CN104111108A - Torsional vibration impact signal characteristic extracting method for rotating mechanism - Google Patents

Abstract

The invention discloses a torsional vibration impact signal characteristic extracting method for a rotating mechanism. The torsional vibration impact signal characteristic extracting method for the rotating mechanism sees torsional vibration impact response signals as a group of attenuation response signals, indicates that the torsional vibration amplitude attenuation process can be simulated through an enveloping curve and presents that using Hilbert conversion to extract the enveloping characteristic of the amplitude attenuation process. Based on said proposal, the impact response energy is evaluated according to the area of the enveloping curve, and exponential functions are used to approach the enveloping curve. A logarithmic enveloping curve is obtained through taking the logarithm of the enveloping curve. An oblique line is used to approach the logarithmic enveloping curve, and the impact response signal damping characteristic and impact amplitude are calculated through the slope of the oblique line and a starting point value. The torsional vibration impact signal characteristic extracting method for the rotating mechanism fully uses the enveloping analysis ability of Hilbert conversion for transient signals, simplifies a Prony algorithm, and is easy and convenient to perform.

Description

A kind of rotating machinery torsional oscillation impact signal feature extracting method

Technical field

The present invention relates to a kind of characteristic of rotating machines vibration signal feature extracting method, can give prominence to the shock characteristic in reflection torsional vibration signals, while being particularly useful for the analysis of rotating machinery delivering polarization monitoring, torsional oscillation shock characteristic extracts demand, and then helps unit to carry out torsional state monitoring and fatigue life prediction.Main application fields comprises: the large rotating machineries such as power, metallurgy, petrochemical industry, aviation, and as compressor, generator, gas turbine, pump, blower fan, motor etc.

Background technology

Under the extraneous factor effects such as asynchronous parallelizing, Power System Disturbances, removal of load, reclosing, line switching operation, in Turbo-generator Set rotating shaft, can receive transient state torsion excitation, excited sub-synchronous oscillation, formed impact torsional oscillation, unit durability has been produced to very large harm.The energy of impact torsional vibration signals and feature are the important indicators that Turbo-generator Set torsional oscillation on-line monitoring and unit durability loss are evaluated.

During Turbogenerator Torsional Oscillation Analysis, Prony method and ITD method are two kinds of the most frequently used transient signal analytical approachs.Prony algorithm is a kind ofly can directly estimate according to sampled value the analytical approach of signal amplitude, frequency, decay factor, initial phase angle.This method hypothesis can be simulated (N >=2p) by the linear combination of p exponential function by the N of an equal interval sampling data point.For match value is approached to actual value, according to square error minimum principle, take least square method to try to achieve each coefficient value.During the actual use of this method, there are following several point defects: (1) algorithm is complicated; (2) observation window people need to be carried out to identification for being divided into some sections; (3) determining of model order affects larger on recognition result; (4) very sensitive to noise.ITD (Random Decrement) method regards environmental excitation as arbitrary excitation as, according to a plurality of random response calculated signals structure free damping responses that detect.This method thinks, the response of structure transient impact is comprised of determinacy part (pulse or step signal) and random partial.The response signal of testing is carried out after abundant sample mean, and the random partial of response is on average fallen, and residue is impulse response signal.This method can be extracted faint impulse response signal from signals and associated noises, but has following shortcoming: the dependency degree that (1) test result is chosen initial value is very large.Result of calculation possibility difference under different initial values is very large, has affected practicality and the reliability of the method; (2) need to have sufficiently long sample to do on average, the required sample size of a sample analysis is larger.

The present invention proposes a kind of new feature extraction of torsional oscillation impact signal and analytical approach.The present invention sees torsional oscillation impulse response signal as one group of convergent response signal, points out that torsional oscillation amplitude attenuation process can simulate with enveloping curve, proposes to convert to extract by Hilbert the envelope trait of amplitude attenuation process.On this basis, according to enveloping curve area, evaluate shock response energy, with exponential function, approach enveloping curve, with oblique line, approach the curve after the envelope signal value of taking the logarithm, by oblique line slope and threshold value, try to achieve impulse response signal damping characteristic and impact amplitude.This method takes full advantage of the Envelope Analysis ability of Hilbert transfer pair transient signal, has simplified Prony algorithm, and method is easy, feasible.

Summary of the invention

The present invention proposes a kind of new rotating machinery torsional oscillation impact signal feature extracting method.

The technical solution adopted for the present invention to solve the technical problems is:

A rotating machinery torsional oscillation impact signal feature extracting method, feature of the present invention is:

1) torsional oscillation impulse response signal is seen as to one group of convergent response signal, with Hilbert, converted to extract amplitude attenuation process

Enveloping curve;

2) according to enveloping curve area evaluation shock response energy;

3) ask for logarithmic envelope curve, with oblique line, approach logarithmic envelope curve, by oblique line slope and threshold value, try to achieve shock response

Signal damping characteristic and impact amplitude;

Its concrete steps are:

(1) measure original torsional vibration signals y (t), with 5 smoothing method noise reductions, obtain signal y to be analyzed ₁(t);

(2) signal y to be analyzed is asked in application Hilbert conversion ₁(t) envelope signal; First, calculate signal y to be analyzed ₁(t) Hilebrt figure signal y ₂(t)

$y_2\left(t\right)=\mathrm{hilbert}\left[y_1\left(t\right)\right]=\frac{1}{\mathrm{\π}}{\∫}_{-\∞}^{\∞}\frac{y_{1\left(t\right)}}{\mathrm{\τ}-t}\mathrm{d\τ};$

By original signal y ₁(t) the new analytic signal of signal configuration and after hilbert conversion:

y ₃(t)＝y ₁(t)+jy ₂(t)

The amplitude of analytic signal is exactly real signal y ₁(t) envelope signal y ₄(t):

$y_4\left(t\right)=\sqrt{y_1^2\left(t\right)+y_2^2\left(t\right)};$

(3) activation threshold value of setting impact signal is b, by the enveloping curve y in the time period to be analyzed ₄(t) be divided into some groups of impact event A _{i, i=1,2 ..., n}.One group of impact event A _ibe defined as follows:

Start point signal place: y ₄(t) <b, y ₄(t+ Δ t)>=b

Signaling destination point place: y ₄(t) >b, y ₄(t+ Δ t)≤b

Wherein, Δ t is sampling time interval;

(4) calculate one group of impact event A _ienvelope size s _i

$s_i=\underset{t1}{\overset{t2}{\&Integral;}}{y}_4\left(t\right)\mathrm{dt}$

Wherein, t ₁, t ₂for the corresponding starting point of impact event and the terminal moment;

(5) use decaying exponential function y ₅(t) approach enveloping curve y ₄(t)

y ₅(t)＝Ae ^-σt

Wherein, A is shock response amplitude, and σ is for the amount of reflection impact event decay speed degree, relevant with system damping;

(6) solve A and σ

Taken the logarithm in above formula equal sign both sides

y ₆(t _i)＝ln[y ₅(t)]＝ln[Ae ^-σt]＝ln(A)-σt

T=0 is signal y constantly ₆(t _i) functional value be ln (A), signal y ₆(t _i) slope be-σ.

The beneficial effect of the signal analysis method that compared with prior art, the present invention proposes is as follows:

(1) the present invention applies hilbert by the signal to after 5 smoothing and noise-reducing processes and converts, and and then extract torsional oscillation impact signal decay characteristics, algorithm is easy;

(2) the present invention proposes to solve the logarithm of discrete envelope signal, from the initial value of discrete logarithm envelope signal, asks for impact amplitude, from the slope value of discrete logarithm signal, asks for signal attenuation speed degree, has simplified parameter identification process;

(3) the present invention proposes to assess impact strength by envelope size, is a kind of new appraisal procedure.

Accompanying drawing explanation

Fig. 1 is the noisy torsional oscillation impulse response signal that emulation obtains;

Fig. 2 is the vibratory impulse response signal after level and smooth through 5;

Fig. 3 is the envelope signal being extracted by Hilbert conversion;

Fig. 4 asks for by envelope signal some groups of impact events that obtain;

Fig. 5 is the signal after the impact event envelope signal value of taking the logarithm;

Fig. 6 is with the impact event after the oblique line matching envelope signal value of taking the logarithm;

Fig. 7 is this method implementing procedure figure.

Embodiment

A rotating machinery torsional oscillation impact signal feature extracting method, feature of the present invention is:

1) torsional oscillation impulse response signal is seen as to one group of convergent response signal, with Hilbert, converted to extract amplitude attenuation process

Enveloping curve;

2) according to enveloping curve area evaluation shock response energy;

3) ask for logarithmic envelope curve, with oblique line, approach logarithmic envelope curve, by oblique line slope and threshold value, try to achieve shock response

Signal damping characteristic and impact amplitude;

Its concrete steps are:

(1) measure original torsional vibration signals y (t), with 5 smoothing method noise reductions, obtain signal y to be analyzed ₁(t);

(2) signal y to be analyzed is asked in application Hilbert conversion ₁(t) envelope signal; First, calculate signal y to be analyzed ₁(t) Hilebrt figure signal y ₂(t)

$y_2\left(t\right)=\mathrm{hilbert}\left[y_1\left(t\right)\right]=\frac{1}{\mathrm{\&pi;}}{\&Integral;}_{-\&infin;}^{\&infin;}\frac{y_{1\left(t\right)}}{\mathrm{\&tau;}-t}\mathrm{d\&tau;}$

By original signal y ₁(t) the new analytic signal of signal configuration and after hilbert conversion:

y ₃(t)＝y ₁(t)+jy ₂(t)

The amplitude of analytic signal is exactly real signal y ₁(t) envelope signal y ₄(t):

$y_4\left(t\right)=\sqrt{y_1^2\left(t\right)+y_2^2\left(t\right)};$

(3) activation threshold value of setting impact signal is b, and the enveloping curve in the time period to be analyzed is divided into some groups of impact event A _{i, i=1,2 ..., n}; One group of impact event A _ibe defined as follows:

Start point signal place: y ₄(t) <b, y ₄(t+ Δ t)>=b

Signaling destination point place: y ₄(t) >b, y ₄(t+ Δ t)≤b

Wherein, Δ t is sampling time interval.

(4) calculate one group of impact event A _ienvelope size s _i

$s_i=\underset{t1}{\overset{t2}{\&Integral;}}{y}_4\left(t\right)\mathrm{dt}$

Wherein, t ₁, t ₂for the corresponding starting point of impact event and the terminal moment;

(5) use decaying exponential function y ₅(t) approach enveloping curve y ₄(t)

y ₅(t)＝Ae ^-σt

Wherein, A is shock response amplitude, and σ is for the amount of reflection impact event decay speed degree, relevant with system damping;

(6) solve A and σ

Taken the logarithm in above formula equal sign both sides

y ₆(t _i)＝ln[y ₅(t)]＝ln[Ae ^-σt]＝ln(A)-σt

T=0 is signal y constantly ₆(t _i) functional value be ln (A), signal y ₆(t _i) slope be-σ.

The present embodiment is that to take certain emulation torsional oscillation impact signal be example, carries out extraction and the analysis of torsional oscillation impulse fault feature, with reference to Fig. 1～7.

(1) with sensor and analyser, gather torsional vibration signals y (t).In order to improve signal analysis precision and reliability, during signals collecting, advise: sample frequency >10 * best result is analysed frequency sampling time >5s;

(2) by 5 smoothing methods, reduce the noise component in signal y (t), obtain signal y to be analyzed ₁(t);

(3) signal y to be analyzed is asked in application Hilbert conversion ₁(t) Hilebrt figure signal y ₂(t)

$y_2\left(t\right)=\mathrm{hilbert}\left[y_1\left(t\right)\right]=\frac{1}{\mathrm{\&pi;}}{\&Integral;}_{-\&infin;}^{\&infin;}\frac{y_{1\left(t\right)}}{\mathrm{\&tau;}-t}\mathrm{d\&tau;};$

(4) by original signal y ₁(t) the signal y and after Hilbert conversion ₂(t) construct new analytic signal y ₃(t)

y ₃(t)＝y ₁(t)+jy ₂(t)；

(5) ask for envelope signal y ₄(t)

$y_4\left(t\right)=\sqrt{y_1^2\left(t\right)+y_2^2\left(t\right)};$

(6) activation threshold value of setting impact signal is b, and the enveloping curve in the time period to be analyzed is divided into some groups of impact event A _{i, i=1,2 ..., n}.One group of impact event A _ibe defined as follows:

Start point signal place: y ₄(t) <b, y ₄(t+ Δ t)>=b

Signaling destination point place: y ₄(t) >b, y ₄(t+ Δ t)≤b

Wherein, Δ t is sampling time interval;

(7) calculate one group of impact event A _ienvelope size s _i

$s_i=\underset{t1}{\overset{t2}{\&Integral;}}{y}_4\left(t\right)\mathrm{dt}$

Wherein, t ₁, t ₂starting time and terminal time for one-shot event;

(8) use decaying exponential function y ₅(t) approach enveloping curve y ₄(t)

y ₅(t)＝Ae ^-σt

Wherein, A is shock response amplitude, and σ is for the amount of reflection impact event decay speed degree, relevant with system damping; (9) solve logarithmic envelope curve y ₆(t _i)

y ₆(t _i)＝ln[y ₅(t)]；

(10) ask for A and σ

y ₆(t _i)＝ln[Ae ^-σt]＝ln(A)-σt；

As can be seen from the above equation, t=0 moment signal y ₆(t _i) value be ln (A), signal y ₆(t _i) slope be-σ.

Claims (1)

1. a rotating machinery torsional oscillation impact signal feature extracting method, is characterized in that:

1) torsional oscillation impulse response signal is seen as to one group of convergent response signal, with Hilbert, converted to extract amplitude attenuation process

Enveloping curve;

2) according to enveloping curve area evaluation shock response energy;

3) ask for logarithmic envelope curve, with oblique line, approach logarithmic envelope curve, by oblique line slope and threshold value, try to achieve shock response

Signal damping characteristic and impact amplitude;

Its concrete steps are:

(1) measure original torsional vibration signals y (t), with 5 smoothing method noise reductions, obtain signal y to be analyzed ₁(t);

(2) signal y to be analyzed is asked in application Hilbert conversion ₁(t) envelope signal; First, calculate signal y to be analyzed ₁(t) Hilebrt figure signal y ₂(t)

$y_2\left(t\right)=\mathrm{hilbert}\left[y_1\left(t\right)\right]=\frac{1}{\mathrm{\&pi;}}{\&Integral;}_{-\&infin;}^{\&infin;}\frac{y_{1\left(t\right)}}{\mathrm{\&tau;}-t}\mathrm{d\&tau;}$

By original signal y ₁(t) the new analytic signal of signal configuration and after Hilbert conversion:

y ₃(t)＝y ₁(t)+jy ₂(t)

The amplitude of analytic signal is exactly real signal y ₁(t) envelope signal y ₄(t):

$y_4\left(t\right)=\sqrt{y_1^2\left(t\right)+y_2^2\left(t\right)};$

(3) activation threshold value of setting impact signal is b, by the enveloping curve y in the time period to be analyzed ₄(t) be divided into some groups of impact event A _{i, i=1,2 ..., n}; One group of impact event A _ibe defined as follows:

Start point signal place: y ₄(t) <b, y ₄(t+ Δ t)>=b

Signaling destination point place: y ₄(t) >b, y ₄(t+ Δ t)≤b

Wherein, Δ t is sampling time interval;

(4) calculate one group of impact event A _ienvelope size s _i

$s_i=\underset{t1}{\overset{t2}{\&Integral;}}{y}_4\left(t\right)\mathrm{dt}$

Wherein, t ₁, t ₂for the corresponding starting point of impact event and the terminal moment;

(5) use decaying exponential function y ₅(t) approach enveloping curve y ₄(t)

y ₅(t)＝Ae ^-σt

Wherein, A is shock response amplitude, and σ is for the amount of reflection impact event decay speed degree, relevant with system damping;

(6) solve A and σ

Taken the logarithm in above formula equal sign both sides

y ₆(t _i)＝ln[y ₅(t)]＝ln[Ae ^-σt]＝ln(A)-σt

T=0 is signal y constantly ₆(t _i) functional value be ln (A), signal y ₆(t _i) slope be-σ.