Method and apparatus for isolating and identifying periodic Doppler signals in a turbine

• US 5,471,880 A
• Filed: 04/28/1994
• Issued: 12/05/1995
• Est. Priority Date: 04/28/1994
• Status: Expired due to Fees

• Current Assignee: Electric Power Research Institute
• Original Assignee: Electric Power Research Institute
• Inventors: Leon, Robert L., Lang, George F.
• Primary Examiner(s): Raevis, Robert
• Assistant Examiner(s): MILLER, ROSE MARY
• Application Number:
  US08/234,192
• Time in Patent Office: 586 Days
• Field of Search: 73/579, 73/593, 73/660, 73/658, 73/659, 364/508
• US Class Current: 73/660
• CPC Class Codes: G01H 1/006 of the rotor of turbo machines
• CPC Class Codes: G01H 3/08 Analysing frequencies present in complex vibrations, e.g. comparing harmonics present acoustic presence detection G01V1/001
• CPC Class Codes: G01N 2291/02872 Pressure
• CPC Class Codes: G01N 29/14 using acoustic emission techniques echo of particles G01N29/046; measuring mechanical vibrations or acoustic waves in solids in general G01H1/00

Abstract

An improved method and apparatus for detecting and identifying one or more resonantly vibrating blades of a turbine through the use of acoustic sensors imbedded in the stationary casing of the turbine is disclosed. The acoustic sensor signals are processed to separate the characteristic Doppler waveform from the accompanying random noise and periodic noise. Improved signal processing techniques which involve the evaluation of both an information-bearing resonant signal and its Hilbert transform via direct synthesis are disclosed.

Claims

1. An apparatus for detecting a resonantly vibrating blade in a row of blades secured to a shaft rotating about a rotor axis, in an operating turbine, comprising:

an acoustic sensor positioned on a stationary member with respect to said shaft to receive sound waves emanating from said rotating blades as said blades rotate about said rotor axis and to provide an output signal representative said received sound waves, said acoustic sensor positioned with respect to said shaft so that said vibrating blade approaches and departs from said acoustic sensor in the course of one rotation of said shaft about said rotor axis;

reference means for obtaining a reference signal indicative of shaft position at least once each time said shaft completes a revolution about said rotor axis and for determining shaft speed;

sampling means for sampling said sensor output signal to obtain samples related to shaft position as said shaft completes a multiplicity of revolutions at a plurality of different shaft speeds;

first memory means for temporarily storing the shaft speed from the reference means and a corresponding plurality of samples synchronized with the shaft position obtained by said sampling means, said plurality of samples together comprising a sample series representing at least one shaft rotation;

ensemble averager means for averaging multiple sample series to develop a plurality of synchronous ensemble averages differentiated by shaft speed;

second memory means for storing the plurality of synchronous ensemble averages differentiated by shaft speed, and an accompanying shaft speed histogram;

Fourier decomposition means for generating a plurality of complex order-normalized spectra differentiated by shaft speed from the synchronous ensemble averages stored in the second memory means, each complex order-normalized spectrum having a real component and an imaginary component;

curve fitter means for curve fitting the plurality of complex order-normalized spectra using the shaft speed histogram as a weighting function to generate a series of generally linear shaft speed functions fit to the real and imaginary Fourier coefficients of the complex order-normalized spectra, wherein for a linear curve fit, the Fourier slope coefficients represent a characteristic Doppler signal and the Fourier intercept coefficients represent periodic background noise; and

synthesis means for constructing a resonant signal characterizing the vibration of the resonantly vibrating blade from the generally linear shaft speed functions generated by the curve fitter means, an instantaneous amplitude thereof being used to identify the resonantly vibrating blade in the monitored row of blades.

2. The apparatus of claim 1 wherein the first memory means temporarily stores sixty-four time spaced samples reflecting acoustic pressure variations during a single most recent revolution of the shaft.

3. The apparatus of claim 1 wherein the sensor output signal is processed in real-time and every shaft revolution is characterized by a period.

4. The apparatus of claim 1 wherein the curve fitter means fits a linear function to each of the real components of the complex order-normalized spectrum in accordance with the equation:

space="preserve" listing-type="equation">A.sub.in =Ω

.sub.n i+ω

.sub.n

where

A_in =a cosine coefficient for the nth order term of the ith speed average;

Ω

_n =a slope of a best fit straight line through the A_in data points;

ω

_n =an intercept of the straight line;

i=an index to the speed segregation, wherein (0≦

i≦

I), I=a number of shaft speed categories segregated during data acquisition; and

n=a harmonic order number (1≦

n≦

N), N=the maximum harmonic order decomposed.

5. The apparatus of claim 4 wherein the curve fitter means substantially simultaneously fits a linear function to each of the imaginary components of the complex order-normalized spectrum in accordance with the equation:

space="preserve" listing-type="equation">B.sub.in =Δ

.sub.n i+δ

.sub.n

where

B_in =a sine coefficient for the nth order term of the ith speed average;

Δ

_n =a slope of a "best fit" straight line through the B_in data points;

δ

_n =an intercept of the straight line;

i=an index to the speed segregation, wherein (0≦

i≦

I), I=a number of shaft speed categories segregated during data acquisition; and

n=a harmonic order number, wherein (1≦

n≦

N), N=the maximum harmonic order decomposed.

6. The apparatus of claim 1 wherein the synthesis means constructs the resonant signal from the Fourier slope coefficients and forms a Hilbert transform of the resonant signal via direct synthesis.

7. The apparatus of claim 6 wherein the synthesis means constructs the resonant signal from the Fourier slope coefficients via direct synthesis by multiplying the real slope coefficients by sampled unit amplitude cosine waves of a corresponding harmonic order and multiplying the imaginary slope coefficients by sampled unit amplitude sine waves of a corresponding harmonic order and summing all of the resulting products at each sample point.

8. The apparatus of claim 7 wherein the Hilbert transform of the resonant signal is formed by direct synthesis substantially simultaneously using the same Fourier slope coefficients, wherein the Fourier slope coefficients are arranged to produce a 90°

phase shift in the frequency domain.

9. The apparatus of claim 7 wherein an arbitrary phase shift is imposed upon the synthesized signal to change a temporal alignment of the synthesized signal to align data from more than one acoustic pressure sensor means located at different azimuths relative to the shaft, the synthesis of the resonant signal defined in accordance with the equation:

##EQU10## and the synthesis of the Hilbert transform of the resonant signal defined in accordance with the equation;

##EQU11## where R(b)=the resonant signal;

H{R(b)}=the Hilbert transform of the resonant signal;

B=the number of blades in a row of blades;

b=a blade number (1≦

b≦

B);

d=an azimuth angle between the sensors;

n=a harmonic order number (1≦

n≦

N);

N=a maximum harmonic order decomposed;

Δ

_n =slope of a best fit straight line through the B_in data points, B_in =a sine coefficient for the nth order term of the of the ith speed average; and

Ω

_n =slope of a best fit straight line through the A_in data points, A_in =a cosine coefficient for the nth order term of the ith speed average.

10. The apparatus of claim 9 wherein the azimuth angle between the sensors (d) is negative to shift a trace from a sensor located to receive a later blade arrival.

11. The apparatus of claim 9 wherein the azimuth angle between the sensors (d) is positive to shift a trace from a sensor located to receive an earlier blade arrival.

12. An apparatus for detecting a resonantly vibrating blade in a row of blades secured to a shaft rotating about a rotor axis from data acquired at more than two different operating shaft rotational speeds, in an operating turbine, comprising:

an acoustic sensor positioned on a stationary member with respect to said shaft to receive sound waves emanating from said rotating blades as said blades rotate about said rotor axis and to provide an output representative of said received sound waves, said acoustic sensor positioned with respect to said shaft so that said vibrating blade approaches and departs from said acoustic sensor in the course of one rotation of said shaft about said rotor axis;

reference means for obtaining a reference signal indicative of shaft position at least once each time said shaft completes a revolution about said rotor axis and for determining shaft speed;

a sample control circuit for establishing a rate for sampling the sensor output signal over a multiplicity of shaft revolutions;

first memory means for temporarily storing a plurality of samples of said sensor output signal the data acquisition means, said plurality of samples together comprising a time history per blade revolution;

Fourier decomposition means for generating a multiplicity of complex order-normalized spectra from the time history, each complex spectrum having a real component and an imaginary component, the spectrum reflecting the spectral content of a single most recently completed shaft revolution;

first ensemble averager means for forming and accumulating an average complex order-normalized spectrum from the multiplicity of complex order-normalized spectra;

an address deviation look-up table cooperating with the reference means for generating an instantaneous speed deviation;

multiplier means for multiplying the complex order-normalized spectrum associated with the most recent revolution and the associated instantaneous speed deviation to generate a multiplicity of speed weighted complex order-normalized spectra;

second ensemble averager means for forming and accumulating an average speed weighted complex order-normalized spectrum from the multiplicity of speed weighted complex order-normalized spectra;

a deviation statistics means for averaging instantaneous shaft speed deviations to generate deviation statistics;

curve fitter means for simultaneously fitting a linear function of the shaft speed deviation to the real and the imaginary components of the averaged complex order-normalized spectrum and the averaged speed weighted complex order-normalized spectrum using the deviation statistics, to generate a series of generally linear shaft speed functions and;

synthesis means for constructing a resonant signal characterizing the vibration of the resonantly vibrating blade from the generally linear shaft speed functions generated by the curve fitter means, an instantaneous amplitude thereof being used to identify the resonantly vibrating blade in the monitored row of blades.

13. The apparatus of claim 12 wherein the deviation statistics comprise a mean value and a mean square value of the shaft speed deviations.

14. The apparatus of claim 12 wherein the reference means comprises a tachometer and a period detector, the period detector for measuring time intervals between successive tachometer pulses, wherein the first sample point is synchronized with the tachometer.

15. The apparatus of claim 12 wherein the Fourier decomposition means comprises a co-location curve fitter utilizing the shaft rotational speed per shaft revolution to fit a harmonic series of sine and cosine waves plus a mean value to the time history data to generate the complex order-normalized spectrum.

16. The apparatus of claim 12 wherein the first ensemble averager means forms and accumulates the average real and the average imaginary components of the complex order-normalized spectra in accordance with the equations:

##EQU12## A_n =average real component value of the nth harmonic of shaft speed;

B_n =average imaginary component value of the nth harmonic of shaft speed;

A_jn =real component value of the nth harmonic during the jth revolution;

B_jn =imaginary component value of the nth harmonic during the jth revolution;

n=a harmonic order number (1≦

n≦

N), N=a maximum harmonic order decomposed;

j=a revolution index (1≦

j≦

J); and

J=number of revolutions averaged.

17. The apparatus of claim 12 wherein the second ensemble averager means forms and accumulates the average speed weighted real and the average speed weighted imaginary components of the speed weighted complex order-normalized spectra in accordance with the equations:

##EQU13## α

_n =average speed weighted real component of the nth harmonic of shaft speed;

β

_n =average speed weighted imaginary component of the nth harmonic of shaft speed;

S_j =speed deviation during the jth revolution;

A_jn =real component value of the nth harmonic during the jth revolution;

B_jn =imaginary component value of the nth harmonic during the jth revolution;

n=a harmonic order number (1≦

n≦

N), N=a maximum harmonic order decomposed;

j=a revolution index (1≦

j≦

J); and

J=number of revolutions averaged.

18. The apparatus of claim 12 wherein the sensor output signal is processed in real time such that every revolution of the shaft contributes to the averages formed by the first and second ensemble averager means.

19. The apparatus of claim 12 wherein the sample rate is a fixed number per revolution.

20. The apparatus of claim 19 wherein the fixed number per revolution is sixty-four.

21. The apparatus of claim 19 wherein the Fourier decomposition means performs a simple unweighted fast Fourier transform.

22. The apparatus of claim 19 wherein the sample control circuit comprises a shaft encoder to fix the sample number per shaft revolution.

23. The apparatus of claim 19 wherein the sample control circuit comprises a phase locked loop having a phase assertion means for fixing the sample number per shaft revolution.

24. An apparatus for detecting a resonantly vibrating blade in a row of blades secured to a shaft rotating about a rotor axis in an operating turbine, from data acquired at more than two different operating shaft rotational speeds, to identify the vibrating blade, comprising:

an acoustic sensor positioned on a stationary member with respect to said shaft to receive sound waves emanating from said rotating blades as said blades rotate about said rotor axis and to provide an output signal representative of said received sound waves, said acoustic sensor positioned with respect to said shaft so that said vibrating blade approaches and departs from said acoustic sensor in the course of one rotation of said shaft about said rotor axis;

reference means for obtaining a reference signal indicative of shaft position at least once each time said shaft completes a revolution about said rotor axis and for determining shaft speed;

a sample control circuit for establishing a rate for sampling the output signal sensor over a multiplicity of shaft revolutions;

first memory means for temporarily storing a plurality of samples of the sensor output signal, said plurality of samples together comprising an instantaneous angle history;

a period detector circuit cooperating with a shaft speed deviation look-up table for generating an instantaneous speed deviation;

a deviation statistics module for averaging instantaneous speed deviations to generate deviation statistics;

first ensemble averager means for generating an average angle history from the instantaneous angle history;

multiplier means for multiplying the instantaneous speed deviation and the instantaneous angle history to generate a speed weighted angle history;

second ensemble averager means for generating an average speed weighted angle history from the speed weighted angle history;

first Fourier decomposition means for generating an average complex order-normalized spectrum from the average angle history;

second Fourier decomposition means for generating an average speed weighted complex order-normalized spectrum from the average speed weighted angle history;

curve fitter means for simultaneously fitting a linear function of the shaft speed deviation to the real and the imaginary components of the averaged complex order-normalized spectrum and the averaged speed weighted complex order-normalized spectrum using the deviation statistics, to generate a series of generally linear shaft speed functions;

synthesis means for constructing a resonant signal characterizing the vibration of the resonantly vibrating blade from the generally linear shaft speed functions generated by the curve fitter means, an instantaneous amplitude thereof being used to identify the resonantly vibrating blade in the monitored row of blades.

25. The apparatus of claim 24 wherein the sample control circuit generates a sample rate which is a fixed number per shaft revolution.

26. The apparatus of claim 25 wherein the sample rate is proportional to the instantaneous rotational speed of the shaft.

29. The apparatus of claim 24 wherein the first and the second Fourier decomposition means each perform a simple unweighted fast Fourier transform.

30. The apparatus of claim 24 wherein the acoustic sensor and the reference means operate in real time such that every revolution of the shaft contributes to the average angle history.

31. The apparatus of claim 25 wherein the fixed number is sixty-four.

32. The apparatus of claim 24 wherein the acoustic sensor includes an analog-to-digital convertor for converting the sensor output signal to digital signals to generate a shaft angle history comprising amplitude samples acquired as a sequence of evenly spaced shaft rotational angles.

33. The apparatus of claim 24 wherein the deviation statistics comprise a mean value and a mean square value of the shaft speed deviations.

34. The apparatus of claim 24 wherein the curve fitter means fits a linear function of the shaft speed deviation to the real and the imaginary components of the average complex order-normalized spectrum and the average speed weighted complex order-normalized spectrum in accordance with the equations:

space="preserve" listing-type="equation">A.sub.n (S)=Ω

.sub.n S+ω

.sub.n ;

and

space="preserve" listing-type="equation">B.sub.n (S)=Δ

.sub.n S+δ

.sub.n,

wherein

S=the instantaneous speed deviation of the shaft;

n=a harmonic order number (1≦

n≦

N), N=the maximum harmonic order decomposed;

A_n (S)=a cosine coefficient for the nth order term at speed deviation S;

Ω

_n =a slope of the linear A_n function;

ω

_n =an intercept of the linear A_n function;

B_n (S)=a sine coefficient for the nth order term at speed deviation S;

Δ

_n =a slope of the linear B_n function; and

δ

_n =an intercept of the linear B_n function.

35. The apparatus of claim 34 wherein the required Ω

_n and ω

_n slopes and the required Δ

_n and δ

_n intercepts are derived by least squares minimization.

27. The apparatus of 26 wherein the sample control circuit comprises a multiple pulse per revolution encoder affixed to the shaft.

28. The apparatus of 26 wherein the sample control circuit comprises a phase locked loop having a phase assertion means for fixing the sample number per shaft revolution.

36. A method of isolating from periodic background noise an acoustic signal characterizing the vibration of a resonantly vibrating blade in a row of turbine blades secured to a shaft rotating about a rotor axis from data acquired at more than two different operating shaft rotational speeds to identify the vibrating blade, comprising the steps of:

acquiring acoustic pressure wave data from pressure waves emanating from the rotating blades from one or more stationary sensors located proximate the turbine blades;

identifying the angular position of the shaft and the shaft speed at least once each time said shaft completes a revolution about said rotor axis;

sampling the pressure wave data acquired by the one or more sensor at a predetermined sample rate;

temporarily storing in a first memory the shaft speed and a corresponding plurality of samples of the pressure wave data synchronized with the shaft position, said plurality of samples together comprising a sample series representing at least one shaft rotation;

averaging multiple sample series to develop a plurality of synchronous ensemble averages differentiated by shaft speed;

storing in a second memory the plurality of synchronous ensemble averages differentiated by shaft speed and an accompanying shaft speed histogram;

performing a Fourier decomposition on the stored synchronous ensemble averages to generate a plurality of complex order-normalized spectra differentiated by shaft speed, each complex order-normalized spectrum having a real component and an imaginary component;

curve fitting the plurality of complex order-normalized spectra using the shaft speed histogram as a weighting function to generate a series of generally linear shaft speed functions fit to the real and imaginary Fourier coefficients of the complex order-normalized spectra, wherein the Fourier slope coefficients represent a characteristic Doppler signal and the Fourier intercept coefficients represent periodic background noise;

constructing a resonant signal characterizing the vibration of the resonantly vibrating turbine blade from the generally linear shaft speed functions; and

determining an instantaneous amplitude of the constructed resonant signal to identify the vibrating blade in the monitored row of blades.

37. A method of isolating from periodic background noise an resonant signal characterizing the vibration of a resonantly vibrating turbine blade in a row of turbine blades secured to a shaft rotating about a rotor axis from data acquired at more than two different operating shaft rotational speeds to identify the vibrating blade, comprising the steps of:

acquiring acoustic pressure wave data from pressure waves emanating from the rotating blades from one or more stationary sensors located proximate the turbine blades;

identifying the angular position of the shaft and the shaft speed at least once each time said shaft completes a revolution about said rotor axis;

sampling the pressure wave data at a predetermined sample rate;

temporarily storing in a first memory a plurality of samples of the data acquired, said plurality of samples together comprising a time history per blade revolution;

performing a Fourier decomposition on the time history data to generate a multiplicity of complex order-normalized spectra, each complex spectrum having a real component and an imaginary component, the spectrum reflecting the spectral content of a single most recently completed shaft revolution;

forming and accumulating an average complex order-normalized spectrum from the multiplicity of complex order-normalized spectra;

generating an instantaneous speed deviation from the angular position and the speed of the shaft using an address deviation look-up table;

multiplying the complex order-normalized spectrum associated with the most recent revolution and the associated instantaneous speed deviation to generate a multiplicity of speed weighted complex order-normalized spectra;

averaging the multiplicity of speed weighted complex order-normalized spectra to form and accumulate an average speed weighted complex order-normalized spectrum;

generating deviation statistics based on instantaneous shaft speed deviations;

curve fitting a linear function of the shaft speed deviation to the real and the imaginary components of the averaged complex order-normalized spectrum and the averaged speed weighted complex order-normalized spectrum using the deviation statistics, to generate a series of generally linear shaft speed functions;

constructing a resonant signal characterizing the vibration of the resonantly vibrating turbine blade from the generally linear shaft speed functions via Fourier-based synthesis; and

determining an instantaneous amplitude of the constructed resonant signal to identify the resonantly vibrating blade in the monitored row of blades.

38. A method of isolating from periodic background noise an resonant signal characterizing the vibration of a resonantly vibrating blade in a row of turbine blades secured to a shaft rotating about a rotor axis from data acquired at more than two different operating shaft rotational speeds to identify the vibrating blade, comprising the steps of:

acquiring acoustic pressure wave data from pressure waves emanating from the rotating blades from one or more sensors located proximate the turbine blades;

identifying the angular position of the shaft and the speed of the shaft at least once each time said shaft completes a revolution about said rotor axis;

sampling the pressure wave data acquired at a predetermined sample rate;

temporarily storing in a first memory a plurality of samples of the sampled data, said plurality of samples together comprising an instantaneous angle history;

generating an instantaneous speed deviation using a period detector circuit cooperating with a shaft speed deviation look-up table;

generating deviation statistics based on instantaneous shaft speed deviations;

generating an average angle history from the instantaneous angle history;

multiplying the instantaneous speed deviation and the instantaneous angle history to generate a speed weighted angle history;

generating an average speed weighted angle history from the speed weighted angle history;

generating an average complex order-normalized spectrum by performing a Fourier decomposition on the average angle history;

generating an average speed weighted complex order-normalized spectrum by performing a Fourier decomposition on the average speed weighted angle history;

fitting a linear function of the shaft speed deviation to the real and the imaginary components of the averaged complex order-normalized spectrum and the averaged speed weighted complex order-normalized spectrum substantially simultaneously using the deviation statistics to generate a series of generally linear shaft speed functions;

constructing a resonant signal characterizing the vibration of the resonantly vibrating turbine blade by performing Fourier-based synthesis on the linear shaft speed functions; and

determining an amplitude of the constructed resonant signal to identify the resonantly vibrating blade in the monitored row of blades.