.

5. A method as claimed in claim 1 wherein the method includes generating separate frequency response functions for each vibration sensor.

6. A method as claimed in claim 1 wherein the method includes the step of calculating the auto spectrum of detected vibration and characteristic operational electrical signals and the cross spectrum of detected vibration and characteristic operational electrical signals.

7. A method as claimed in claim 6 wherein the method includes the step of monitoring the calculated auto and cross spectra over time.

8. A method as claimed in claim 6 wherein the monitoring of the calculated auto and cross spectra involves any one or more of: calculating frequency response functions, coherence; transmissibility (operational transfer path analysis); or principal component analysis.

9. A method as claimed in claim 1 wherein the method includes: generating a force frequency response function from the vibration frequency response function; and comparing the determined force generated frequency response function to a model force frequency response function.

10. A method as claimed in claim 9 wherein determination of the force frequency response function from the measured vibration frequency response function is achieved by inverse methods.

11. A method as claimed in claim 1 wherein the method includes classifying the operation of the equipment in response to any identified variations between the generated frequency response function and the model frequency response function or any identified variations in the calculated auto and/or cross spectra.

12. A method as claimed in claim 8 wherein the method includes the analysis of any identified variations to determine whether a fault has occurred and/or the identity of the fault.

13. A method as claimed in claim 12 wherein the method includes the further step of outputting a signal indicative of the fault; generating a maintenance notification including an indication of the identified fault; or outputting a command signal to shut down all or part of the equipment.

14. A method as claimed in claim 1 wherein the model frequency response function is generated by modelling the expected frequency response of the equipment.

15. A method as claimed in claim 1 wherein the model frequency response function is generated by way of a calibration process.

16. A method as claimed in claim 15 wherein the calibration process is carried out at the completion of manufacture of the equipment; on installation of the equipment; periodically or after servicing.

17. A method as claimed in claim 1 wherein the method includes converting signals from the frequency domain to the time domain for analysis.

18. A method as claimed in claim 1 wherein the method includes monitoring multiple items of electrical equipment.

19. A method as claimed in claim 1 wherein the method includes monitoring machinery linked to electrical equipment.

20. An apparatus for monitoring the condition of electrical equipment, the apparatus comprising: a vibration sensor operable to detect vibration of the equipment and output a signal indicative thereof; an electrical sensor operable to detect a characteristic operational electrical signal of the equipment and to output a signal indicative thereof; a spectrum generator operable to receive the output of the vibration sensor and electrical input sensor and to thereby generate a frequency spectrum of the detected vibration and a frequency spectrum of the detected characteristic operational electrical signal; and a processing unit operable to process the respective frequency spectrums to generate a frequency response function and to compare the generated frequency response function to a model frequency response function so as to identify any variations between the generated frequency response function and the model frequency response function and output an indication thereof.

21. An apparatus as claimed in claim 20 wherein there are multiple characteristic operational electrical signal sensors.

22. An apparatus as claimed in claim 20 wherein the electrical signal sensors are operable to detect any attribute of the characteristic operational electrical signal.

23. An apparatus as claimed in claim 20 wherein there are multiple vibration sensors.

24. An apparatus as claimed in claim 20 wherein the spectrum generator is operable to calculate the auto spectrum of detected vibration and characteristic operational electrical signals and the cross spectrum of detected vibration and characteristic operational electrical signals.

25. An apparatus as claimed in claim 24 wherein the processing unit is operable to monitor the calculated auto and cross spectra over time.

26. An apparatus as claimed in claim 20 wherein the apparatus is operable to generate a force frequency response function from the vibration frequency response function; and compare the generated force frequency response function to a model force frequency response function.

27. An item of electrical equipment monitored according to the method of claim 1.

28. A system comprising a plurality of items of electrical equipment according to claim 27.

29. An item of electrical equipment comprising an apparatus according to claim 20.

30. A system comprising a plurality of items of electrical equipment according to claim 29.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0042] In order that the invention may be more clearly understood one or more embodiments thereof will now be described, by way of example only, with reference to the accompanying drawings, of which:

[0043] FIG. 1 is a schematic block diagram of an apparatus for monitoring the condition of electrical apparatus according to the present invention;

[0044] FIG. 2a illustrates a detected characteristic operational electrical signal of a DC electrical motor in operation;

[0045] FIG. 2b illustrates a detected vibration signal corresponding to the motor operation in FIG. 2a;

[0046] FIG. 3a is a frequency spectrum generated from the detected characteristic operational electrical signal of FIG. 2a;

[0047] FIG. 3b is a frequency spectrum generated from the detected vibration signal of FIG. 2b;

[0048] FIG. 4a illustrates acceleration frequency response spectra generated using three different input voltage levels;

[0049] FIG. 4b illustrates a model acceleration frequency response function generated by combining the frequency response spectra of FIG. 4a;

[0050] FIG. 4c illustrates an acceleration autospectrum as used for analysis in prior art techniques;

[0051] FIG. 5a illustrates blocked force frequency response spectra determined from the corresponding acceleration frequency response spectra of FIG. 4a;

[0052] FIG. 5b illustrates a model blocked force frequency response function generated by combining the frequency response spectra of FIG. 5a;

[0053] FIG. 6 shows two alternative implementations of the invention in respect of monitoring the condition of a fan;

[0054] FIG. 7a illustrates vibration frequency spectrums generated in the method of the present invention in respect of the fan of FIG. 6a;

[0055] FIG. 7b illustrates frequency response functions generated in the method of the present invention in respect of the fan of FIG. 6a;

[0056] FIG. 8 illustrates a comparison between frequency response functions for healthy and faulty fans of FIG. 6a; and

[0057] FIG. 9 illustrates a comparison between frequency response functions for healthy and faulty fans of FIG. 6b.

[0058] Turning now to FIG. 1 there is shown an apparatus 10 for monitoring the condition of an item of electrical equipment 1 whilst in operation. Typically, the electrical equipment 1 is a motor, generator or the like and the electrical equipment 1 is typically connected to one or more mechanical components.

[0059] The apparatus 10 comprises a vibration sensor 11. Depending on the circumstances, the vibration sensor 11 can comprise an accelerometer, piezoelectric sensor, pressure sensor, Hall effect sensor, microphone, optical fibre grating, strain gauge or the like. The vibration sensor 11 is operable to detect vibration of the equipment 1 during use and output a signal indicative thereof. Whilst the example of FIG. 1 shows a single vibration sensor 11, the skilled man will appreciate that multiple vibration sensors 11 can be provided if necessary.

[0060] The apparatus 10 also comprises an electrical sensor 12. The electrical sensor is operable to detect a characteristic operational electrical signal of the equipment 1. The characteristic operational electrical signal might be an input signal for power consuming equipment such as a motor and an output signal for a generator. The electrical sensor 12 can be operable to detect a single attribute of the characteristic operational electrical signal (voltage, current, power) or multiple attributes. As such, the electrical sensor 12 may comprise a voltmeter, ammeter or power meter as required. Whilst the simple example of FIG. 1 shows a single electrical sensor 12, the skilled man will appreciate that multiple electrical sensors 12 can be provided if necessary.

[0061] Turning now to FIG. 2, illustrative examples of both a detected characteristic operational electrical signal (FIG. 2a) and a detected vibration signal (FIG. 2b) are shown. The characteristic operation electrical signal is the electrical input signal for a DC motor operating at a constant speed. Whilst the DC motor may be set to operate at a particular voltage input (such as 10V illustrated in FIG. 2a), variation in coil position and brush operation during operation generates a measurable ripple on the base level of the input DC signal, with a period inversely related to the operation speed of the motor. The corresponding vibration signal resulting from operation of the electric motor according to the operating conditions of FIG. 2a is illustrated at FIG. 2b. The output of the vibration sensor 11 and the electrical sensor 12 is supplied to a spectrum generator 13. The spectrum generator 13 is operable to generate a frequency spectrum of the detected vibration and a frequency spectrum of the detected characteristic operational electrical signal. Whilst the example in FIG. 1 shows a single spectrum generator 13, in alternative embodiments, dedicated spectrum generators 13 may be provided for each sensor 11, 12. Turning now to FIGS. 3a and 3b, illustrations of frequency spectra generated from the time domain signals of FIGS. 2a and 2b respectively are shown. Peaks at the operational frequency of the motor (and harmonics of the operational frequency can be observed in the respective frequency spectra.

[0062] The output of the spectrum generator 13 is passed to a processing unit 14. The processing unit 14 is operable to process the respective frequency spectrums to generate a frequency response function. The processing unit 14 can generate the frequency response spectrums by use of a multichannel FFT (fast Fourier transform) to generate auto spectrums and cross spectrums as required or by direct calculation from the complex Fourier spectra from which the auto and cross spectra are derived.

[0063] Once a frequency response function is generated, the processing unit 14 is operable to compare the generated frequency response function to a model frequency response function. This allows any variations between the generated frequency response function and the model frequency response function to be identified. The occurrence of such variations provides an indication that the equipment 1 is not functioning according to the model frequency response function. This could be indicative of a fault. Indeed, in some instances, the nature of the variation could provide an identification of the nature of the fault.

[0064] Additionally, the auto spectrum and cross spectra generated may be used to calculate the coherence of the vibration and characteristic operational electrical signals. Where the coherence is equal to 1, there is a constant relationship between input and output whereas if the coherence drops to less than 1, then the relationship between input and output has changed, which can indicate the development of a fault. Where there are multiple vibration sensors 11 and/or multiple electrical sensors 12, the auto and cross spectra may be additionally utilised for virtual coherence analysis, operational transfer path analysis or principal component analysis.

[0065] Where variations from the model frequency response function are identified, the processing unit 14 may provide an output signal indicative of the variations. Optionally, the output signal can result in an indication being output on an output interface 15. The output interface can include means for visual output (such as one or more indicator lamps or a display screen) and means for audio output such as a buzzer, bell, alarm or loudspeaker). Additionally, the output interface may have means for communicating a notification to a remote device or means for printing a local notification. In such embodiments, the output signal may provide details of the nature of the fault (if identified) and/or instructions to perform maintenance.

[0066] In some embodiments, where a sufficiently serious fault is identified by comparison of the generated frequency response function to the model frequency response function, the processing unit 14 may be operable to output a command signal to the equipment 1. The command signal may cause operation of the equipment to be shut down. This can forestall potential danger or damage associated with continued operation of faulty equipment.

[0067] Optionally, as is shown in FIG. 1, the apparatus may incorporate a data store 16 or be connected to a remote data store 16. The data store can store the model frequency response function. The data store can also maintain a store of generated frequency response functions. This can allow an audit of monitoring activity to take place from time to time. Beneficially, this may help identify early indications of potential future faults.

[0068] The model frequency response function can be generated by modelling the expected frequency response of the equipment or by way of a calibration process. In the calibration process, equipment that is understood to be in good operating condition may be operated over the full range of expected operating conditions. During this operation, the vibration sensor 11 and electrical sensor 12 are operable to detect vibration and the characteristic operational electrical signal. The detected signals are processed by the spectrum generator 13 and processing unit 14 to generate frequency response functions. The frequency response functions so generated are stored as model frequency response functions. Where appropriate, a single model frequency response function that describes all operating conditions is generated. Where this is not possible, a series of model frequency response functions may be generated.

[0069] This is illustrated in FIG. 4. In FIG. 4a, acceleration frequency response spectra generated using three different input voltage levels are shown. In FIG. 4b, a model frequency response function generated by combining the frequency response spectra of FIG. 4a is illustrated. The skilled man will understand that in practice frequency response spectra from more different input signals may be utilised, as required or as desired.

[0070] FIG. 4c shows the acceleration autospectrum from which the acceleration frequency response function and master curve of FIG. 4a and FIG. 4b are derived. The autospectrum of FIG. 4c is what might be analysed in a conventional monitoring technique. As can readily be seen from comparison of the elements of FIG. 4, the present invention both reduces the dynamic range and simplifies the spectrum structure compared to conventional techniques. This can improve the accuracy and ease of fault identification.

[0071] Where the calibration process is used, the calibration can be undertaken at the completion of manufacture or installation of the equipment. In some cases, the calibration process can be repeated periodically, for instance after planned servicing. This can enable evolution in the frequency response of the equipment 1 due to use to be taken into account in monitoring.

[0072] In some embodiments, the calibration can also involve measuring the accelerance A of the equipment which is essentially the structural frequency response function describing the acceleration of a structure due to a unit force input. Where the accelerance A is known, and the vibration sensor 11 measures acceleration, the present invention can be used to determine the blocked force f.sub.bl at the location of vibration sensor 11. The blocked force f.sub.bl may be calculated from

f.sub.bl=A.sup.−1{dot over (a)}

where {dot over (a)} is the acceleration at the measurement point and F is a unit force input. As such, the accelerance A can be used to remove the influence of resonances from the model frequency response function. This is illustrated in FIG. 5, where blocked force frequency response functions were determined from the corresponding acceleration frequency response functions of FIG. 4. The resultant functions of FIG. 5a generated from the different input signals and the resultant model function of FIG. 5b are simplified by the removal of structural resonances. This can thereby make it easier to identify faults in general and/or to classify identified faults.

[0073] In one specific example of the implementation of the invention, as illustrated schematically in FIG. 6a, the equipment 1 comprises a rotary fan driven by an electric motor and the vibration sensor 11 comprises a sound pressure sensor spaced at 1 m from the fan 1 in a semi anechoic room. FIG. 7a shows a series of one-third octave band frequency spectrums 20 generated by the spectrum generator 13 for vibration detected by sound pressure sensor 11 for a healthy fan 1 running at different speeds determined by the identified DC input voltages. The sound pressure level measured varies widely demonstrating that a broad range of sound pressure levels are observable for healthy items of electrical equipment.

[0074] In the present invention, the spectrum generator 13 is also operable to generate a frequency spectrum of the electrical power supplied to fan 1 using an electrical sensor 12 connected to the input power supply. The processing unit 14 is then operable to generate both auto spectra and cross spectra of the detected vibration with respect to input voltage, current or power supplied to the fan 1. The auto and cross spectra are then used to determine frequency response functions 30 relating the vibration to the electrical input as shown in FIG. 7b. In this particular example, the electrical sensor 12 has detected electrical power supplied to the fan 1 and the detected electrical power has been used to generate a frequency response functions 30 relating input electrical power to vibration for the same range of input voltages as used in FIG. 3a. As can be seen for a healthy fan there is relatively little variation in the frequency response function 30 over the full operational range of the fan. As such, it is possible to represent healthy operation of the fan 1 using a single model frequency response function. Nevertheless, the skilled man will appreciate that if greater accuracy is required, a set of model frequency response functions may be used instead of a single model frequency response function.

[0075] Turning now to FIG. 8, a comparison is shown between the model frequency response function 31 of a healthy fan 1 and the generated frequency response function 32 of a fan 1 with a faulty blade as measured by sound pressure sensor 12. The difference between the two frequency response functions 31, 32 can be clearly identified by the increase in the magnitude of the generated frequency response function 32 in the frequency range 600-5000 Hz. As a result of identification of this difference: a warning alarm may be output, a maintenance notification generated and/or operation of the fan can be shutdown as required.

[0076] To better identify faults or to more accurately detect particular faults, it may be necessary to use different types of vibration sensor 11 or mount vibration sensors 11 at different points on or near the equipment 1. For example, as shown in FIG. 2b, the vibration sensor 11 could comprise an accelerometer fitted to the housing of an equivalent fan to that of FIG. 6a. Once again, a model frequency response function 31 can be generated for the fan 1. In the case that the fan of FIG. 6b has an imbalance fault, the model and generated frequency response functions 31, 32 related to the accelerometer output are shown in FIG. 9. The fault condition can be identified in this case by an increase in the magnitude of the generated frequency response function 32 in the frequency range 20-60 Hz.

[0077] The skilled man will therefore appreciate that the particular type, placement and number of vibration sensors 11 (and electrical sensors 12) will be determined by the nature of the equipment and the nature of the expected faults.

[0078] The above embodiment is described by way of example only. Many variations are possible without departing from the scope of the invention as defined in the appended claims.