JPH0460544B2 - - Google Patents

Description

[Detailed description of the invention] [Technical field to which the invention pertains] The present invention relates to a rubbing detection method that uses acoustic vibration to detect the presence or absence of a so-called rubbing phenomenon in which a rotating part and a stationary part come into contact with each other during operation of a rotating machine. The present invention relates to a device, and particularly to a device configuration that can perform detection without being influenced by changes in vibration level that are not caused by rubbing phenomena of acoustic vibrations or by the absolute value of the vibration level.

[Prior art and its problems]

If a rubbing phenomenon occurs during operation of a rotating machine such as a steam turbine or a generator, abnormal vibrations may occur, making it impossible to continue operation or leading to a serious accident. Therefore, it is necessary to detect the occurrence of the rubbing phenomenon at an early stage and take measures to avoid the rubbing, such as lowering the rotational speed of the rotating part. The rubbing phenomenon is usually detected by detecting the acoustic vibrations generated in rotating machinery due to rubbing (hereinafter, these vibrations may be referred to as rubbing vibrations), and this rubbing vibration occurs in synchronization with the rotation of the rotating parts. There is a characteristic that

Figure 4 is a configuration diagram of a conventional rubbing detection device.
First, the configuration and operation of the apparatus shown in FIG. 4 will be explained with reference to the waveform explanatory diagram shown in FIG. 5. In both figures, 1a, 1b, 1c, and 1d are the rotating shaft, casing, bearing metal, and bearing cover of the rotating machine 1, respectively, and 1e is fixed to the rotating shaft 1a and is attached to the casing 1.
It is a rotating body that rotates within b. 2 is an acoustic sensor using a piezoelectric element formed to have a resonance frequency in a high frequency region, and this sensor 2 is connected to a bearing metal 1.
c, and detects acoustic vibrations in this bearing metal and outputs a signal 2a corresponding to the vibrations. Reference numeral 3 denotes a pulse generator that outputs a pulse signal 3a synchronized with the rotation of the rotating shaft 1a, and the signal 3a is an electric pulse that is output once every rotation of the rotating shaft 1a. 4 is an amplifier that amplifies the output signal 2a, and 5 is a bandpass filter to which the output signal 4a of the amplifier is input, and which extracts the dominant frequency component in the rubbing vibration from this input signal to improve the S/N ratio. . For example, when the casing 1b and the rotating body 1e are rubbing at position 100, the amplifier output signal 4a
Although the vibration waveform based on the rubbing vibration 100a appears to be buried in the noise, the filter 5 removes the noise as described above, so the output signal 5a of the filter does not correspond to the periodic rubbing vibration 100a. A corresponding vibration appears, albeit slightly. However, even in the signal 5a, high frequency noise components of several hundred KHz or higher still predominate. Reference numeral 6 designates an envelope detector which performs envelope detection on such a signal 5a to remove high frequency noise and obtain an output signal 6a in a low frequency band that is easy to process in subsequent steps. Since the detector 6 can perform such signal processing, when rubbing is present, vibrations corresponding to the rubbing vibrations 100a appear quite prominently in the detector output signal 6a. Hereinafter, this vibration may also be referred to as rubbing vibration. Reference numeral 20 denotes a noise removal section which includes an amplifier 4, a filter 5, and a detector 6, and removes noise from the signal 2a as described above.

7 is an average value obtained by inputting signals 6a and 3a, removing white noise included in signal 6a and not synchronized with signal 3a, and outputting a series of temporally continuous signals as output signal 7a. During calculation,
The arithmetic unit 7 is composed of an arithmetic unit 71 to which the signals 6a and 3a are input, and an output unit 72 that outputs the signal 7a as a signal according to the arithmetic result in the arithmetic unit 71. , for example, is configured to perform an average calculation operation as described below to remove white noise. That is, in the arithmetic unit 71, when a pulse constituting the signal 3a is input, the waveform of the signal 6a from the pulse input time until a predetermined data collection time has elapsed is stored in time series, and then the waveform of the signal 6a is stored in time series. When the pulses constituting 3a are input, a similar memorization operation is performed again, and the memorized waveform at this time is superimposed on the already memorized waveform, and such waveform superimposition operation is repeated a predetermined number of times N. After that, the waveform value of the cumulative superimposed waveform of the (N+1) waveforms formed as described above is divided by (N+1), and a time series signal 71a corresponding to the division result is output.
The signal 71a is then converted into the signal 7a at the output section 72. Therefore, the white noise still contained in the signal 6a is removed from the signal 7a, resulting in a signal with a high S/N ratio, and the rubbing vibration normally appears clearly in the signal 7a as a triangular pulse-like waveform. P shown in FIG. 5 is the rubbing vibration waveform obtained in this manner. Hereinafter, the above-described arithmetic operation of the arithmetic unit 7 may be simply referred to as an average arithmetic operation.

8 receives the signal 7a, compares this input signal level with the built-in threshold level S, and outputs the signal 7a.
9 is a comparator that outputs an alarm signal 8a when the level of S is higher than S, and 9 is an alarm that operates as an alarm when the signal 8a is input. 10 is a CRT display for monitoring the waveform of the signal 7a.

In FIG. 4, each part is configured as described above, so if the arithmetic unit output signal 7a is composed of the rubbing vibration waveform P and a flat waveform with no rubbing vibration, the warning signal 8a will cause the rubbing to occur. However, since the signal 7a is not actually formed only by the waveform P and the flat waveform, as will be explained below, the rubbing detection device shown in FIG. There is a problem with detection.

That is, FIG. 6 is an actual waveform diagram of the main part of the detection device in an experiment conducted using the detection device of FIG. 4, and in the figure, A is an actual measurement diagram of the pulse signal 3a,
-D are different actual measurement diagrams of the arithmetic unit output signal 7a. The inventor then discovered that the triangular pulse in diagram B indicates rubbing vibration,
It has been experimentally confirmed that the waveform changes in Figures C and D are not caused by rubbing vibration, and the inventor has also confirmed that waveform changes that are not caused by rubbing as shown in Figures C and D often appear in practice. I'm experiencing it. Therefore, the rubbing detection device shown in FIG. 4 has a problem in that it may malfunction due to waveform changes in the signal 7a that are not based on rubbing vibrations.

In FIG. 4, since the gain of the amplifier 4 is changed according to the level of the signal 2a, the level of the signal 7a in the comparator 8 corresponds to the absolute value of the acoustic vibration level, that is, the level of the signal 2a. When comparing with the threshold value S set as
It is necessary to correct the gain. For this reason, the detection device shown in FIG. 4 has the problem that handling of the device is troublesome.

[Purpose of the invention]

The present invention solves the above-mentioned problems with conventional devices, and easily performs rubbing detection without being influenced by changes in vibration level that are not caused by the rubbing phenomenon in acoustic vibrations or by the absolute value of the acoustic vibration level. The object of the present invention is to provide a rubbing detection device that can perform the following steps.

[Key points of the invention]

In order to achieve the above object, the present invention provides a first calculation means that periodically performs a predetermined average calculation on a signal from which noise has been removed from an output signal of an acoustic sensor and holds the calculation result each time as time series data; The rubbing detection device is configured to have a second calculation means that inputs the data of the first calculation means and performs a kurtosis calculation, and performs the rubbing by comparing the kurtosis calculation output value of the second calculation means with a preset reference value. By doing this, rubbing detection is not performed based on the level of acoustic vibration, and rubbing detection detects changes in vibration level that are not caused by rubbing phenomena in acoustic vibration. It is designed so that it is not affected by the sound vibration level, and is not affected by the absolute value of the acoustic vibration level.

[Embodiments of the invention]

FIG. 1 is a block diagram of one embodiment of the present invention, and the rotating machine 1 shown in FIG. 4 is omitted in this figure. In this case, the calculation unit 71 uses the pulse of the pulse signal 3a as a trigger signal and superimposes the time-series waveform of the detector output signal 6a every rotation of the rotating shaft 1a,
It is configured to perform an average calculation operation, and is configured to output a calculation end signal 71b when the average calculation is completed. 11 is a memory that holds the contents of the time series signal 71a as time series data;
consists of a calculation section 71 and a memory 11, and is a first calculation means that performs an average calculation on the output signal 6a of the noise removal section 20 and holds the calculation result as time series data. Reference numeral 12 designates a kurtosis calculator as a second calculation means that calculates a statistic K called kurtosis using the time series data held in the memory 11 when the calculation end signal 71b is input, and this calculator 12 calculates the calculation result. It is configured to output a value 12a corresponding to the value 12a. Kurtosis K is a statistic expressed by equation (1), and in this case, K is the detector output signal 6
This is an amount that is a guideline for the degree of prominence of a waveform such as a rubbing vibration waveform in a.

K=(1/σ ⁴ )・_o 〓 ⁱ⁼¹ (Xi−) ⁴ ...(1) Here, X _i is the individual time series data held in the memory 11, n is the number of this data, ,σ
are the mean value and standard deviation of X _i , respectively.

FIG. 2 is an explanatory diagram of kurtosis for various waveforms, in which the values of K for square waves, sine waves, triangular waves, and random waves are theoretical values, and the values of K for rubbing waves are actually measured values. As can be seen from the figure, the kurtosis K of normal alternating waveforms, including random waves, is K
3, but in a triangular pulse waveform such as the rubbing vibration waveform in the detector output signal 6a, the value of K is very large, K=15 to 30. Therefore, it is possible to compare the magnitude of K with a preset value to determine the presence or absence of a rubbing phenomenon.
Reference numeral 13 in FIG. 1 is a comparator which has a reference value for determining the presence or absence of a rubbing phenomenon, and when the input value 12a exceeds this reference value, it outputs an alarm signal 13a to indicate that there is rubbing and activates the alarm 9. . As mentioned above, the value of K of the wrapping wave and the K of the normal alternating waveform are
Since it is significantly different from the value of K=5, the above reference value is
The inventor of the present invention has confirmed through experiments that there is no problem in rubbing detection even if the value is selected to be about 10. In FIG. 1, each part operates as described above, and the comparator 13 performs the final rubbing detection. There is no such thing as occurring.

FIG. 3 is an explanatory diagram illustrating the relationship between the shape of a rubbing wave and the magnitude of kurtosis K. The rubbing waveform in the figure is an actually measured waveform, and the value of K is an actually measured value.
Further, θ is the rotation angle of the rotation shaft 1a mentioned above, and all the illustrated waveforms correspond to one rotation of the rotation shaft 1a. In FIG. 3, section A shows the initial stage of the occurrence of rubbing, and at this stage the degree of rubbing is light and the angle θ at which the rotating part and the stationary part come into contact is small. Therefore, the rubbing waveform has a narrow pulse width and the kurtosis K takes on a large value of 30. However, as the above-mentioned rubbing continues, the value of θ usually gradually increases, so the rubbing waveform becomes as shown in section B, and K becomes, for example,
It will be a small value like 15. In this case, since θ is large, the rubbing is severe, and it can be seen from the above that K changes depending on θ.
In other words, the change in the value of K of the rubbing wave in Fig. 2 is based on the change in the rubbing contact angle θ, and as mentioned above, K usually
shows a large value, so the detection device shown in FIG. 1 can effectively detect rubbing at an early stage.

Furthermore, the detection device shown in FIG. 1 performs rubbing detection by calculating kurtosis K, and as can be seen from equation (1), K is a dimensionless quantity that is unrelated to the level of vibration in the output signal 2a of the acoustic sensor 2. be. Therefore, in the detection device shown in FIG. 1, since the rubbing detection is not affected by the absolute value of the vibration level of the acoustic vibration, there is no need to correct the amplifier gain for the arithmetic unit output value 12a. The device becomes easier to handle.

〔Effect of the invention〕

As described above, the present invention includes a first calculation means that periodically performs a predetermined average calculation on a signal obtained by removing noise from an output signal of an acoustic sensor, and stores the calculation result each time as time-series data; The rubbing detection device is configured to have a second calculation means that inputs the data of the first calculation means and performs a kurtosis calculation, and performs rubbing detection based on the output value of the second calculation means, thereby detecting acoustic vibrations. Disabled rubbing detection based on level. Therefore, in the present invention, rubbing detection is
It has the effect of not being affected by changes in the vibration level that are not caused by the rubbing phenomenon in acoustic vibration, and it also has the effect of not being affected by the absolute value of the acoustic vibration level, that is, the level of the output value of the acoustic sensor. be. Further, since the above-mentioned kurtosis usually has a large value at the initial stage of rubbing, the present invention also has the effect of being able to detect the rubbing phenomenon at an early stage.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an embodiment of the present invention, Figs. 2 and 3 are explanatory diagrams of different kurtosis with respect to vibration waveforms, Fig. 4 is a block diagram of a conventional rubbing detection device, and Fig. 5 is a block diagram of a conventional rubbing detection device. FIG. 6 is an explanatory diagram of the waveforms of the main parts in the figure, and FIG. 6 is an actual measurement diagram of the waveforms of the main parts in FIG. 1...Rotating machine, 2...Acoustic sensor, 12...
Kurtosis computing unit as second computing means, 14...
...first calculation means, 20...noise removal section.

Claims (1)

[Claims] 1. An acoustic sensor that detects acoustic vibrations of a rotating machine, a noise removal unit that removes noise from the output signal of the acoustic sensor, and a predetermined average calculation for the output signal of the noise removal unit, and calculates the calculation result and time series. It is equipped with a first calculation means for holding data as data, and a second calculation means for inputting the data of the first calculation means to perform a kurtosis calculation and generate an output value, and the output value of the second calculation means is set in advance. A rubbing detection device characterized by detecting rubbing in the rotating machine by comparing it with a reference value.