CN101532920A - Chaos-based method for detecting weak signals of low speed and heavy-duty device - Google Patents

Abstract

The invention relates to a chaos-based method for detecting weak signals of a low speed and heavy-duty device, which is used to detect weak signals with given frequency. In the method, a vibration pickup is used to extract early weak failure symptom signals of the low speed and heavy-duty device, amplify and filter the signals; then a analog-to-digital converter is used to convert analog signals into digital signals, carry out autocorrelation and demodulating treatment on the digital signals, de-mean and normalize the treated signals and add the signals to a preset chaos oscillating system; a computer is used to refer to Hu invariant moment indicators to automatically distinguish phase diagram change of system output so as to distinguish if weak failure symptom signals with given frequency are contained in the signals. The method has the advantages that a chaos oscillator can be used to detect the weak modulation signals, the computer can automatically distinguish phase diagram change of the chaos oscillator to determine if the weak failure symptom signals are contained in the signals to be detected, can detect early weak failure symptom signals of the low speed and heavy-duty device and find hidden danger for failures in advance.

Description

A kind of method for detecting weak signals of low speed and heavy-duty device based on chaos

Technical field

The present invention relates to a kind of method for detecting weak signals of low speed and heavy-duty device based on chaology, this method is mainly used in the feeble signal that detects given frequency.

Background technology

In heavy equipmentes such as iron and steel, metallurgy, low-speed heavy-loaded gear is the important motivity drive disk assembly.These equipment are being played the part of important role in continuous transportation of producing and cycle operation.Yet because these equipment working conditions are abominable, overload is serious in short-term, impact greatly, the gear teeth are in boundary lubrication condition, and the probability that gear breaks down is big, damage in case gear occurs, continuous transportation and cycle operation with having influence in the production cause serious consequence and economic loss.Simultaneously, bad or material is uneven and alignment error because design is improper as rigidity, can make the on-stream generation impact shock of gear.When gear operation,, also can cause vibration because gear is subjected to the relative slip alternate of the alternate load and the flank of tooth.For low-speed heavy-loaded gear, its most common failure is generally gear uniform wear, spot corrosion, flank of tooth gummed, gear teeth plastic yield and broken teeth.These typical faults more can cause abnormal vibrations, and along with the deterioration amplitude of flank of tooth situation constantly enlarges.Therefore, the vibration that a variety of causes produces all has certain characteristic frequency.

Low-speed heave-load device is being carried out in the process of fault diagnosis, because the commentaries on classics of slow-speed shaft is generally several hertz to several hertz of zero points frequently, the meshing frequency of slow speed turbine stage is generally tens hertz, the interval time that causes each fault to be impacted is longer, on slow-speed shaft, be difficult to detect fault-signal, need carry out comprehensive fault analysis fault.

In the lower environment of signal to noise ratio (S/N ratio), low-speed heave-load device is carried out fault diagnosis, low-frequency weak signal is often flooded by noise, has brought certain difficulty and trouble to gear on the slow-speed shaft or bearing fault feature extraction.

Utilize the useful signal that floods in the chaotic oscillator system detection noise, be exactly as the hormetic perturbation of chaotic oscillator system cycle with measured signal, though noise is strong, but not influence of change to system state, and in case specific signal is arranged, because chaos system is to the susceptibility of cycle small-signal, even amplitude is less, system is undergone phase transition, by the identification system state, whether the decidable signal exists, thereby reaches the purpose that faint cycle small-signal under the strong background noise is detected.Why the chaotic oscillator system can detect faint periodic signal, is exactly because it is extremely responsive to the tiny signal close with the system forced frequency, on the contrary, noise is but had very strong immunity.

Yet, based on the Detection of Weak Signals of chaotic oscillator, judging that phase diagram all is human eye identification mostly when changing at present, the computer automation degree is poor.In addition,, only use demodulation method to handle, generally effect and bad for faint modulation signal.

Summary of the invention

The objective of the invention is to overcome the above-mentioned defective of conventional detection, a kind of method for detecting weak signals of low speed and heavy-duty device based on chaos has been proposed, this method not only can be discerned the phase diagram state of chaotic oscillator system output automatically, but also can detect faint modulation signal in conjunction with other coherent signal disposal route.

To achieve these goals, the present invention has taked following technical scheme:

Method for detecting weak signals of low speed and heavy-duty device among the present invention may further comprise the steps:

1) determines the fault characteristic frequency that may comprise in the tested modulation signal according to the system under test (SUT) drive mechanism, set up corresponding chaotic oscillator system;

2) according to the fault characteristic frequency in the step 1), the phase diagram of determining the output of described chaotic oscillator system with reference to Hu formula invariant moments becomes the HuShi invariant moments threshold value of large scale cycle status and the amplitude thresholds of corresponding reference signal by chaos state, chaos state when then the amplitude of the reference signal in the chaotic oscillator system being made as the amplitude thresholds less than reference signal, have only reference signal in the chaotic oscillator system this moment;

3) extract the early stage Weak fault characteristic signal of low-speed heave-load device by vibration transducer, be converted to digital signal with after this signal amplification filtering by analog to digital converter, digital signal is carried out auto-correlation processing, again the signal after the auto-correlation is carried out demodulation analysis, then the signal after the demodulation is gone average and normalized and will handle after signal join step 2) in the chaotic oscillator system that sets, if the phase diagram of chaotic oscillator system output becomes the large scale state, then illustrate in the signal that vibration transducer extracts and have fault characteristic frequency signal to be measured, if the phase diagram of chaotic oscillator system output still is in chaos state, then illustrate in the signal that vibration transducer extracts not have fault characteristic frequency signal to be measured;

The method that the phase diagram that judgement chaotic oscillator system exports in the step 3) is in large scale state or chaos state is: a quantizating index that adopts the HuShi invariant moments to change as the identification chaotic oscillator output phase figure of system, when the HuShi invariant moments numerical value in the chaotic oscillator system is higher than step 2) during the HuShi invariant moments threshold value determined, the chaotic oscillator system is in the large scale cycle status; When the HuShi invariant moments numerical value in the chaotic oscillator system is less than or equal to step 2) during determined HuShi invariant moments threshold value, then the chaotic oscillator system is in chaos state.

Compare with conventional detection, the present invention has the following advantages:

The present invention can detect the known Weak fault characteristic signal of early stage frequency, and can discern the phase diagram situation of change of chaotic oscillator system output automatically by HuShi invariant moments index.

Description of drawings

Fig. 1 fundamental diagram of the present invention

Fig. 2 is based on the signal processing algorithm process flow diagram of the detection feeble signal of chaotic oscillator and other theory

Fig. 3 workflow diagram of the present invention

The time-domain diagram of the damage of bearings signal (original signal) that Fig. 4 (a) the present invention realizes

The amplitude spectrum of the damage of bearings signal (original signal) that Fig. 4 (b) the present invention realizes

The HuShi invariant moments changing trend diagram that Fig. 5 the present invention realizes

The time history diagram of the chaotic oscillator system output that Fig. 6 (a) the present invention realizes when not adding original signal

The phase diagram of the chaotic oscillator system output that Fig. 6 (b) the present invention realizes when not adding original signal

Fig. 7 (a) be the present invention realize original signal is carried out time-domain diagram after the auto-correlation processing

Fig. 7 (b) be the present invention realize original signal is carried out amplitude spectrum after the auto-correlation processing

What Fig. 8 (a) the present invention realized carries out time-domain diagram after the Hilbert demodulation to the signal after the auto-correlation

What Fig. 8 (b) the present invention realized carries out amplitude spectrum after the Hilbert demodulation to the signal after the auto-correlation

Fig. 9 (a) the present invention realizes removes Hilbert restituted signal among Fig. 7 to be input in the chaotic oscillator system time history diagram of this system's output this moment after average and the normalization as signal to be checked

Fig. 9 (b) is that realize Hilbert restituted signal among Fig. 7 is gone of the present invention is input in the chaotic oscillator system as signal to be checked after average and the normalization, the phase diagram of this system's output this moment

Embodiment

The invention will be further described below in conjunction with accompanying drawing:

Data acquisition in the present embodiment and early stage handle as shown in Figure 1, and the internal signal Processing Algorithm flow process of chaotic oscillator system as shown in Figure 2.At first determine to comprise which fault characteristic frequency in the measured signal, set up corresponding chaotic oscillator system according to the system under test (SUT) drive mechanism.Then according to this fault characteristic frequency, the phase diagram of determining the output of described chaotic oscillator system with reference to the HuShi invariant moments becomes the HuShi invariant moments threshold value of large scale cycle status and the amplitude thresholds of corresponding reference signal by chaos state, chaos state when the amplitude of the reference signal in the chaotic oscillator system is made as the amplitude thresholds that is slightly smaller than reference signal, have only reference signal in the chaotic oscillator system this moment.Then vibration signal is carried out auto-correlation processing, again the signal after the auto-correlation processing is carried out demodulation analysis, what take in the concrete embodiment is Hilbert (Hilbert) demodulation method.Afterwards, signal after the demodulation is gone average and normalized, the signal of handling is joined in the chaotic oscillator system that sets before, if the phase diagram of system's output becomes the large scale state, then illustrate and have fault characteristic frequency signal to be measured in the signal, if phase diagram still is in chaos state, then explanation does not exist.Simultaneously can allow the Computer Automatic Recognition system whether become the large scale cycle status with reference to HuShi invariant moments index from chaos state.

Use Detection of Weak Signals scheme that this method carries out low-speed heave-load device as shown in Figure 3, mainly solve and improve signal to noise ratio (S/N ratio), quantitatively discern problems such as CHAOTIC PHASE figure variation, modulation signal analysis.Signal is carried out auto-correlation processing can give prominence to the fault signature signal, improve signal to noise ratio (S/N ratio).A quantizating index that adopts the HuShi invariant moments to change as identification CHAOTIC PHASE figure, when the HuShi invariant moments of chaotic oscillator system correspondence is higher than the HuShi invariant moments threshold value of this chaotic oscillator system, phase diagram is in the large scale cycle status, when being lower than this threshold value, then is in chaos state.In the signal of gathering, often have modulation signal, for the ease of the chaotic oscillator systematic analysis, demodulation methods such as the use Hilbert of elder generation are handled signal, then the data that obtain after the demodulation are gone average and normalized, the signal after handling is joined in the chaotic oscillator system.

Damage of bearings data instance with certain high line bar factory, the data of using are data of 20 days nearly before fault takes place, and sample frequency is 10kHz, and sampling number is 2048, the bearing inner race fault characteristic frequency that calculates is 116.619Hz, and the oscillogram of original signal and amplitude spectrum are as shown in Figure 4.Set up following chaotic oscillator system:

$\left\{\begin{array}{c}\stackrel{\&CenterDot;}{x}=2\mathrm{\&pi;}{f}_{0}y\\ \stackrel{\&CenterDot;}{y}=2\mathrm{\&pi;}{f}_0(-\mathrm{cy}+x-x^3+{\mathrm{<figure-callout\; id="0"\; label="F"\; filenames="A200910082043E00111.png,A200910082043E00121.png"\; state="\{\{state\}\}">F</figure-callout>}}_0\cos \left(2\mathrm{\&pi;}{f}_{0}t\right)+n\left(t\right))\end{array}\right.---\left(1\right)$

In the formula: f ₀Be the frequency of chaotic oscillator internal system periodic perturbation power, c is for being ratio of damping, F ₀Cos (2 π f ₀T) be reference signal, n (t) is a signal to be checked, F ₀Be the amplitude of reference signal ,-x+x ³Be nonlinear restoring force, 2 π f ₀Y is the differential of x, and y is an intermediate variable.

For bianry image, at R ²P+q rank square on the plane is:

$m_{\mathrm{pq}}=\underset{y}{\mathrm{\&Sigma;}}\underset{x}{\mathrm{\&Sigma;}}f(x,y)x^p{y}^q---\left(2\right)$

In the formula, (x is that image is in coordinate points (x, y) gray scale on y) to f.

m _PqDepend on the position of image in coordinate, do not possess translation invariance, p+q rank central moment μ _PqSatisfy translation invariance, it is defined as

${\mathrm{\&mu;}}_{\mathrm{pq}}=\underset{y}{\mathrm{\&Sigma;}}\underset{x}{\mathrm{\&Sigma;}}{(x-\stackrel{\&OverBar;}{x})}^p{(y-\stackrel{\text{-}}{y})}^{q}f(x,y),---\left(3\right)$

In the formula (x, y) barycenter of representative image: $\left\{\begin{array}{c}\stackrel{\&OverBar;}{x}=\frac{m_{10}}{m_{00}}\\ \stackrel{\&OverBar;}{y}=\frac{m_{01}}{m_{00}}\end{array}\right..$

To μ _PqCarry out normalization process and obtain η _PqExpression formula be

${\mathrm{\&eta;}}_{\mathrm{pq}}=\frac{{\mathrm{\&mu;}}_{\mathrm{pq}}}{{\mathrm{\&mu;}}_{00}^{1+\frac{p+\mathrm{<figure-callout\; id="2"\; label="q"\; filenames="A200910082043E00111.png,A200910082043E00122.png"\; state="\{\{state\}\}">q</figure-callout>}}{2}}},p+q\&GreaterEqual;2---\left(4\right)$

η _PqSatisfied translation and flexible unchangeability, but do not satisfied rotational invariance, Hu has obtained 72 complete rank and 3 rank invariant moments by researching and analysing

:

The metric of presentation video degree of divergence, the degree of divergence of image is big more, then

Big more;

The symmetric metric of presentation video, image symmetrical are good more, then

More little.

According to the various characteristics of square, the present invention selects for use

Invariant moments finds that by a large amount of emulation phase diagram has different at different states as the quantizating index of discerning phase diagram automatically And its variation has certain rules, under the chaos state and large scale under the cycle

Difference clearly, thereby can quantitatively discern the state of phase diagram, reach the purpose of automatic identification.

For the such bianry image of CHAOTIC PHASE figure, when calculating invariant moments, can think R ²All identical (f (x, y)=1), phase diagram do not have the gray scale of each coordinate points of process to be 0 (to be f (x, y)=0), in substitution formula then (2)～(4), to calculate the invariant moments of CHAOTIC PHASE figure the gray scale of each coordinate points of phase diagram process on the plane

For formula (1), during initial setting up chaotic oscillator system, no signal to be checked in this system, promptly n (t) is 0.Make ratio of damping c=0.5, reference signal frequency ┌ is made as the calculating fault features frequency 116.619Hz of bearing, counting is N=2048, F ₀Initial value be 0.42, F ₀Change step be 0.01, stop value is 0.92, the calculating step-length h=0.0001 of Duffing oscillator adopts 4 rank Runge-Kutta algorithms to find the solution, initial value x (0)=0, y (0)=0, the invariant moments that obtains

Trend as shown in Figure 5.

As can be seen from Figure 5, the phase diagram of this chaotic oscillator system output is become the HuShi invariant moments threshold value of large scale cycle status by chaos state

Be about 2.3, the amplitude thresholds F of corresponding reference signal _bBe about 0.65.So regulate F ₀Make the chaotic oscillator system be in and be slightly smaller than threshold value F _bState, the F in this example ₀Be made as 0.628, this moment, the phase plane trajectory of output was in chaos state, as shown in Figure 6.Original signal is carried out auto-correlation processing earlier, and time-domain diagram after the auto-correlation and amplitude spectrum are as shown in Figure 7.Again the signal after the auto-correlation is carried out the Hilbert demodulation, time-domain diagram after the demodulation and the local amplitude spectrum that amplifies are as shown in Figure 8.Then the signal after the demodulation is gone average and normalized, at last the signal after handling is input in this chaotic oscillator system as signal to be checked, the HuShi invariant moments that obtain this moment is 2.339, greater than HuShi invariant moments threshold value 2.3, the signal that includes frequency to be checked in this signal to be checked is described, the signal that promptly comprises fault characteristic frequency 116.619Hz, the state of this chaotic oscillator system output simultaneously as shown in Figure 9, phase diagram wherein is in the large scale cycle status really.

As can be seen from Figure 7, vibration signal can be found out the impact phenomenon of some Weak fault characteristic signals a little through after the auto-correlation processing, does not see characteristic frequency from the amplitude spectrum of correspondence; Again signal is carried out after the Hilbert demodulation, also do not see the Weak characteristic frequency in the corresponding amplitude spectrum.So the signal after the demodulation is gone to be input in the chaotic oscillator system as measured signal after average and the normalization.Do not add before the measured signal, the output of chaotic oscillator system as shown in Figure 6, system is in chaos state, add after the signal to be detected, the output of chaotic oscillator system produces great changes, become the large scale cycle status by chaos state, as shown in Figure 9, illustrate to contain the Weak fault characteristic signal that will detect to some extent in the measured signal.

The present invention at first proposes to detect faint modulation signal with the chaotic oscillator system, has solved the detection problem of the faint modulation signal in the low-speed heave-load device.Discern the variation of phase diagram in conjunction with the HuShi invariant moments as quantizating index, use autocorrelation method to improve signal to noise ratio (S/N ratio), differentiate the Weak characteristic signal automatically by computing machine and whether exist.

Claims (1)

1, a kind of method for detecting weak signals of low speed and heavy-duty device based on chaos is characterized in that, this method may further comprise the steps:

1) determines the fault characteristic frequency that may comprise in the tested modulation signal according to the system under test (SUT) drive mechanism, set up corresponding chaotic oscillator system;

2) according to the fault characteristic frequency in the step 1), the phase diagram of determining the output of described chaotic oscillator system with reference to Hu formula invariant moments becomes the HuShi invariant moments threshold value of large scale cycle status and the amplitude thresholds of corresponding reference signal by chaos state, chaos state when then the amplitude of the reference signal in the chaotic oscillator system being made as the amplitude thresholds less than reference signal, have only reference signal in the chaotic oscillator system this moment;

3) extract the early stage Weak fault characteristic signal of low-speed heave-load device by vibration transducer, be converted to digital signal with after this signal amplification filtering by analog to digital converter, digital signal is carried out auto-correlation processing, again the signal after the auto-correlation is carried out demodulation analysis, then the signal after the demodulation is gone average and normalized and will handle after signal join step 2) in the chaotic oscillator system that sets, if the phase diagram of chaotic oscillator system output becomes the large scale state, then there is fault characteristic frequency signal to be measured in the signal that vibration transducer extracts, if the phase diagram of chaotic oscillator system output still is in chaos state, then there is not fault characteristic frequency signal to be measured in the signal that vibration transducer extracts;

The method that the phase diagram that judgement chaotic oscillator system exports in the step 3) is in large scale state or chaos state is: a quantizating index that adopts the HuShi invariant moments to change as the identification chaotic oscillator output phase figure of system, when the HuShi invariant moments numerical value in the chaotic oscillator system is higher than step 2) during determined HuShi invariant moments threshold value, the chaotic oscillator system is in the large scale cycle status; When the HuShi invariant moments numerical value in the chaotic oscillator system is less than or equal to step 2) during determined HuShi invariant moments threshold value, then the chaotic oscillator system is in chaos state.

Images