CN1395085A - Extraction and monitoring method of half-speed vortex on-line stability characteristics of high-speed rotating machine - Google Patents

Abstract

The invention discloses the method for picking-up and monitoring the character of on-line stability of the half speed whirl for the on-the-fly machine. The electric whirl displacement sensor of non-contacted type obtains the vibration displacement signal of the rotator. The vibration displacement signal is preprocessed through the FIR filter technique so as to reserve the component near the half-frequency and filter out other interference component. Then, the RDT techniques pick-up the stability parameter of the system is obtained through the genetic algorithm. The trend map between the stability parameter and time can be drawn out.

Description

Feature extraction of high-speed rotating machine half-speed vortex on-line stability and monitoring method

One, affiliated field

The invention belongs to mechanical fault diagnosis and rotating machinery model analysis field, further can expand to the related application field that relates to signal analysis and feature extraction, feature extraction of particularly a kind of high-speed rotating machine half-speed vortex on-line stability and monitoring method.

Two, background technology

Half-speed vortex is the most common failure of high-speed rotating machine, and it is a kind of autovibration that is caused by the hydrodynamic force between rotating shaft and the hydrodynamic lubrication bearing, is a kind of performance of system's loss of stability (abbreviation unstability).Under certain condition, in case system's unstability, the rotor eddy amplitude can sharply increase, and makes unit produce judder, and having of vibration is sudden.In addition, since the half-speed vortex frequency be approximately the rotor gyro frequency half (be generally rotor change frequently 0.42 to 0.48 between), it is a kind of metasynchronism vibration, therefore in rotor, can produce alterante stress, than the synchronous vibration that is caused by reasons such as rotor imbalances bigger harmfulness being arranged, usually is the main cause that causes a serious accident.So high-speed rotating machine all will carry out Calculation on stability and checking computations in the design phase usually, even sets up 1: 1 experiment table stability is carried out modal test.Yet numerous examples show, because many factors all can influence stability, as oil temperature, oil pressure, bearing clearance and rotor balancing, centering situation etc., even the good unit of stability Design destabilization problems can occur equally after moving several months or several years.Therefore, the on-line monitoring of half-speed vortex is very important.

Existing half-speed vortex on-line monitoring method mainly contains: spectrum analysis, waterfall figure and shaft core position monitoring.Frequency spectrum analysis method is to realize according to occurring half vibration of composition frequently after the unstability, mainly is by half existence of composition frequently in the spectrum analysis monitors vibration signal in unit running process.If find to have in the vibration of rotor half composition frequently, promptly think the possibility that has half-speed vortex; Otherwise,, think that then system is in steady state (SS) if find half frequency composition.The major defect of this method is: have only when system is in instability status and could find, be difficult to stability status is carried out trend analysis the generation of unpredictable half-speed vortex.Because half-speed vortex has sudden and rotor is produced alterante stress, usually in the very short time that whirling motion occurs, unit is worked the mischief and destroy, therefore this method is obviously not ideal enough.Waterfall drawing method and spectrum analysis ratio juris are basic identical, also are to utilize spectrogram to monitor half frequency composition, and different is to carry out in unit raising speed or reduction of speed process.This method writes down the vibration signal of unit and tries to achieve spectrogram under different rotating speeds, the spectrogram that different rotating speeds is obtained down is stacked together by rotating speed then, form a frequency, rotating speed three-dimensional picture (being called waterfall figure), therefore can observe the stability status under the different rotating speeds with respect to amplitude.But this method exists equally to be had only after half-speed vortex takes place and could find, and the defective that can't carry out stable trend analysis and prediction.The shaft core position monitoring method is to be undertaken by the monitoring static equilibrium center of rotor motion and the relative position of bearing geometric center, the variation intermediate description of shaft core position change of stability.Can determine stabilized zone and the unstable region that shaft core position changes in advance by theoretical analysis and test method.When unit operation, if shaft core position falls into the stabilized zone, illustrative system is in steady state (SS); Otherwise,, then may be in instability status if shaft core position falls into unstable region.The advantage of this method is to carry out on-line analysis and prediction to stable trend and state, but it is very high that defective is that the installation of actual measurement precision, sensor to the bearing geometric parameter and debugging etc. require, be difficult in fact realize that we once attempted repeatedly all failing.In addition, the shaft core position monitoring is to calculate stability status by the variation of indirect monitoring shaft core position, is suitable for qualitative analysis, is difficult to provide quantitative results.

Except that above three kinds of monitoring methods, another kind of method for analyzing stability commonly used is that system damping is calculated and experiment, and this method is used the system damping (for example damping ratio) that obtains and estimated stability status.System damping can influence the combined influence of the various factors of stability to stability by quantitative response, in stability theory and experiment, use very general, its major advantage is to carry out engine sta bility margin and trend analysis quantitatively, is a kind of very good method for analyzing stability theoretically.But, this in the past method is to obtain stability parameter by Theoretical Calculation and mode experiment in the design of unit and manufacture process, and for the unit in actual motion, these stability parameters are difficult to obtain, therefore, with regard to prior art, this method can't be carried out stable on-line monitoring.

Three, summary of the invention

Defective or deficiency at prior art, the objective of the invention is to needs based on rotary machinery fault diagnosis and half-speed vortex stability on-line monitoring, feature extraction of a kind of high-speed rotating machine half-speed vortex on-line stability and monitoring method are provided, this method is extracted stability parameter from the rotating machinery vibrating displacement signal, thus realize to rotatory mechanical system engine sta bility margin and trend online, quantitatively monitor and predict.(1) Fa Ming design basis

1. theory and practice confirms: on average the go in ring frequency of speed of the frequency of half-speed vortex and fluid between rotating shaft and the hydrodynamic lubrication bearing equates, and the on average belt speed of fluid is a distinctive natural frequency of Rotor-Bearing System.Therefore, half-speed vortex and common system are similar, have characteristics such as damping, resonance equally.

Large rotating machine such as centrifugal compressor, turbine generators group in the course of the work, hydrodynamic force and steam-operating power can produce arbitrary excitation to rotor, can evoke the proper vibration of system, thereby have excited half-speed vortex.If system is in steady state (SS) (there are enough dampings in system), the very fast decay under the effect of system damping of whirling motion meeting, disappearance; If owing to temperature, rotating speed, oil pressure, bearing change (system damping at this moment can't suppress whirling motion) when making system surpass stable critical value to moderate condition, whirling motion is then kept with the form of resonance and is no longer disappeared.

The present invention is exactly under above-described two prerequisites, feature according to half-speed vortex, use signal processing technology and will represent that the parameter of system damping extracts as eigenwert, these characteristic parameters have been described the damping characteristic of system, thereby can be used as measuring of possibility size that the assessment half-speed vortex takes place.(2) Fa Ming realization principle

1. half-speed vortex is that vibration displacement by rotor shows, so the present invention's non-contact electric eddy shift sensor of using present main flow obtains the vibration displacement signal of rotor.Based on this signal, therefrom extract the stability parameter of half-speed vortex.

2. the external force that excites half-speed vortex is the arbitrary excitation that the hydrodynamic force that has of system itself and steam-operating power produce, so the very strong RDT technology (Random decrement technique) of employing practicality is come the damping characteristic (as damping ratio) of extraction system.The principle of this technology is based on carries out the multistage intercepting to the vibration displacement signal with a fixing value, then the multistage signal is asked population mean, and averaged result is the free damping response of system's proper vibration.For the present invention at half-speed vortex, this technology also can't be used.Reason is, in the rotating machinery vibrating displacement signal, exist powerful interference component usually, mainly be the determinacy speed-frequency that causes of the remaining out-of-balance force of rotor and the vibration of frequency multiplication thereof, these vibrations are very strong, and usually the response with arbitrary excitation overrides fully.On the other hand, the core of RDT technology when aforementioned determinacy interference is very serious, can not correctly be carried out the multistage intercepting of signal at arbitrary excitation, therefore also just can not obtain satisfied result.For solving this technical matters, the present invention uses the FIR wave filter technology, before implementing the RDT technology, the vibration displacement signal is carried out pre-service, and only near the composition the reservation half frequently all filters out other interference components.This measure is one of gordian technique of being achieved of the present invention.

3. after having obtained the free damping response, can obtain the damping characteristic parameter (damping ratio or logarithmic decrement) of system by genetic algorithm.In unit running process, at set intervals (as, one day, two weeks or one month) obtain stability parameter with said method, the stability parameter of these acquisitions is drawn out trend map over time, can realize stable development trend is estimated, thus the generation of prediction half-speed vortex.(3) working of an invention step, rule and algorithm＜1〉signals collecting

Fig. 1 is bearing in rotating machinery-rotor-support-foundation system synoptic diagram and sensor scheme of installation.Fig. 2 is the signals collecting process flow diagram.Wig is x (t) with eddy current displacement sensor measuring machine group bearing with respect to the vibration displacement signal of axle journal, by A/D conversion (sampling) x (t) is converted to discrete digital signal x _n, promptly

x _n=x (nT), n=0,1,2 ..., N-1 (1) wherein, T is a sampling interval, N is a sample length.For guaranteeing that the parameter of extracting has enough precision, sampling interval T and sample length N press following regular value: $\frac{40}{N\·f_r}\≤T\≤\frac{1}{10\·f_r}---\left(2\right)$ F in the formula _rBe the speed-frequency of machine group rotor, calculate by following formula $f_r=\frac{n_r}{60}\left(\mathrm{Hz}\right)---\left(3\right)$ N wherein _rBe the rotating speed of unit, unit is RPM (a commentaries on classics/per minute).The sample length N of sampling should be not less than 1024, promptly

The value of the common N in N 〉=1024 (4) is the bigger the better, and for example 2048,4096 or the like, can choose according to the capacity of sampling system.＜2〉digital filtering

With the signal x of Finite Impulse Response filter to gathering _n(n=0,1,2 ..., N-1) carry out bandpass filtering, eliminate the influence of interference component.If filtered signal is y _n, filtering is as follows: $y_n=\underset{m=-\mathrm{<figure-callout\; id="2"\; label="M\; M\; sin"\; filenames="A0211468900152.png"\; state="\{\{state\}\}">M</figure-callout>}}{\overset{M}{\mathrm{\&Sigma;}}}\frac{\sin }{}\frac{2\mathrm{\&pi;m}{f}_{h}T-\mathrm{<figure-callout\; id="2"\; label="sin"\; filenames="A0211468900152.png"\; state="\{\{state\}\}">sin</figure-callout>}2\mathrm{\&pi;m}{f}_1\mathrm{<figure-callout\; id="2"\; label="T"\; filenames="A0211468900152.png"\; state="\{\{state\}\}">T</figure-callout>}}{2\mathrm{m\&pi;}}[1+\cos (\mathrm{m\&pi;}/N)]\&CenterDot;x_{n-m}---\left(5\right)$ Wherein, f _lAnd f _hBe respectively the lower limit and the upper limiting frequency of bandpass filter passband.f _lAnd f _hSelection principle be with half frequently (half of generating unit speed frequency) be included between the two, the present invention's suggestion is by following value: (5) in the formula, M is a filter length, and M is big more usually, and precision is high more, but calculated amount is also big more.In our experience, M is not less than 40 and can have satisfied accuracy requirement.＜3〉extract signature waveform

By step＜2〉obtain system arbitrary excitation response y _n(n=0,1,2 ..., N-1), use the signature waveform (being the free damping response of system) that the RDT technology can extract the computational stability parameter.The RDT technology is with certain constant C _bFor reference value signal is carried out the multistage intercepting, then the multistage calculated signals population mean of intercepting is realized.Specific algorithm is as follows:

1. choose the reference value of signal being carried out the multistage intercepting

In theory can be with signal y _nRoot-mean-square value y _RMSAs benchmark, promptly $y_{\mathrm{RMS}}=\sqrt{\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{y}_n^2}---\left(7\right)$

2. revise reference value

Because y _nBe digital signal, its sample value can not by chance equal certain specific constant usually (as y _RMS), therefore should get y _RMSA near small codomain is as the benchmark of the sub-segment signal of intercepting, promptly $y_{\mathrm{RMS}}-\frac{\mathrm{\&Delta;<figure-callout\; id="2"\; label="C"\; filenames="A0211468900152.png"\; state="\{\{state\}\}">C</figure-callout>}}{2}\&le;C_b\&le;y_{\mathrm{RMS}}+\frac{\mathrm{\&Delta;C}}{2}---\left(8\right)$

For the Δ C in the following formula, the following formula value is pressed in the present invention's suggestion

ΔC＝2.2frTy _RMS (9)

In the formula, T is a sampling interval, f _rBe the generating unit speed frequency.

3. calculate the starting point of sub-segment signal

With signal y _n(n=0,1,2 ..., each sample value N-1) and reference value C _bCompare, with reference value C _bThe sample of signal point that equates is as the starting point of sub-segment signal.If obtain I starting point (obvious I≤N), establish i sub-segment signal again and be designated as altogether

, k=0,1,2 ..., K-1, K are the sample length of sub-segment signal, then the starting point of the I of Huo Deing sub-segment signal is $u_0^{\left(i\right)}=y_i{|}_{y_i=C_b},i=\mathrm{0,1,2},...,I-1---\left(10\right)$

4. intercept sub-segment signal

The starting point of I sub-segment signal obtaining is pressed signal y respectively _nOrder is got K sample value, promptly obtains M sub-segment signal, promptly $u_k^{\left(i\right)}=y_{i+k},k=\mathrm{0,1,2},...,K-1$

i＝0，1，2，...，I-1 (11)

5. for guaranteeing that the signature waveform that extracts has enough quantity of information, advise that the length K of sub-segment signal satisfies $K\&GreaterEqual;\frac{4}{f_{r}T}---\left(12\right)$

In the formula, T is a sampling interval, f _rBe the generating unit speed frequency.

6. calculate the population mean of each sub-segment signal

Calculate each sub-segment signal

Promptly obtain the signature waveform signal v that will extract with respect to i average _k, promptly $v_k=\frac{1}{I}\underset{i=0}{\overset{I-1}{\mathrm{\&Sigma;}}}{u}_k^{\left(i\right)},k=\mathrm{0,1,2},...,K-1---\left(13\right)$ ＜4〉computational stability parameter

System's free damping response can be expressed from the next

V (t)=Ae ^-utIn sin (ω t+ ) (14) formula, u (〉=0) is system damping, and ω is a natural frequency, and is a phase place, and A is a peak swing.By step＜3〉obtain signature waveform signal v _k(k=0,1,2 ..., K-1) afterwards, we can come match (14) formula and v with least square method _kThereby, try to achieve stability parameter.If objective function is $E=\underset{k=0}{\overset{K-1}{\mathrm{\&Sigma;}}}[v\left(\mathrm{kT}\right)-v_k{]}^2=\min ---\left(15\right)$

To above nonlinear optimal problem, the present invention recommends the genetic algorithm for solving with stable performance.Find the solution (15) formula, can obtain the estimated value of u, ω, and A, wherein can obtain the dampingratio of system by parameters u relevant and ω with system stability $\mathrm{\&zeta;}=\frac{u}{\sqrt{{\mathrm{\&omega;}}^2+u^2}}---\left(16\right)$

Dampingratio is an important parameter of estimating half-speed vortex stability.ζ is big more, and stability is good more; ζ is more little, and stability is poor more; When ζ was zero, system had been in instability status, is accompanied by strong half-speed vortex.＜5〉half-speed vortex STABILITY MONITORING

Rotation unit to operation regularly repeats above step＜1〉to step＜4 〉, for example once a day, weekly, first quarter moon once or January once, the damping ratio that at every turn obtains is drawn trend map in chronological order, can be quantitatively, intuitively half-speed vortex stability is carried out trend analysis, is estimated engine sta bility margin, and then realize supervision and prediction to half-speed vortex stability.Specifically referring to embodiment.

Repeated multiple times of the present invention is verified with field data, has obtained satisfied result (referring to embodiment) in large rotating machine group half-speed vortexes such as centrifugal compressor, turbodynamo stability on-line monitoring.

The breakthrough point compared with existing method (for example spectrum analysis, waterfall figure etc.) of invention is: development trend that can in-service monitoring half-speed vortex stability, estimate set steady nargin, thus the possibility of prediction half-speed vortex generation.Secondly, invention has quantitatively, effect, and the result who obtains intuitively has physics and simple, the clear and definite characteristics of engineering significance, is convenient to the field staff and grasps.In addition, working of an invention is simple and easy to do, and only needing to measure unit axle journal vibration displacement signal can realize, wide accommodation, and implementation cost is low.

Four, description of drawings

Fig. 1 is bearing in rotating machinery-rotor-support-foundation system synoptic diagram and sensor scheme of installation;

Fig. 2 is the signals collecting process flow diagram;

Fig. 3 is the vibration displacement signal of one of them sensor output;

Fig. 4 is the characteristic signal that extracts from signal shown in Figure 3;

Fig. 5 is half-speed vortex stability change trend map (damping ratio over time).

Five, embodiment

For a more clear understanding of the present invention, the present invention is described in further detail according to embodiment that technical scheme of the present invention has been done below in conjunction with accompanying drawing and inventor.

Embodiment:

With CO of certain petro-chemical corporation ₂Centrifugal compressor units is an example.This unit has been installed the current vortex sensor that U.S. Bently company produces, and has been equipped with 7200 series instruments, can obtain the vibration displacement signal of rotor journal easily from the instrument lead-out terminal.By being based upon line monitoring system, we have carried out the tracking and monitoring of several years to this unit.The half-speed vortex fault took place one time in unit during this time.Now be described below:

Fig. 3 is the vibration displacement signal of one of them sensor output.The rotating speed of unit is 12600 _RPM, can be regarded as to such an extent that speed-frequency is by (3) formula

f _r＝12600/60＝210(Hz) (17)

Set by step＜1〉choose sampling interval

T=1/3000 (18) and sample length

N=1024 (19) T and N satisfy the requirement of (2) formula and (4) formula.Signal is sampled, obtain discrete digital signal x _n(n=0,1,2 ..., N-1).

(6) formula selecting filter upper and lower bound frequency set by step＜3〉is Filter length is taken as 40.(5) formula of pressing is to signal x _nCarry out digital filtering, obtain signal y after the filtering _n(n=0,1,2 ..., N-1).

(7) formula signal calculated y set by step＜4 〉 _nRoot-mean-square value y _RMS, by the reference value C of (8) formula signal calculated multistage intercepting _b, wherein Δ C calculates by (9) formula, promptly

Δ C=2.2 * 210/3000 * y _RMS=0.154y _RMS(21) use reference value C _b(10) formula of pressing is obtained the starting point of sub-segment signal, and intercepts sub-segment signal by (11) formula, obtains $u_k^{\left(i\right)},k=\mathrm{0,1,2},...,K-1;i=\mathrm{0,1,2,3},...,I-1---\left(22\right)$

Wherein, get

K=60 (23) satisfies (12) formula.(13) formula of application is asked the population mean of each sub-segment signal, promptly obtains the signature waveform of free damping response, referring to Fig. 4.

Set by step＜5〉draw trend map, half-speed vortex is carried out in-service monitoring and prediction.Present embodiment, we provide the trend map (Fig. 5) that measuring point of above-mentioned unit is drawn in chronological order at the system damping ratio that obtains with the present invention every two weeks in two years, use the present invention so as to explanation and are rotated mechanical half-speed vortex on-line stability monitoring and analytical approach.In half-speed vortex shown in Figure 5 stability trend map, the dampingratio of system has the process that descends faster in 3 months, and damping ratio has dropped to approximately 0.1 from about 0.4 till 3rd month, illustrate that stability is in deterioration.In half a year subsequently (from 3rd month to 9th month), damping ratio is fluctuation in a narrow margin between 0.1 to 0.2, although damping ratio is less during this period, its variation is comparatively steady.Since 9th month, damping ratio had a comparatively faster decline process again, and system stability further worsens.Arrived about the 10th first quarter moon, damping ratio has dropped to and has approached zero (about 0.03), is indicating that half-speed vortex may take place at any time, therefore tackles unit and keeps in repair.But consider that the preplanned overhaul phase closes on, for keeping the production cycle of plan, company's decision unit continues operation, simultaneously unit is strengthened monitoring.In the time of 11 months, damping ratio equals zero at unit operation to the, and half-speed vortex takes place.Because the whirling motion amplitude is not very big, so unit adheres to moving about two weeks again.To youngster's first quarter moon, the equipment of whole company enters the overhaul phase according to plan, and this unit has also carried out the disintegration overhaul.After the overhaul, unit has recovered stable operation, between the 12nd first quarter moon to the 24 months, and damping ratio fluctuation between 0.3 to 0.4 always, system is in steady operational status, half-speed vortex do not occur.By present embodiment as seen, the half-speed vortex on-line stability monitoring method that provides of the present invention is effective.When the damping ratio that is obtained by the present invention reduces rapidly in time, should give special concern.Especially, when damping ratio approaches zero, indicating that then half-speed vortex may take place.

Claims (1)

1. high-speed rotating machine half-speed vortex on-line stability feature extraction and monitoring method is characterized in that, carry out by the following method:

1) non-contact electric eddy shift sensor of the present main flow of application obtains the vibration displacement signal of rotor; Based on this signal, use the stability parameter that the RDT technology is therefrom extracted half-speed vortex;

2) before implementing the RDT technology, use the FIR wave filter technology vibration displacement signal is carried out pre-service, only near the composition the reservation half frequently all filters out other interference components;

3) after having obtained the free damping response, can obtain the damping ratio of system or the damping characteristic parameter of logarithmic decrement by genetic algorithm; In unit running process, obtain stability parameter with said method at set intervals, the stability parameter of these acquisitions is drawn out trend map over time;

Concrete implementation step, rule and algorithm are as follows:＜1〉signals collecting

Suppose that be x (t) with eddy current displacement sensor measuring machine group bearing with respect to the vibration displacement signal of axle journal,, x (t) is converted to discrete digital signal x by A/D conversion/sampling _n, promptly

x _n＝x(nT)，n＝0，1，2，...，N-1 (1)

Wherein, T is a sampling interval, and N is a sample length;

For guaranteeing that the parameter of extracting has enough precision, sampling interval T and sample length N press following regular value: $\frac{40}{N\&CenterDot;f_r}\&le;T\&le;\frac{1}{10\&CenterDot;f_r}---\left(2\right)$

F in the formula _rBe the speed-frequency of machine group rotor, calculate by following formula $f_r=\frac{n_r}{60}\left(\mathrm{Hz}\right)---\left(3\right)$

N wherein _rBe the rotating speed of unit, unit is RPM (a commentaries on classics/per minute); The sample length N of sampling should be not less than 1024, promptly

N≥1024 (4)

Usually the value of N is the bigger the better, and chooses according to the capacity of sampling system;＜2〉digital filtering

With the signal x of Finite Impulse Response filter to gathering _n(n=0,1,2 ..., N-1) carry out bandpass filtering, eliminate the influence of interference component;

If filtered signal is y _n, filtering is as follows: $y_n=\underset{m=-M}{\overset{M}{\mathrm{\&Sigma;}}}\frac{\sin 2\mathrm{\&pi;m}{f}_{h}T-\sin 2\mathrm{\&pi;m}{f}_{l}T}{2\mathrm{m\&pi;}}[1+\cos (\mathrm{m\&pi;}/N)]\&CenterDot;x_{n-m}---\left(5\right)$

In the formula, M is a filter length, and M is big more usually, and precision is high more;

Wherein, f _lAnd f _hBe respectively the lower limit and the upper limiting frequency of bandpass filter passband, f _lAnd f _hSelection principle be will half frequently be included between the two, by following value:

＜3〉extract signature waveform

By step＜2〉obtain system arbitrary excitation response y _n(n=0,1,2 ..., N-1), use the signature waveform that the RDT technology can extract the computational stability parameter, i.e. the free damping of system response; The RDT technology is with certain constant C _bFor reference value signal is carried out the multistage intercepting, then the multistage calculated signals population mean of intercepting is realized; Specific algorithm is as follows:

1. choose the reference value of signal being carried out the multistage intercepting

In theory can be with signal y _nRoot-mean-square value y _RMSAs benchmark, promptly $y_{\mathrm{RMS}}=\sqrt{\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{y}_n^2}---\left(7\right)$

2. revise reference value

Because y _nBe digital signal, its sample value y _RMSUsually certain specific constant can not be by chance equaled, therefore y should be got _RMSA near small codomain is as the benchmark of the sub-segment signal of intercepting, promptly $y_{\mathrm{RMS}}-\frac{\mathrm{\&Delta;C}}{2}\&le;C_b\&le;y_{\mathrm{RMS}}+\frac{\mathrm{\&Delta;C}}{2}---\left(8\right)$ Δ C in the following formula presses the following formula value

Δ C=2.2frTy _RMS(9) in the formula, T is a sampling interval, f _rBe the generating unit speed frequency; 3. the starting point of calculating sub-segment signal is with signal y _n(n=0,1,2 ..., each sample value N-1) and reference value C _bCompare, with reference value C _bThe sample of signal point that equates is as the starting point of sub-segment signal.If obtain I starting point (obvious I≤N), establish i sub-segment signal again and be designated as altogether , k=0,1,2 ..., K-1, K are the sample length of sub-segment signal, then the starting point of the I of Huo Deing sub-segment signal is $u_0^{\left(i\right)}=y_i{|}_{y_i=C_h},i=\mathrm{0,1,2},...,I-1---\left(10\right)$

4. intercept sub-segment signal

The starting point of I sub-segment signal obtaining is pressed signal y respectively _nOrder is got K sample value, promptly obtains M sub-segment signal, promptly $u_k^{\left(i\right)}=y_{i+k},k=\mathrm{0,1,2},...,K-1$ i＝0，1，2，...，I-1 (11)

5. for guaranteeing that the signature waveform that extracts has enough quantity of information, advise that the length K of sub-segment signal satisfies $K\&GreaterEqual;\frac{4}{f_{r}T}---\left(12\right)$

In the formula, T is a sampling interval, f _rBe the generating unit speed frequency;

6. calculate the population mean of each sub-segment signal

Calculate each sub-segment signal Promptly obtain the signature waveform signal v that will extract with respect to i average _k, promptly $v_k=\frac{1}{I}\underset{i=0}{\overset{I-1}{\mathrm{\&Sigma;}}}{u}_k^{\left(i\right)},k=\mathrm{0,1,2},...,K-1---\left(13\right)$ ＜4〉computational stability parameter

System's free damping response can be expressed from the next

v(t)＝Ae ^-utsin(ωt+) (14)

In the formula, u (〉=0) is system damping, and ω is a natural frequency, and is a phase place, and A is a peak swing.By step＜3〉obtain signature waveform signal v _k(k=0,1,2 ..., K-1) afterwards, come match (14) formula and v with least square method _kThereby, try to achieve stability parameter; If objective function is $E=\underset{k=0}{\overset{K-1}{\mathrm{\&Sigma;}}}[v\left(\mathrm{kT}\right)-v_k{]}^2=\min ---\left(15\right)$

To above nonlinear optimal problem, recommend genetic algorithm for solving with stable performance; Find the solution (15) formula, can obtain the estimated value of u, ω, and A, wherein can obtain the dampingratio of system by parameters u relevant and ω with system stability $\mathrm{\&zeta;}=\frac{u}{\sqrt{{\mathrm{\&omega;}}^2+u^2}}---\left(16\right)$

Dampingratio is an important parameter of estimating half-speed vortex stability; ζ is big more, and stability is good more; ζ is more little, and stability is poor more; When ζ was zero, system had been in instability status, is accompanied by strong half-speed vortex;＜5〉half-speed vortex STABILITY MONITORING

Rotation unit to operation regularly repeats above step＜1〉to step＜4 〉, the damping ratio that at every turn obtains is drawn trend map in chronological order, can be quantitatively, intuitively half-speed vortex stability is carried out trend analysis, is estimated engine sta bility margin, and then realize supervision and prediction to half-speed vortex stability.

Images