JP2001074616A - Device for diagnozing abnormality of rotating machine - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect the peak value of abnormal vibration even when there is noise by detecting the vibration of a rotating machine for adding a conversion signal being subjected to continuous wavelet conversion to the direction of a frequency axis, and by judging whether the vibration is abnormal or not according to the addition result. SOLUTION: By a filter, a frequency required for diagnosis is extracted, at the same time, power supply noise is eliminated, an output signal is subjected to wavelet conversion, a period signal being proportional to the rotary period of a rotating machine is detected, addition to the direction of a frequency axis is made for judging the presence or absence of abnormality. An original waveform A1 is a waveform containing an impulsive signal to be detected, and an original waveform B1 is a sinusoidal wave. When Fourier transform is carried out to the original waveforms, the original waveforms A1 and B1 become frequency spectrum such as A4 and B4, where A4 does not have a peak. The continuous wavelet conversion is made to the A1 and B1, and frequencies A2 and B2 are subjected to frequency addition processing, thus resulting in power A3 and B3 and detecting power according to the period of an impulsive signal in A3.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for diagnosing abnormality of a rotating machine which detects an abnormality when a bearing or the like of the rotating machine is damaged, and diagnoses an abnormality of a rotating machine such as a train. .

[0002]

2. Description of the Related Art A conventional abnormality diagnosis device for a rotating machine bearing is, for example, an abnormality detection device based on a frequency spectrum using a Fourier transform for a vibration signal as disclosed in Japanese Patent Publication No. 7-7043, "Electric motor diagnosis device". Tokiko Sho 61-57
No. 491, "Rolling Bearing Abnormality Diagnosis Apparatus", mainly detects an abnormality based on an envelope of a vibration signal.

When a flaw (or flaw on the shaft side) occurs in a bearing of a rotating machine, an impulse-like vibration signal is generated when the flaw is hit. This impulse-like vibration signal is generated at a period proportional to the rotation period. FIG. 6 shows an example of an impulse-like signal.

In the case of the former document, in the frequency spectrum using the Fourier transform on the signal as shown in FIG. 6, the spectrum as shown in FIG. Therefore, it is difficult to detect an impulse-like vibration signal as shown in FIG. 6 by the method using the frequency spectrum. Therefore, when there is an abnormality such as a damaged bearing, an impulse-like vibration signal is generated, but there is a problem that a peak value cannot be detected in the frequency spectrum as shown in FIG.

On the other hand, in the case of the latter document, the envelope on the positive side of the signal shown in FIG. 6 is as shown in FIG. 8, and a periodic signal can be detected. However, if the impulse-like vibration signal is hidden by noise as shown in FIG. 9, the envelope becomes as shown in FIG. 10, and the vibration signal cannot be detected.

Therefore, there is a problem that an abnormality cannot be detected unless a filter is applied to a frequency band in which an impulse-like vibration signal exists to make the impulse signal visible in the vibration signal.

[0007]

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and can detect a peak value of abnormal vibration. It is an object of the present invention to provide an apparatus for diagnosing abnormalities of a rotating machine.

[0008]

(1) An abnormality diagnosis apparatus for a rotating machine according to the present invention detects a vibration of the rotating machine, performs continuous wavelet transform, adds the converted signal in a frequency axis direction, and adds the added signal. It is to determine whether or not there is an abnormality according to the result.

(2) Further, the vibration of the rotating machine is detected, continuous wavelet transform is performed through a filter for removing noise, the converted signal is added in the frequency axis direction, and it is determined whether or not there is an abnormality according to the addition result. Is what you do.

(3) In the above (1) or (2), the autocorrelation of the addition result is calculated, and it is determined whether or not there is an abnormality according to the calculation result.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 FIG. 1 is a diagram showing a configuration of a rotating machine abnormality diagnosis device according to a first embodiment of the present invention. FIG. 2 is a graph showing a result of the diagnosis. FIG. It is a figure showing comparison with processing.

In FIG. 1, reference numeral 1 denotes a rotating machine to be subjected to vibration diagnosis, and 2 denotes a vibration detecting means for detecting the vibration of the rotating machine 1 and outputting a vibration signal in accordance with the detected vibration. Reference numeral 3 denotes a filter for extracting frequencies necessary for diagnosis and removing power supply noise (50 Hz, 60 Hz). Reference numeral 4 denotes a wavelet transform unit, which performs a wavelet transform on the output signal from the filter unit and detects a periodic signal proportional to the rotation period of the rotating machine. Reference numeral 5 denotes an adding means for adding the signals after the wavelet transform in the frequency axis direction.

Next, the operation will be described. If a continuous wavelet transform (described later) using Gabor as a basis function is used for the vibration signal as shown in FIG. 6, a time change of the vibration signal power for each frequency can be obtained. Then, by performing the addition process in the frequency axis direction on the band in which the peak shown in FIG. 7 is unclear, a time variation of the power as shown in FIG. 2 can be obtained. With this method, even if the impulse-like vibration signal is hidden by noise as shown in FIG. 9, it is possible to detect the impulse-like vibration signal.

Here, the processing of the present invention based on the continuous wavelet transform will be described in comparison with the Fourier transform. FIG. 3 shows an explanatory diagram thereof. The original waveform (A1) is a waveform containing an impulse-like signal to be detected in the present invention, and the original waveform (B1)
Is a sine wave. Actually, noise is superimposed on these waveforms, but is omitted here for simplicity.

When Fourier transform is performed on these original waveforms, the original waveform (A1) becomes (A4) and the original waveform (B1) becomes (B1).
The frequency spectrum is as shown in 4). In (B4), there is a peak at a specific frequency, but in (A4), there is no peak, and the Fourier transform cannot detect an impulse-like signal as in (A1).

Next, when continuous wavelet transform is performed on the original waveforms (A1) and (B1), the original waveform (A1) becomes (A1).
2) The original waveform (B1) is as shown in (B2). (B
For 1), the time changes with a constant value at a specific frequency. On the other hand, for (A1), the power is distributed such that it spreads on the frequency axis at the time when the impulse-like signal is generated. At this time, the power at each frequency is small.

When frequency addition processing is performed between (A2) and (B2) between two broken lines, (A3) and (B2)
The result is as shown in 3). In (B3), the power changes with time at a constant value. On the other hand, in (A3), since the power dispersed by the addition processing is added, the power can be detected in accordance with the period of the impulse-like signal. Therefore, an impulse-like signal such as (A1) can be detected by the method of the present invention based on the continuous wavelet transform.

The frequency spectrum of the vibration signal shown in FIG. 6 is as shown in FIG. The frequency distribution at this time is shown in FIG.
It changes depending on the condition of the wound. Large wounds that can be heard by the ear tend to be in the lower frequency band, and small wounds that are difficult to detect in the ear tend to be located in the higher frequency band. Therefore, several bands for frequency addition are prepared.

Further, since a steep filter is not obtained in the wavelet transform, a periodic signal can be more clearly obtained as a result of the wavelet transform and the addition process by performing a filter process on a necessary band in advance. it can.

As an example, an abnormality can be diagnosed in the following range. Acceleration sensor band: up to about 10 KHz Sampling frequency: about 20 KHz Number of rotations of rotating machine: about 300 to 1000 rotations

Next, the processing means of the continuous wavelet transform will be described. The Fourier transform of the time-series signal is expressed as in equation (1).

[0022]

(Equation 1)

[0023] Fourier transform because it convoluting the basis functions that do not converge to 0 to infinity of e ^-j ω ^t in a time-series signal f (t), we can not see the time change. Therefore, it is temporally close to 0 (it does not matter if it is not close to 0, but it is often the case that it is close to 0 in practice) and Ωo in terms of frequency.
Consider a basis function Ψ (t) localized in a region near (an arbitrary frequency selected as a basis function frequency). The convolution of this function with the time series signal is the impulse response

[0024]

(Equation 2)

Equation (2) is equivalent to convolving a filter with a time-series signal. (￣ is complex conjugate)

[0026]

(Equation 3)

W (b) is obtained by extracting a frequency near Ωo from f (t), and b is called a shift coefficient. In order to obtain components of different frequencies, different 周波 数 (t) is required. In the wavelet transform, the same expression (3) is obtained by expanding and contracting the same Ψ (t), and this is realized.

[0028]

(Equation 4)

A is what is called a scaling function. When these are put together, equation (4) is obtained, and equation (5) is obtained.

[0030]

(Equation 5)

[0031]

(Equation 6)

Equation (5) is called a wavelet transform. What continuously changes a and b is called a continuous wavelet transform. In the continuous wavelet transform, the power distribution of Expression (5) is displayed two-dimensionally with 1 / a on the vertical axis and b on the horizontal axis, and the time change of the time-series signal is observed.

In the continuous wavelet transform, when a Gabor wavelet is used as a basis function, a temporal change can be seen with the frequency on the vertical axis. The basis function of the Gabor wavelet is represented by Expression (6).

[0034]

(Equation 7)

W obtained by the basis function of the equation (6)
From (a, b), the amplitude (square root of power) can be obtained by equation (7), as in the frequency spectrum obtained from the Fourier transform.

[0036]

(Equation 8)

In this equation (7), Re is a function representing a real part and Im is a function representing an imaginary part. The result shown in FIG. 3 (A2) is obtained, and the time change of the amplitude can be seen.

Next, the result is added on the frequency axis by the adding means 5 shown in FIG. As mentioned above, the frequency axis is 1
/ A, so if 1 / a = c, the addition output (P)
Becomes the equation (9).

[0039]

(Equation 9)

In the equation (8), f1 and f2 are calculated as shown in FIG.
2) The low frequency and the high frequency of FIG.
(A3). FIG. 2 shows the result of wavelet transform of FIG. 6 or 9 and addition in the direction of the frequency axis.

As described above, according to the first embodiment, if the bearing of the rotating machine is damaged, when the rotating machine rotates, the vibration signal proportional to the rotation cycle can be reliably detected by continuous wavelet transform and frequency addition. To diagnose the abnormality of the rotating machine. In FIG. 1, although the noise is removed by the filter unit 3, the filter unit 3 may be omitted when the noise is small.

Embodiment 2 FIG. 4 is a configuration diagram of a rotating machine abnormality detection device according to Embodiment 2 of the present invention. 4 differs from FIG. 1 in that an autocorrelation calculating means 6 is added next to the adding means 5.

A certain time signal is defined as x (t), and its LAG
The autocorrelation when (delay) is m is expressed by equation (9).

[0044]

(Equation 10)

In this way, as shown in FIG. 5, by performing the autocorrelation processing as compared with FIG. 2 of the first embodiment, a periodic signal proportional to the rotation period can be more clearly detected.

At this time, the value calculated by the above equation after subtracting the average of x (t) is the self-covariance. This self-dispersion may be used.

Embodiment 3 In the above embodiment, the diagnosis target is the bearing of the rotating machine. However, the diagnosis target may be a shaft of the rotating machine corresponding to the bearing, and can be applied to the detection of vibration abnormality of the rotating machine in general. The rotating machine is intended for an electric motor, a generator, wheels of a vehicle, and other rotating objects.

[0048]

As described above, according to the present invention, since the vibration waveform of the rotating machine is wavelet-transformed and added, the peak value of the abnormal vibration can be detected, and noise can be detected. Also, abnormal vibration can be detected.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a rotating machine abnormality diagnosis device according to Embodiment 1 of the present invention.

FIG. 2 is a diagram showing an abnormality detection processing result according to the first embodiment of the present invention.

FIG. 3 is a diagram showing a comparison between an abnormality detection process according to the first embodiment of the present invention and a conventional process.

FIG. 4 is a configuration diagram of a rotating machine abnormality diagnosis device according to Embodiment 2 of the present invention;

FIG. 5 is a diagram showing an abnormality detection processing result according to the second embodiment of the present invention.

FIG. 6 is a waveform diagram of abnormal vibration of the rotating machine.

FIG. 7 is a diagram showing a conventional frequency spectrum.

FIG. 8 is a diagram showing an envelope of a conventional vibration signal.

FIG. 9 is a waveform diagram when noise is included in a vibration signal.

FIG. 10 is a diagram showing an envelope of the waveform of FIG. 9;

FIG. 11 is a diagram showing a frequency spectrum showing a difference in frequency band distribution depending on a state of a damage of a bearing.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Rotating machine 2 Vibration detection means 3 Filter means 4 Wavelet transformation means 5 Addition means 6 Autocorrelation calculation means

Claims (3)

[Claims] An apparatus for diagnosing abnormality of a rotating machine, which detects vibration of the rotating machine, performs continuous wavelet transform, adds the converted signal in the frequency axis direction, and determines whether or not there is an abnormality according to the addition result. 2. A rotating machine that detects vibration of the rotating machine, performs continuous wavelet transform through a filter for removing noise, adds the converted signal in the frequency axis direction, and determines whether or not there is an abnormality based on the addition result. Abnormality diagnosis device. 3. An abnormality diagnosis apparatus for a rotating machine according to claim 2, wherein an autocorrelation of an addition result is calculated,
An abnormality diagnosis device for a rotating machine that determines whether or not there is an abnormality according to the calculation result.