JPH09229762A - Method and apparatus for monitoring of abnormality of instrument - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and an apparatus for monitoring of a plurality of instruments by sound, which remove a disturbance sound and detect various abnormal sounds with high sensitivity. SOLUTION: In a preliminary experiment, sound power at the monitoring frequency of a normal sound is analyzed statistically (Steps 101 to 103), the threshold value of acoustic power for abnormality detection is set at a value in which an acoustic-power change width is added to the man value of the normal sound, and the acoustic-power change width is decided in such a way that the abnormality sound is not detected erroneously (Step 104). In addition, on the basis of a plurality of disturbance sounds, to be removed, which are decided by the duration time and by the acoustic power at the monitoring frequency, the disturbance sound in which the product of the duration time by the acoustic power becomes a maximum value is selected (Steps 105 to 107), the maximum value of the product of the duration time by the acoustic power is divided by the acoustic-power change width, and the averaging time of the acoustic power is preliminary decided (Step 108).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to abnormality monitoring of various equipment such as in a power plant, for example, equipment installed in a gas turbine or a water supply pump room in a combined cycle of a thermal power plant, using acoustic signals or the like. The present invention relates to a device abnormality monitoring method and device suitable for the above.

[0002]

2. Description of the Related Art Conventionally, for abnormality monitoring of a power plant including a gas turbine, a water supply pump room, etc., an acceleration sensor or an AE sensor is attached to a specific device to detect an abnormality peculiar to each monitored device. I've been trying. But,
There are many devices used in power plants, etc., and with the development of conventional technology, it is difficult to constantly monitor all devices because the number of monitoring signals increases as the number of monitoring targets increases. . Therefore, under the present circumstances, during actual operation of the plant, abnormality monitoring is also performed by the visual and auditory senses of patrol personnel.

In recent years, in abnormality monitoring of a power plant or the like, an acoustic signal is used in place of the auditory sense of a patrolman, in other words, the human sense of hearing, to complement an acceleration sensor, an AE sensor, or the like, or to combine them. There is a move to monitor more advanced plants. However, while using acoustic signals, multiple devices can be monitored simultaneously,
There are many types of abnormal sounds to be monitored, and the acoustic properties at the time of abnormality are various, and it is difficult to detect all abnormal sounds. Further, since a plurality of disturbance sounds are usually mixed, it is necessary to distinguish between the disturbance sound and the abnormal sound and monitor the abnormality while removing the mixed disturbance sound.

On the other hand, as a plant abnormality monitoring technique using sound, a method using a frequency spectrum of an acoustic signal is known. This is to perform spectrum analysis of normal sound in advance to determine a normal range of frequency spectrum intensity, and monitor whether the frequency spectrum of the monitoring sound deviates from this normal range. In this method, the average number of times the frequency spectrum is added is determined based on conditions such as frequency spectrum stabilization, disturbance noise removal, and constraint on abnormal monitoring time. The range is set.

As related arts, for example, JP-A-63-281025 and JP-A-4-316198.
Japanese Unexamined Patent Publication (Kokai) No. H10-264, pp. 157-334 is known.

[0006]

By the way, various sounds are generated in a power plant or the like, and can be classified into normal sound, abnormal sound and disturbance sound from the viewpoint of sound monitoring. Normal sound is the sound generated by the monitored device during normal operation, and it is expected that the operating status of the plant and changes over time will occur. The abnormal sound is exactly the sound associated with the abnormality of the device and is the sound to be detected by the present invention. Disturbance sound is a sound that is not associated with an abnormality of the equipment, such as that generated with the work of a worker (for example, when a spanner is dropped), and should be removed so as not to make an erroneous determination.

Of these, when the abnormal sound and the disturbance sound are classified according to their characteristics, as shown in Table 1 below, the abnormal sound is a continuous sound and an impact sound, and the disturbance sound is a continuous sound, an impact sound and an intermittent sound. Each can be classified. The continuous abnormal sound has a long duration, and the sound power varies from large to small.
Abnormal shock noise has a short duration and very large sound power, and is likely to lead to a serious accident in a short time. In addition, regarding the disturbance sound, the sound of any classification has a relatively small duration and sound level.

[0008]

[Table 1]

The present invention has been made in view of the circumstances of the prior art as described above, and an object thereof is to detect a normal sound having fluctuation (variation) in order to monitor abnormality of a plurality of devices by utilizing sound. An object of the present invention is to provide a device abnormality monitoring method and device capable of removing disturbance noise and detecting abnormal noise of various properties with high sensitivity.

[0010]

In order to achieve the above object, a first means is a device abnormality monitoring method for acoustically monitoring an abnormality during operation of a device to be monitored. A second step of determining the acoustic power fluctuation range based on a threshold value obtained by adding the acoustic power fluctuation range to the average value so as not to erroneously detect a normal driving sound as an abnormal sound. Process,
A third step of selecting a disturbance sound having a maximum value of a product of the duration time and the sound power from a plurality of disturbance sounds to be removed defined by the duration time and the sound power of the monitoring frequency, and the maximum value. Is divided by the fluctuation range of the acoustic power as a averaging time of the acoustic power at the monitoring frequency, and a preliminary step including a fourth step, and the averaging time of the acoustic power of the sound signal at the monitoring frequency is determined. During the period, the monitoring step of monitoring whether or not the value after the arithmetic mean exceeds the threshold value is determined, and whether or not an abnormal sound is generated from the monitored device is determined by whether or not the value exceeds the threshold value. It is characterized by doing.

In this case, a plurality of monitoring frequencies are set,
In the preliminary step, the first to fourth steps are performed for each monitoring frequency, and in the monitoring step, it is monitored whether or not the acoustic power after the averaging exceeds a preset threshold value for each monitoring frequency. You can also Further, instead of the acoustic power of the monitoring frequency, an overall value which is the total energy of the acoustic signal can be used, or an integrated value of the acoustic power of the frequency section can be used.

A second means is a device abnormality monitoring method for monitoring an abnormality of a monitored device during operation, wherein a sensor is arranged in the monitored device, and a feature amount is extracted from a detection signal of the sensor, Obtain the average value of the characteristic amount during normal operation in advance, and determine the variation width so as not to be erroneously determined to be abnormal during normal operation with a threshold value obtained by adding the variation width of the characteristic amount to this average value. A disturbance signal that gives the maximum value of the product of the duration and the characteristic amount level is selected from a plurality of disturbance signals to be removed defined by the characteristic amount level, and a value obtained by dividing the maximum value by the fluctuation range is calculated. A preliminary step of determining as the arithmetic mean time of the characteristic amount, and a monitoring step of monitoring whether or not the value after arithmetic mean of the characteristic amount of the detection signal of the sensor for the arithmetic mean time exceeds the threshold value. Including It is characterized by determining the presence or absence of abnormality in the monitoring target device depending on whether it exceeds the threshold.

A third means is a device abnormality monitoring device for acoustically monitoring the abnormality of a monitored device, and means for detecting an acoustic signal of the monitored device and acoustic power from the acoustic signal detected by this detecting means. Means for calculating, and means for determining the abnormality determination threshold and the averaging time from the acoustic power calculated by the calculating means,
A means for storing the abnormality determination threshold value and the averaging time determined by the deciding means, and a sound of the monitoring frequency calculated by the calculating means during the averaging time stored in the storing means. A means for averaging the power and a means for judging an abnormality from the acoustic power after the averaging calculated by the means for averaging, based on the abnormality judgment threshold value stored in the means for storing. It is characterized by having.

In this case, the calculating means may calculate the acoustic power for each of a plurality of monitoring frequencies from the acoustic signal detected by the detecting means. Further, instead of the acoustic power of the monitoring frequency, an overall value which is the total energy of the acoustic signal can be used, or an integrated value of the acoustic power of the frequency section can be used.

A fourth means is a device abnormality monitoring device for monitoring an abnormality of a monitored device during operation, and a feature quantity is extracted from a sensor arranged at a monitoring position of the monitored device and a detection signal of this sensor. Means, means for determining an abnormality determination threshold value and arithmetic mean time from the feature amount detected by this extracting means, means for storing the abnormality determination threshold value and arithmetic mean time determined by this determining means, and During the arithmetic mean time stored in the storing means, from the means for arithmetic mean processing of the feature amount extracted by the extracting means, and the characteristic amount after arithmetic mean calculated by the arithmetic mean processing means And a means for judging an abnormality based on the abnormality judgment threshold value stored in the storing means.

The third and fourth means may further be provided with man-machine interface means for changing the processing contents of the determining means and the determining means according to the judgment of an observer.

That is, according to the present invention, the frequency of the monitored acoustic signal is analyzed, the acoustic powers are added and averaged to remove disturbance noise, and it is monitored whether the added and averaged characteristic acoustic powers exceed a threshold value. did. In this method, it is necessary to previously determine the threshold value for determination and the number of times of averaging performed for removing disturbance noise for each monitoring frequency. As the threshold value, a normal sound is statistically analyzed, an average value and a standard deviation thereof are calculated, and a value that does not cause a false determination of the normal sound is set. As for the averaging count, an optimum value is determined for each monitoring frequency according to the detection limit curve as shown below.

That is, when the abnormal sound and the disturbance sound shown in Table 1 are plotted on the plane of the duration X and the acoustic power Y for each frequency, the regions of the abnormal sound and the disturbance sound are as shown in FIG. Here, the duration and sound power of the abnormal sound that can be detected are X and Y, respectively, the addition average time is t, and the threshold sound power is α. The threshold value α is a value obtained by adding the sound power fluctuation width m to the average value s of normal sounds.

First, consider the case where the duration X of the abnormal sound is smaller than the averaging time t. Since the abnormal sound is added for X hours in the arithmetic mean time of t hours, the component of the abnormal sound in the acoustic power after the arithmetic mean is X / t times the acoustic power Y. Further, since the normal sound component is the average value s, the acoustic power after the arithmetic mean is the sum of XY / t and s. If the acoustic power after the averaging is larger than the threshold value α, it can be detected as an abnormality, and therefore the detectable area is in the range of Y (X / t) + s> α ... (1).

Next, consider the case where the duration X of the abnormal sound is larger than the averaging time t. Since the abnormal sound is added during the averaging, the component of the abnormal sound in the acoustic power after the averaging is the acoustic power Y. Therefore, the detectable area is in the range of Y + s> α (2).

From the above results, the abnormality detectable area at the monitoring frequency is shown by the shaded area in FIG. Here, the curve formula Y = mt / X [= (α-s) t / X] (3) that determines the abnormality detection limit is defined as the detection limit curve.

In FIG. 4, in order to improve the abnormality detection sensitivity, it is necessary to bring the detection limit curve represented by the equation (3) close to the origin or reduce the acoustic power fluctuation range Y = m (4). There is. However, the fluctuation range of sound power m
Is set based on the result of statistical analysis of normal sound so that normal sound is not erroneously determined, and cannot be arbitrarily reduced. Therefore, the detection limit curve is brought close to the origin so as not to make an erroneous determination of disturbance noise, that is, the averaging time t is set under a certain acoustic power fluctuation width m.
To improve the detection sensitivity.

In other words, in this method, the disturbance sound is analyzed, the distribution of the disturbance sound is obtained for each monitoring frequency as shown in FIG. 4, and the disturbance sound is not included in the detectable area, and is expressed by the above equation (3). Determine the detection limit curve shown. This is equivalent to finding the maximum value of the product of the duration of the disturbance sound and the sound power. Further, the normal sound is statistically analyzed, the sound power fluctuation width m is obtained, and further, the X coordinate of the intersection of the detection limit curve and the equation (4) is obtained, and the addition average time t is set.

The above object is achieved by performing the abnormality monitoring by using the optimum addition time t and the acoustic power fluctuation width m obtained for each monitoring frequency as described above.

When monitoring a specific abnormal sound,
The frequency with high acoustic power of the abnormal sound is obtained in advance as the monitoring frequency, and the distribution of the disturbance sound at the monitoring frequency is shown in FIG.
And the optimum detection limit curve (Y = mt
/ X), the normal sound is statistically analyzed to determine the sound power fluctuation width m, and the optimum averaging time t is determined and monitored based on these results, thereby erroneously determining the disturbance sound. It is possible to detect an abnormal sound with high sensitivity without doing so.
Further, by monitoring the frequency with high acoustic power of a specific abnormal sound, it becomes possible to detect with high sensitivity even an abnormal sound having the same distribution as the disturbance sound at other frequencies.

When an impact sound having a high acoustic power and a short duration is detected, an overall value (total energy of all frequencies) is used instead of the acoustic power of the monitoring frequency as shown in FIG. By setting the averaging time t to be the shortest (without averaging) and setting the sound power fluctuation width m to be larger than the maximum overall value of the normal sound and the disturbance sound, as shown in FIG. It is possible to detect an abnormal impact sound with high sensitivity.

The integrated value of the sound power in the frequency section, for example, the sound power of each band in the 1/3 octave analysis is monitored instead of the sound power of a single frequency, and the optimum sound power fluctuation width m is obtained for each band. By obtaining the addition average time t and monitoring it in all bands, it is possible to detect an unknown abnormal sound with high sensitivity.

By performing the processing as described above, it becomes possible to detect the continuous abnormal noise, the shock abnormal noise, and the unknown abnormal noise with high sensitivity.

[0029]

DETAILED DESCRIPTION OF THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

1. Abnormality Monitoring Process FIG. 1 is a flowchart showing a general processing procedure of the device abnormality monitoring method according to the embodiment of the present invention. This abnormality monitoring process includes a process at the time of preliminary experiments and a process at the time of monitoring. The process during the preliminary experiment corresponds to the preliminary process in the claims, and the process during the monitoring corresponds to the monitoring process.

1.1 Processing in Preliminary Experiment First, the processing procedure in the preliminary experiment will be described. In the preliminary experiment, a normal sound analysis mode M ₁ , a disturbance sound analysis mode M ₂ ,
There are three modes, namely, the decision parameter decision mode M ₃ . In the normal sound analysis mode M ₁ , first, the sound signal of the normal sound of the monitoring target device that is normally operating is taken in (step 101), and then the sound power of the monitoring frequency is calculated (step 102). The average value s is calculated by the statistical analysis of the sound (step 103). Further, in order to eliminate erroneous determination of normal sound, a value (m + s) obtained by adding the sound power fluctuation width m of normal sound to the average value s, that is, α = s + m (5) is set as the abnormality determination threshold ( Step 104).

In the disturbance sound analysis mode M ₂ , the disturbance sound is generated while the monitored device is stopped and the acoustic signal is taken in (step 105), and the duration of the disturbance sound and the distribution of the acoustic power are obtained (step 106). ), The maximum value β of the product of the acoustic power and the duration in the distribution is obtained (step 107).

In the determination parameter determination mode M ₃ , the product of the acoustic power fluctuation duration m obtained by the normal sound analysis mode M ₁ and the acoustic power and duration in the disturbance sound distribution region obtained by the disturbance sound analysis mode M ₃ is calculated. Maximum value β
Then, the arithmetic mean time t is determined by t = β / m (6) (step 108).

1.2 Processing During Monitoring Next, a processing procedure during actual monitoring will be described.

At the time of monitoring, first, the acoustic signal of the device to be monitored is fetched (step 109), the acoustic power of the monitoring frequency is calculated (step 110), and the acoustic power is calculated in the decision parameter determination mode M _{3 and} the arithmetic mean time t is calculated.
During step (111), and the acoustic power after the averaging is determined by the normal sound analysis mode M ₁ for the abnormality determination threshold α
When it exceeds, it is determined to be abnormal (step 112).

2. Basic Functional Configuration of Device Abnormality Monitoring Device FIG. 2 is a functional block diagram showing the basic configuration of the device abnormality monitoring device according to the present embodiment.

This device abnormality monitoring device is provided with the sound detection unit 1
An acoustic signal processing unit 2, a determination parameter calculation unit 3,
The arithmetic mean processing unit 4, the abnormality determination unit 5, the determination parameter storage unit 6, and the man-machine interface 7 are basically configured.

In the device abnormality monitoring device having such components, the man-machine interface 7 by the monitor indicates that the normal sound analysis mode M ₁ during the preliminary experiment is the normal sound analysis mode during the preliminary experiment. It is input to the determination parameter calculation unit 3 and the abnormality determination unit 5 via.
Then, the acoustic detector 1 detects an acoustic signal of a normal sound of the monitoring target device that is operating normally (step 10).
1), the acoustic power of the monitoring frequency is calculated by the acoustic signal processing unit 2 (step 102), and the determination parameter calculation unit 3
It is stored in the determination parameter storage unit 6 via. The determination parameter calculation unit 3 statistically processes the sound power of normal sound to obtain an average value s, standard deviation, etc. (step 10
3) Based on the result, the sound power fluctuation range m is determined, and the abnormality determination threshold value α is determined as in Expression (5) (step 104). The average value s of the normal sound and the sound power fluctuation range m are determined.
The abnormality determination threshold value α is stored in the determination parameter storage unit 6.

Therefore, in this embodiment, the sound detection unit 1 serves as means for detecting the sound signal of the monitored device, the sound signal processing unit 2 serves as means for calculating the sound power, and the judgment parameter computing unit 3 serves as the abnormality judgment threshold value. And the means for determining the averaging time, the determination parameter storage unit 6 for storing the abnormality determination threshold and the averaging time, and the averaging processing unit 4 for averaging the acoustic power of the monitoring frequency. The section 5 corresponds to each means for determining abnormality.

In the disturbance sound analysis mode M ₂ , first, the observer inputs the disturbance sound analysis mode at the time of the preliminary experiment to the judgment parameter calculation unit 3 and the abnormality judgment unit 5 via the man-machine interface 7. Then, a disturbance sound is generated while the monitored device is stopped, the acoustic signal is detected by the acoustic detection unit 1 (step 105), and the acoustic power and duration of the monitoring frequency are calculated by the acoustic signal processing unit 2 (step 106). ), And input to the determination parameter calculation unit 3. The determination parameter calculator 3 obtains the distribution of the disturbance sound as shown in FIG. 4, and the disturbance sound can be detected in the detection limit curve given by the following equation Y = β / X (7) That is, the minimum β (maximum value of the product of the sound power of the disturbance sound and the duration) that does not satisfy Y> β / X (8) is calculated (step 107) and stored in the determination parameter storage unit 6.

In the decision parameter determination mode M ₃ which is the last in the preliminary experiment, the decision parameter calculation section 3 and the abnormality determination section 5 are sent by the observer via the man-machine interface 7 to the decision parameter determination mode in the preliminary experiment. To enter. Then, in the determination parameter calculation unit 3, from the determination parameter storage unit 6, the sound power fluctuation width m
And β, which is the parameter of the detection limit curve, are taken in, and the averaging time t is determined by the equation (6) (step 108) and stored in the abnormality determination parameter storage unit 6.

When the abnormality monitoring mode is entered, the monitoring parameter indicates that it is in the abnormality monitoring mode through the man-machine interface 7 by the determination parameter calculation unit 3 and the abnormality determination unit 5.
To enter. Then, the acoustic signal of the monitoring target device is detected by the acoustic detection unit 1 (step 109), and the acoustic power of the monitoring frequency is calculated by the acoustic signal processing unit 2 (step 11).
0), the acoustic power is arithmetically averaged during the arithmetic mean time stored in the judgment parameter storage unit 6 by the arithmetic mean processing unit 4 (step 111), and the abnormality judgment unit 5 stores the acoustic power after the arithmetic mean and the judgment parameter. The presence or absence of abnormality is determined by the abnormality determination threshold value α stored in the unit 6 (step 1
12) Inform the monitoring person of the determination result via the man-machine interface 7.

3. Device abnormality monitoring device applied to thermal power plant Next, an embodiment in which the device abnormality monitoring device according to the present invention is applied to an abnormality monitor of a gas turbine of a thermal power plant will be described with reference to the drawings.

FIG. 6 is a block diagram showing a hardware configuration diagram of the device abnormality monitoring apparatus according to this embodiment. The device abnormality monitoring device according to this embodiment includes a computer 10.
Microphone 11, A / D converter 12, DSP
It is composed of a (digital signal processor) 13, a storage device 14 and a man-machine interface 15. In this configuration, the microphone 11 detects the acoustic signal of the monitored device, the A / D converter 12 converts the acoustic signal into a digital signal, the DSP 13 calculates the acoustic power of the monitoring frequency, and the arithmetic mean is calculated.
At the time of the preliminary experiment, various abnormality determination parameters are calculated based on the calculation result of the DSP 13 and stored in the storage device 14, and at the time of actual monitoring, the calculation result of the DSP 13 and the storage device 1
The abnormality determination is performed based on the abnormality determination parameter stored in 4, and the determination result is notified to the monitoring person through the man-machine interface 15, and the mode conversion such as the preliminary experiment and the abnormality monitoring is performed through the man-machine interface 15. This is done according to the instructions of the observer.

Therefore, in this embodiment, the computer 11 serves as a means for detecting the acoustic signal of the equipment to be monitored, the A / D converter 12 and the DSP 13 serve as a means for calculating the acoustic power, and the computer 10 uses the abnormality determination threshold value and the addition. Corresponding to the means for deciding the averaging time, the storage device 14 for storing the abnormality judgment threshold value and the averaging time, and the means for the computer 10 averaging the acoustic power of the monitoring frequency and the means for judging the abnormality There is.

FIG. 7 shows the relationship between the duration of the abnormal sound to be detected and the disturbance sound to be removed and the overall value in this embodiment. As the abnormal sound, there are a surging sound that is an impact sound, an air leak sound that is a continuous sound, etc., and it is considered that an unexpected abnormal sound is generated. The disturbance sound includes a metal dropping sound which is an impact sound, a crane traveling sound which is a continuous sound, and a paging sound which is an intermittent sound. In this embodiment, the sound power for each 1/3 octave band is used instead of the sound power for a single monitoring frequency. The sound power for each 1/3 octave band is obtained by integrating the sound power spectrum for a section having a certain width around the frequency given by the geometric progression of 1/3 of 2 For details of filter characteristics, etc., see JIS C15131.
983. The 1/3 octave center frequency to be monitored is 16 from 250Hz to 8kHz.
It is a band and uses a plurality of bands in the high frequency band of audible frequencies. This is because, as shown in FIG. 8, the air leak sound has high acoustic power in the high frequency band in the audible range, and the abnormality monitoring is performed by a plurality of bands in order to detect an unexpected abnormal sound. In addition to the sound power for each 1/3 octave band, the overall value is also monitored to detect surging. Further, the normal sound of the gas turbine varies greatly depending on its operating condition.

Therefore, the load signal of the generator is used as the process data, and the plant operating region is divided into five in this example as shown in FIG. Here, the generator load is taken as a parameter of the operating region. By dividing in this way, the fluctuation range of the normal sound in one monitoring area becomes small, and the sound power fluctuation range m at the time of determining the abnormality determination threshold can be made small. As a result, the abnormality detectable area in FIG. 4 is expanded, and the abnormal sound can be detected with high sensitivity.

3.1 Processing Procedure in Normal Sound Analysis Mode During Preliminary Experiment FIG. 1 shows the processing procedure in normal sound analysis mode during preliminary experiment.
This will be described with reference to the flowchart of FIG.

In this process, first, the number of measurements k is set to 0 (step 1001), the acoustic signal of each sound is taken in by the microphone 11 (step 1002), and A /
The signal captured by the D converter 12 is converted into a digital signal (step 1003), the digital signal is captured by the DSP 13 (step 1004), and the fast Fourier transform process (FFT process) is performed (step 1005), and then the 1/3 octave analysis is performed. Then, the acoustic power for each monitoring frequency band is calculated (step 1006) and the overall value is calculated (step 1007). At this time, the load signal of the generator is taken in by the computer 10, and the acoustic power and overall value for each monitoring frequency band are stored in the storage device 14 as Octs (i, j) k (step 1008). Here, i is 5 in each operating region, j is 17 in each frequency band and overall value (overall value is i = 1), and k is the number of measurements. This process is repeated n times of measurement (step 1009 → step 1002).

Next, the Octs stored in the storage device 14 is stored.
(I, j) k is loaded into the computer 10 (step 1010), and the operating region and the beginning of the frequency band (i =
j = 1) to (step 1011), the computer 10
Then, statistical processing of a normal sound is performed based on the Octs (i, j) k to calculate an average value α and a standard deviation Sd (step 1012). Based on the result, the standard deviation of the standard deviation is calculated as shown in FIG. The quadrupling is determined as the sound power fluctuation width m (step 1013). However, the sound power fluctuation range m and the average value α are 8 in total for each frequency band (16 bands) and overall value for each operating region (5 regions).
It is necessary to determine five (step 1014 → 101)
2, step 1015 → step 1012). And
Eighty-five sound power fluctuation range m (i, j) and average value α
Each of (i, j) is stored in the storage device 14 (step 1016).

Although four times the standard deviation is determined as the sound power fluctuation width m here, the value of four times the standard deviation can be appropriately changed according to the detection state and the detection target. Yes, practically 4 times standard deviation 3 dB
The value obtained by adding is expected as the maximum value of the acoustic fluctuation width m.

3.2 Disturbance Sound Analysis Mode in Preliminary Experiment The signal processing procedure in the disturbance sound analysis mode in the preliminary experiment will be described with reference to the flowchart of FIG.

In this process, the number of times of measurement k is set to 0 (step 1101), and a metal dropping noise when the gas turbine is stopped,
Disturbances such as paging and crane traveling sounds are generated, the acoustic signal of each sound is captured by the microphone 11 (step 1102), and converted into a digital signal by the A / D converter 12 (step 1103), and D
At SP13, the digital signal is fetched and FFT processed (step 1104), and then 1/3 octave analysis is performed to calculate the acoustic power and overall value for each monitoring frequency band (steps 1105 and 1106), and Octg.
(J) k (j: overall and each frequency band,
k: number of times of measurement) is stored in the storage device 14 (step 1107), and this process is repeated n times of number of times of measurement (step 1108 → step 1102). Then, the storage device 1
The Octg (j) k stored in No. 4 is fetched into the computer 10 (step 1109), and the computer 10 causes the disturbance for each monitored frequency band from the beginning of the operating region and the frequency band based on the fetched Octg (j) k. Find the sound power and duration of a sound.

Specifically, for each j (frequency band and overall value), the threshold is set to sl, and Octg
(J) When k changes k, the number of times c that continuously exceeds the threshold value sl is obtained, and a value obtained by multiplying the measurement interval time dt by c is taken as the duration Ct (j) l, and this series of threshold values is exceeded. Oct
The average value of g (j) k is defined as the acoustic power G (j) l. Here, l represents each disturbance sound (steps 1110 to 1110).
1120). Next, the minimum product of β (j) (Ct (j) l and G (j) l) in the detection limit curve given by equation (7) in which the disturbance noise does not enter the detectable region (equation (8)). Of the disturbance sound Oc for the overall value.
The maximum value of tg (l) k is calculated and set as β (l) (steps 1121 to 1127). Then, β (j) (j = 1 to
17) is stored in the storage device 14.

3.3 Judgment Parameter Determination Mode During Preliminary Experiment The signal processing procedure in the judgment parameter determination mode as the final stage of the preliminary experiment will be described with reference to the flowchart of FIG.

16 1/3 octave bands (j = 2
.About.17), the sound power fluctuation range m (i, j) and the parameter β (j) of the detection limit curve are loaded from the storage device 14 into the computer 10 (step 12).
01), the arithmetic mean time t (i, j) is determined using the equation (6) according to the corresponding set of j (step 120).
5). For the overall value (j = 1), β
(1) and m (i, 1) are compared, and the larger one is new m
As (i, 1) (steps 1203, 1204, 12
06, 1207) and store them in the storage device 14 (step 1208).

3.4 Processing at Abnormality Monitoring Next, the processing procedure at the time of actual abnormality monitoring is shown in FIG.
This will be described with reference to the flowchart of FIG.

In this process, first, the acoustic signal of the equipment to be monitored is detected by the microphone 11 (step 1
301), the A / D converter 12 converts the digital signal (step 1302), and the converted digital signal is converted to DP.
After taking in S13 (step 1303) and performing FFT processing (step 1304), 1/3 octave analysis is performed to calculate the overall value and the acoustic power for each monitoring frequency band (steps 1305 and 1310). 1
Regarding the / 3 octave, as described above, the operating area is determined (determined i) from the generator load and stored in the storage device 14, and the acoustic power is added during a certain average time t (i, j). To average. Since the addition time varies for each frequency band, the addition and averaging is stopped when the addition and averaging time reaches each frequency band (steps 1306 and 13).
07, 1308). Next, the acoustic power after the averaging and the generator load are loaded into the computer 10, the operating region is determined (i is determined), and the result is stored in the storage device 14. Then, a normal sound average value α (i, 1) and sound power fluctuation width m (i, j) are taken in, and the normal sound average value α (i, 1) and sound power fluctuation width m (i, j) are summed. Is set as the abnormality determination threshold value and the sound power of the frequency band exceeds even one abnormality determination threshold value, it is determined to be abnormal (steps 1309 and 131).
2) If it does not exceed, it is determined to be normal (step 13).
09, 1313).

As for the overall value, the overall value without addition averaging calculated by the computer 13 in the DSP 13 and the generator load are taken in, the operating region is determined (i is determined), and the normal sound average stored in the storage device 601 is determined. The value α (i, 1) and the sound power fluctuation range m (i,
j) is taken in, and the sum of the normal sound average value α (i, 1) and the sound power fluctuation range m (i, j) is used as the abnormality determination threshold value, and when the overall value exceeds the abnormality determination threshold value, it is determined as abnormal (step 1311, 1312), and if not exceeded, it is determined to be normal (steps 1311, 1313).

3.5 Others In this embodiment, the processing is performed based on the hardware configuration shown in FIG. 6, but the configuration is not limited to that shown in FIG. As shown in, an analog filter 16 is provided instead of the DSP 13,
The output from the analog filter 16 is an A / D converter 12
It is also possible to configure so as to input and output from the computer 10 or the storage device 14 to calculate and monitor the acoustic power.

Further, in this embodiment, the acoustic power fluctuation range m is set to four times the standard deviation as described above, but the present invention is not limited to this, and the standard range is different for each operation band and frequency band. It is also possible to determine using a deviation coefficient or the like, and it may be determined based on the maximum sound power without using the standard deviation.

Further, although the sound power of the 1/3 octave band is used in the present embodiment, it is not limited to this, and for example, the 1/1 octave analysis or any octave band, and further, other than the octave band is used. It is also possible to use acoustic power in any frequency band.

Further, in this embodiment, the preliminary experiment and the abnormality monitoring device are set to be performed by using the same device, but the device for performing the preliminary experiment and the actual abnormality monitoring are different devices, that is, It may be configured such that the abnormality detection parameter is determined by being executed by another device including the sound detection unit, the sound signal processing unit, and the frequency analysis device, and is input to the present device.

Further, although the averaging time is used in this embodiment, each processing can be executed by using the averaging count instead of the averaging time.

Further, in the above-described embodiment, the acoustic signal is detected and various processes are executed to detect the abnormality of the device. However, in addition to this, if the detection target is different, Depending on the detection target, for example, acceleration, vibration,
Displacement, temperature sensor, etc. are used to obtain the characteristic amounts of various parameters detected by these sensors, and based on the characteristic amounts, the same processing as that of the above-described embodiment is performed to monitor the device abnormality. It is also possible to configure.

[0066]

As is apparent from the above description, according to the present invention, the process is divided into a preliminary process and a monitoring process, extracted in the preliminary process, a threshold value is set based on the processed information, and the monitoring process described above is performed. Since it determines whether or not an abnormal sound is generated from the monitored device based on a threshold value, when monitoring a plurality of devices or the like by using acoustics, disturbance noise should be removed and abnormal sound should be detected with high sensitivity. You can

[Brief description of drawings]

FIG. 1 is a flowchart showing a processing procedure of a device abnormality monitoring method of the present invention.

FIG. 2 is a block diagram showing a basic functional configuration of a device abnormality monitoring device of the present invention.

FIG. 3 is a diagram showing a relationship between sound power and duration of abnormal sound and disturbance sound generated in the device.

FIG. 4 is a diagram showing an anomaly detectable region and a detection limit curve in the relationship between sound power and duration.

FIG. 5 is a diagram showing an abnormality detectable region and a detection limit curve in the relationship between the overall value and the duration.

FIG. 6 is a block diagram showing a hardware configuration of an embodiment in which the present invention is applied to a thermal power plant.

FIG. 7 is a diagram showing a relationship between sound power and duration of abnormal sound and disturbance sound generated in a gas turbine.

FIG. 8 is a diagram showing a 1/3 octave analysis result of an air leak sound.

FIG. 9 is a diagram showing a state in which an operating region is divided according to a generator load.

FIG. 10 is a flowchart showing a processing procedure in a normal sound analysis mode during a preliminary experiment in the present embodiment.

FIG. 11 is a flowchart showing a processing procedure in a disturbance sound analysis mode during a preliminary experiment in the present embodiment.

FIG. 12 is a flowchart showing a processing procedure of a determination parameter determination mode at the time of a preliminary experiment in the present embodiment.

FIG. 13 is a flowchart showing a processing procedure at the time of abnormality monitoring in this embodiment.

FIG. 14 is an explanatory diagram showing an abnormality determination threshold value determining method according to the present embodiment.

FIG. 15 is a block diagram showing a hardware configuration of another embodiment using an analog filter.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Acoustic detector 2 Acoustic signal processor 3 Judgment parameter calculator 4 Addition and averaging processor 5 Abnormality judgment unit 6 Judgment parameter storage 7 Man-machine interface 10 Computer 11 Microphone 12 A / D converter 13 DSP 14 Storage device 15 Man-machine Interface 16 Analog filter

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Minoru Yanagibashi 3-2-2 Sachimachi, Hitachi, Ibaraki Prefecture Hitachi Engineering Service Co., Ltd. (72) Inventor Michio Nakayama 3-2, Sachimachi, Hitachi, Ibaraki No. 2 In Hitachi Engineering Services Co., Ltd. (72) Inventor Hideo Nakamura 3-1-1 1-1 Saiwaicho, Hitachi City, Ibaraki Hitachi Co., Ltd. Hitachi Factory

Claims (11)

[Claims] 1. A device abnormality monitoring method for acoustically monitoring an abnormality during operation of a monitored device, comprising: a first step of obtaining an average value of acoustic power at a monitoring frequency of a normal operation sound in advance; Second, based on the threshold value to which the power fluctuation range is added, the sound power fluctuation range is determined so that the normal driving sound is not erroneously detected as an abnormal sound.
A third step of selecting a disturbance sound whose product of the duration and the sound power gives a maximum value from a plurality of disturbance sounds to be removed which are defined by the duration and the sound power of the monitoring frequency, and A preliminary step including a fourth step of determining a value obtained by dividing the maximum value by the fluctuation range of the acoustic power as an averaging time of the acoustic power at the monitoring frequency; and A monitoring step of monitoring whether or not the value after arithmetic averaging exceeds the threshold during the arithmetic averaging time period, and an abnormal sound is generated from the monitored device depending on whether or not the value exceeds the threshold. Device abnormality monitoring method characterized by determining whether or not there is 2. A plurality of monitoring frequencies are set, and in the preliminary step, the first to fourth steps are performed for each monitoring frequency, and in the monitoring step, the acoustic power after the averaging is performed for each monitoring frequency. The device abnormality monitoring method according to claim 1, further comprising monitoring whether or not a preset threshold value is exceeded. 3. Instead of the acoustic power of the monitoring frequency,
3. An apparatus abnormality monitoring method according to claim 1, wherein an overall value which is the total energy of the acoustic signal is used. 4. Instead of the acoustic power of the monitoring frequency,
3. The device abnormality monitoring method according to claim 1, wherein an integrated value of sound power in a frequency section is used. 5. A device abnormality monitoring method for monitoring an abnormality during operation of a monitored device, wherein a sensor is arranged in the monitored device, a characteristic amount is extracted from a detection signal of the sensor, and the characteristic value is detected in advance during normal operation. The average value of the characteristic amount is obtained, and the variation width is determined by a threshold value obtained by adding the variation width of the characteristic amount to the average value so as not to be erroneously determined to be abnormal during normal operation, and is defined by the duration and the characteristic amount level. The disturbance signal that gives the maximum value of the product of the duration and the feature amount level is selected from the plurality of disturbance signals to be removed, and the value obtained by dividing the maximum value by the fluctuation range is the arithmetic mean of the feature amounts. A preliminary step of determining as a time, and a monitoring step of monitoring whether or not the value after arithmetic mean of the characteristic amount of the detection signal of the sensor for the arithmetic mean time exceeds the threshold, Cross over Apparatus abnormality monitoring method characterized by determining the presence or absence of abnormality of the monitored device by whether 6. A device abnormality monitoring apparatus for acoustically monitoring an abnormality of a monitored device, means for detecting an acoustic signal of the monitored device, and means for calculating acoustic power from the acoustic signal detected by the detecting means. Means for determining an abnormality determination threshold value and arithmetic mean time from the sound power calculated by the calculating means, means for storing the abnormality determination threshold value and arithmetic mean time determined by the determining means, and means for storing this During the arithmetic mean time stored in, the means for arithmetic mean processing of the acoustic power of the monitoring frequency calculated by the calculating means, and the acoustic after the arithmetic mean calculated by the means for arithmetic mean processing Means for judging an abnormality from the power based on the abnormality judgment threshold value stored in the storing means. A device abnormality monitoring device characterized by: 7. The device abnormality monitoring device according to claim 6, wherein the calculating unit calculates the acoustic power for each of a plurality of monitoring frequencies from the acoustic signal detected by the detecting unit. 8. Instead of the acoustic power of the monitoring frequency,
The device abnormality monitoring device according to claim 6 or 7, wherein an overall value that is the total energy of the acoustic signal is used. 9. Instead of the acoustic power of the monitoring frequency,
The device abnormality monitoring device according to claim 6 or 7, wherein an integrated value of sound power in a frequency section is used. 10. A device abnormality monitoring device for monitoring an abnormality of a monitored device during operation, a sensor arranged at a monitoring position of the monitored device, and means for extracting a characteristic amount from a detection signal of the sensor, The means for determining the abnormality determination threshold value and the arithmetic mean time from the feature amount detected by the extracting means, the means for storing the abnormality determination threshold value and the arithmetic mean time determined by the means for determining, and the storing means During the period of the stored averaging time, the feature amount extracted by the extracting unit is subjected to the averaging process, and the feature amount after the averaging calculated by the averaging process unit is stored. An apparatus abnormality monitoring device comprising: means for determining an abnormality based on an abnormality determination threshold value stored in the means. 11. The man-machine interface means for changing the processing contents of the determining means and the determining means according to the judgment of an observer is further provided. The device abnormality monitoring device described.