JP2000214052A - Abnormal sound detection system and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an abnormal sound detection system which can detect an abnormal sound generated from a facility surely and quickly even in a noisy production field. SOLUTION: A sound from a non-directional microphone 12 is analyzed by means of an FFT analyzer 21 in order to analyze ambient sound and a sound from a mono-directional microphone 13 is subjected to wavelet analysis thus providing information about the magnitude of various period signals at each time position. According to the method, an abnormal sound can be detected surely.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for effectively controlling the occurrence of abnormal sound in a production line due to an abnormality in the operation of a device in a production facility or an abnormality in a product. The present invention relates to an abnormal sound detection system and a recording medium that detect quickly.

[0002]

2. Description of the Related Art Conventionally, generally, in a building of a production factory, a large number of production facilities are operated independently or in conjunction with each other, and each production facility emits a unique noise. Specifically, such as machine vibration noise and steam outflow noise,
The types vary. In addition, noise generated irregularly in time, such as the operation of a forklift, is added around the production facility.

[0003] Further, when a work carrier is traveling or when a press machine is operating,
To prevent danger, a horn may sound or a simple music melody may be played to alert the surroundings that they are actually driving and to alert them. . As described above, in general, at a production site, a plurality of noises generated continuously or intermittently from a plurality of noise sources, and further, a plurality of noises generated irregularly in time are added, so that the noise is complicated. Overlapping, a unique noise environment is formed.

In such a complicated noise environment,
The monitoring worker patrols the operation status of the production equipment in a timely manner,
At present, the presence of abnormalities is monitored mainly by eyes and ears. In this case, for the operation management items that can be measured every moment, such as temperature, pressure, and transport speed, for example, the presence or absence of occurrence of an abnormality is monitored and observed by observing a measuring instrument monitor attached to a control panel. Things can be done relatively easily.

[0005] However, the abnormal sound emitted from the production equipment must be directly visited by a monitoring worker near the equipment, and must be constantly or frequently observed.
I have to say that its discovery is extremely difficult. Also, the skill of detecting abnormal sounds requires considerable skill. When an abnormal sound can be confirmed at a distance, in many cases, equipment destruction or work damage has already considerably progressed in many cases, which may lead to a serious situation.

[0006] As described above, it is difficult to find abnormal sounds.
As mentioned earlier, this is probably due to the extremely complicated noise environment, but that is not to say.
In order to prevent such a serious situation from occurring, it is extremely important to detect an abnormal sound emitted from a production facility as early as possible in an early stage when the abnormality occurs.

However, because of this, the task of a skilled observer patroling and monitoring a series of long-range production facilities around the clock is too burdensome for the observer. In addition, since a large number of skilled observers are required to continuously monitor a plurality of parts of the production equipment, such a method cannot be a practical measure. That is, a method for finding an abnormality that requires relying on a monitoring worker in the first place cannot be an appropriate method in a production site in a very noisy environment.

[0008] The present invention specifically provides a method for determining the cause of abnormal noise at a production site (a problem relating to the identification of an abnormal noise source), and the origin of the abnormal noise ( Problem related to the search for an abnormal noise source) and an abnormal sound detection system constructed to reliably perform the above. Generally speaking, a mechanical structure vibrates in a complicated manner during operation due to its structural characteristics, mass, rigidity, and the like. Mode). Such a phenomenon is consistent with the idea of "Fourier series" that a complex time signal waveform is given by combining a sine wave and a cosine wave whose amplitude and phase are appropriately adjusted.

In the present invention, the main purpose is not to directly perform vibration analysis in each part of the production equipment individually, but to comprehensively perform noise analysis (so-called sound field analysis) at the production site. Aims to do.
Of course, it is also possible to construct a detection system so that vibration analysis is performed for each individual facility. However, when there are a large number of equipment such as a continuous production line, the equipment cost of the detection equipment and the like is inevitably increased, and as a result, the production cost is greatly reduced. Become.

Therefore, the inventor of the present invention has found that a monitoring operator directly discovers an abnormality in a facility mainly by his eyes and ears, and especially by ears. In some cases, an abnormal sound emitted from a non-existent part can be identified (such as when an abnormality occurs in the equipment box). Focusing on the fact that the abnormal sound can sometimes be identified, we considered performing noise analysis (sound field analysis) at the production site.

In this case, a microphone is used as an acoustic sensor in place of the ear. In order to perform more accurate noise analysis than with the ear, specifically, the search for an abnormal sound source and the cause of the In order to be able to identify whether the noise data is due to the above, a computer data processing system for processing the noise data sampled by a computer is constructed.

The ear is an excellent acoustic sensor, and a skilled person may be able to identify a predetermined noise from a fairly complex noise mixed with all kinds of noises. It must be said that it is extremely difficult to detect a relatively small abnormal sound emitted in the initial stage of the process. Therefore, the present inventor has considered performing a waveform analysis of the noise in a comprehensive noise environment in which all noises are superimposed.

Here, the frequency spectrum indicates the frequency of many sine waves having different frequencies contained in the waveform and their amplitudes, and can be obtained by Fourier transform. In that case, specifically, an FFT analyzer is used. The FFT analyzer has a feature that frequency analysis can be performed with high resolution and analysis can be performed at once over a wide range of frequencies. In addition, since information on the phase included in the signal is also obtained, the information has the characteristic that the information included in the input signal can be handled without omission using the three elements of the signal, namely, frequency, amplitude, and phase.

FIG. 12 is a diagram showing a configuration example of a conventional FFT analyzer. The sound is captured by the microphone 51, converted into an electric signal, and input to the FFT analyzer 52. In the FFT analyzer 52, an analog signal is passed through a low-pass filter 53 that cuts off unnecessary high-frequency components, and then subjected to A / D conversion by an A / D converter 54 for digital signal processing.
, And a discrete Fourier transform (D
FT) to display the image on the CRT display 56 and print the image on the plotter 57. In the FFT analyzer 52, the input sample value data is cut out by a window function and discrete Fourier transform (D
FT). [Definition formula 1]

[0015]

X (kΔf) = Σx (iΔt) exp (−j2πik / N) ‥‥‥ (1) x (iΔt) is obtained by converting the time axis signal to Δ (T = NΔt) seconds.
It is an i-th waveform sampled at N points at a sampling period of t seconds. X (kΔf) is a frequency spectrum, and a periodic signal x (iΔt) having a period of T seconds is represented by Δf (= 1 /
T) and a coefficient when it is regarded as a composite waveform of a cosine wave and a sine wave having a frequency that is an integral multiple of T). Also, the inverse D
FT (IDFT) is based on the following [Definition formula 2]. [Definition formula 2]

[0016]

X (iΔt) = 1 / NΣX (kΔf) exp (j2πik / N) ‥‥‥ (2) Note that the above [Definition Formula 1] and [Definition Formula 2] are calculated using the fast Fourier transform (FFT). ) Can dramatically reduce the amount of calculation. By the way, the waveform of the sound (or vibration) which is focused on in the present invention can be described by the three elements of the signal. Specifically, as described above, the time waveform of the noise (analog signal) ) Can be converted to a frequency spectrum by performing an FFT on it. Also, on the contrary,
By applying the inverse FFT to the frequency spectrum, a “correlation function” expressing the similarity between two data strings is obtained (→
Winner-Hinchin theorem).

The correlation function includes the following [Definition formula 3],
"Autocorrelation function Rxx (τ)" represented by [Definition formula 4]
There are two types of functions called “cross-correlation function Rxy (τ)”.
The autocorrelation function represents the similarity to itself which is obtained by adding a time difference of delay time τ to x (t) which is a time function, and the cross-correlation function is two time functions x added with the time difference of delay time τ.
It represents the similarity between (t) and y (t). [Definition formula 3]

[0018]

(Equation 3)

[Definition formula 4]

[0020]

(Equation 4)

For example, if the time function x (t) is periodic, the value of the autocorrelation function becomes large when the time is shifted by the period, and the time function x (t) is irregular. Then, the value becomes maximum when τ = 0, and becomes 0 in other cases. Using this property, the period of the waveform buried in the noise can be checked.

FIG. 13 is a diagram showing a waveform example of a noise signal. When the production equipment is operating in a steady state, the noise generated from it (including periodically generated alarm sounds) is probably periodic (FIG. 13 (a)). Under such circumstances, the forklift operation noise etc. uttered completely independent of the production equipment,
It is a noise that can be predicted to some extent (that is, noise having a specific waveform) and can be handled separately (see Fig. 13 (b)). Only the noise (abnormal sound) caused by the abnormality is irregular noise accompanied by a transient state change on the time axis (FIG. 13).
In view of (c), the autocorrelation function may be used as a means for finding abnormal sounds.

In the FFT analyzer, an inverse FFT (IF) of a power spectrum (also called a self spectrum) is used.
By performing FT), an autocorrelation function is obtained. Its value is -1 to +1. In addition, since the autocorrelation function is a mean square value of the time function x (t), as is apparent from the above [Definition formula 3],
This represents the signal power. So, specifically,
A time waveform of the input noise is cut out at predetermined time intervals, and a power spectrum is obtained by performing FFT.
An autocorrelation function is obtained by performing FFT, and when the value of Rxx (0) changes transiently on the time axis, it can be determined that an abnormal sound may be occurring. Thus, F
Regarding the time waveform of the noise input to the FT analyzer,
By obtaining the autocorrelation function, it is possible to detect the occurrence of abnormal sound to some extent.

However, simply obtaining the autocorrelation function does not enable accurate detection when the abnormal sound emitted from the production equipment is small. This is due to the DFT (Discrete Fourier Transform) processing itself, in addition to the problem of the sound receiving sensitivity of the microphone. That is, in the DFT, the frequency analysis by the Fourier transform is completed in a finite calculation time (within a time window).
The time information of the signal may be lost. Therefore, a method of short-time Fourier transform that increases the time resolution without lowering the frequency resolution by increasing the apparent time window length and performing the Fourier transform is also conceivable.

[0025]

However, this short-time Fourier transform method is also not suitable for detecting abnormal sounds in a complicated noise environment because it is necessary to select an optimum window function. That is, the waveform is distorted by cutting at the time window length, the peak of the spectrum is lowered,
There is a problem that power leaks to both sides by that much, and there is a possibility that small peaks and the like are hidden. SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has an abnormal sound detection system and recording method capable of reliably and promptly detecting an abnormal sound generated from equipment even in a production site where noise is remarkable. It is intended to provide a medium.

[0026]

An abnormal sound detecting system according to the present invention comprises: a microphone for converting ambient sound into an electric signal; a sound analyzing means for detecting a time change of a frequency distribution of a sound signal from the microphone; Storage means for storing a time change of the frequency distribution of the sound signal in advance; and an abnormal sound for detecting an abnormal sound by comparing the time change of the frequency distribution of the current sound signal with the time change stored in the storage means. And a detecting means.

Further, since the sound analysis means performs a wavelet analysis of the sound signal, the sound can be analyzed by a simple calculation. Furthermore, by providing position detection means for detecting a sound generation position by cross-correlating sound signals from a plurality of microphones, the sound generation position can be accurately detected with a simple configuration. Further, since the position detecting means compares the amplitudes of sound signals from a plurality of microphones, it is possible to improve the detection accuracy of a sound generation position.

Further, the microphone includes a first microphone having a sharper directivity and a second microphone having a lower directivity than the first microphone, and the sound analyzing means includes a sound signal from the first microphone. After the sound signal from the second microphone is canceled out, the time change of the frequency distribution of the sound signal is detected, so that an abnormal sound that needs to be detected by excluding the sound signal from a source other than the production equipment is excluded. Only can be detected. Further, the present invention is a computer-readable recording medium in which a program for causing a computer to function as a sound analysis unit, a storage unit, and an abnormal sound detection unit of the abnormal sound detection system is recorded.

[0029]

BEST MODE FOR CARRYING OUT THE INVENTION When observing the characteristics of mechanical vibration and sound, vibration and sound generated in a steady state are
Vibration and sound generated in a transient state, such as sudden vibration and abnormal sound, are mixed with each other over time, and appear as unsteady signals as a whole. It needs to be considered with great care.
In order to handle such an unsteady signal, a time change (time-frequency distribution) of a frequency spectrum is obtained.

Since the signal can be considered to be locally fluctuating locally, its frequency can be considered to change with time. Therefore, capturing this locally periodic signal in the transition of time, that is, time-frequency analysis for capturing the signal from both time and frequency becomes important. For this purpose, the present embodiment uses a method called wavelet analysis. Here, the wavelet is based on a “mother wavelet” (corresponding to a sine function and a cosine function of DFT) which is a “maternal body”. Provide information.

By performing such a wavelet analysis, it is possible to simultaneously capture temporal and spatial transitions of fluctuations while making use of the features of Fourier analysis for expressing a signal in the frequency domain. The wavelet transform is defined by the following [Definition formula 5], and the inverse wavelet transform is defined by the following [Definition formula 6]. [Definition formula 5] The wavelet transform of the function f (x) by the mother wavelet ψ (x) is

[0032]

(Equation 5)

[Definition formula 6] The inverse wavelet transform is

[0034]

(Equation 6)

In the present embodiment, the autocorrelation function described above is not used as a judging means for finally detecting abnormal sound, but is merely a means for grasping a certain level of surrounding noise state at a certain measurement point. Use as And
For the final detection of the abnormal sound, the above-described wavelet analysis is used. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a diagram showing a microphone installation position in a production plant to which the abnormal sound detection system according to one embodiment of the present invention is applied. In the example of FIG.
The power plant installation area 2 such as a motor or a fan is provided on one side of the production plant 1, and the microphones 3 are installed at appropriate positions on the power facility installation area 2 side. In the example of FIG. 1B, the power equipment installation areas 2 are located on both sides of the production plant 1, and the microphones 3 are installed at appropriate positions on both sides of the production plant 1. In the example of FIG. 1 (c), when there is a noise source 4 such as a steam outlet, the microphone 3
Is to be installed. In this case, steady noise can be reliably captured. However, it is not necessary and necessary to install the equipment right next to all the noise generating places, since this imposes a heavy burden on facilities. Regarding the distance between the microphones, it is assumed that a plurality of workers take a distance that allows each person to sufficiently identify the same noise when standing at the location where each microphone is installed. Try not to be too far away that you can't.

FIG. 2 is a diagram showing an example of the arrangement of two types of microphones 3. An omnidirectional microphone 12 and a unidirectional microphone 13 are individually installed at microphone installation locations beside a noise source 11 in the production plant 1 and connected to a noise waveform analyzer 14. As shown in FIG. 2 (b), the directional characteristics of the omnidirectional microphone 12 equally catch sounds from all directions. Unidirectional microphone 1
As shown in FIG. 2 (c), the directivity of No. 3 can particularly strongly capture sound from a specific direction. The unidirectional microphone 13 is arranged toward the noise source, and the omnidirectional microphone 12 is arranged above the unidirectional microphone 13.

FIG. 3 is a diagram showing a configuration example of the noise waveform analyzer 14. The noise waveform analyzer 14 includes an FFT analyzer 21 (shown in FIG. 12), a # 1 memory 22 for recording the measurement result, a wavelet analyzer 23, and a # 2 memory 24 for recording the measurement result. And a measurement controller 25 for controlling the measurement instruments.

FIG. 4 is a diagram showing an example of an overall system for detecting an abnormal sound. Omnidirectional microphones 12A, 12B,
12C and unidirectional microphones 13A, 13B, 13
C connected to each noise waveform analyzer 14A, 14B, 14
C is a GP-IB general-purpose interface bus (IEEE-
488 standard) or the LAN 31 to the host computer 32. The host computer 32 has a memory 33 and a timer 34, is connected with a keyboard 35 for inputting various data and a display 36 for displaying a result of abnormal sound detection and the like, and each of the noise waveform analyzers 14A, 14A.
The measurement operation of 4B and 14C is controlled. The unidirectional microphone 13 is generally arranged to grasp the situation of noise emitted from the noise utterance source side, and the non-directive microphone 12 is arranged to grasp the surrounding noise situation of the unidirectional microphone 13. Are located in

The input from the omnidirectional microphone 12 is
The time change of the noise (time signal) in the steady state in which the production plant 1 is operating normally is grasped by the autocorrelation function described above. In this case, noise sources that are not directly related to production, such as an alarm sound that is periodically emitted and a forklift operation noise that is emitted irregularly, are also considered. The input from the unidirectional microphone 13 is
Used to directly detect abnormal noise that is transiently generated in an emergency and identify the cause of the abnormal noise while directly measuring and monitoring the noise situation in the steady state of the noise source 11 such as the power source involved in production. Is done. In this case, consideration is given by canceling out the influence of the surrounding noise situation due to the input analysis result from the omnidirectional microphone 12.

FIG. 5 is a diagram schematically showing an arrangement of the omnidirectional microphone 12 in the diffuse sound field. Even if the occurrence of an abnormal sound can be detected by performing a wavelet analysis on the input from the unidirectional microphone 13, it is not possible to specify in which direction the detected sound actually came from. That is, since the sound is diffused and transmitted in the space, each microphone 3 arranged at an appropriate position on the side of the production plant 1 as shown in FIG. 1 has a distance from an abnormal sound source (position unknown). In response to the,
The same abnormal sound waves having different sound pressure levels should have arrived.

Therefore, as shown in FIG. 5, a diffused sound field centered on the abnormal sound source 41 (a sound field that can be assumed to have the same sound energy in any direction passing through any one point). , Reflection, etc.) are assumed as concentric hemispherical spaces (the shape of the production equipment is ignored), and each omnidirectional microphone 12A,
The sound pressure levels of the abnormal sound signals input to 12B and 12C are relatively compared.

Specifically, in the measurement controller 25,
An abnormal sound waveform is extracted from the result of the FFT operation based on the result of the wavelet analysis, and a representative sound pressure level I is obtained. This value depends on the distance d from the abnormal sound source 41 to each of the omnidirectional microphones 12A, 12B, and 12C. For example, since it is a hemispherical diffusion, I =
W / 2πd ² (W is the sound source output), and the sound pressure level at each sound receiving point is a function of the radial distance d.
However, in this case, it is necessary to pay attention to the fact that an abnormal sound is delayed due to the distance d. In the present embodiment, when searching for a sound source of an abnormal sound, the distance d is corrected using the cross-correlation function described above.

The cross-correlation function is generally used for measuring the delay time of the transmission system, and as is clear from the above-mentioned [Definition formula 4], the time functions x (t) and y
The average value of the product of (t) is displayed, and the value is -1 to +1. And in the FFT analyzer,
A cross-correlation function is obtained by inverse FFT (IFFT) of the cross spectrum. That is, the power spectrum of each of the input signal and the output signal is calculated and obtained by FFT, and the power spectrum is multiplied to obtain an average, thereby obtaining a mutual spectrum.

In the case of the present embodiment, the power spectrum of each of the noises input to the three omnidirectional microphones 12A, 12B, and 12C installed at adjacent positions as the above-mentioned two signals used as the calculation bases is shown. Are obtained by FFT, and they are multiplied to obtain an average, thereby obtaining a cross spectrum. Further, by applying an inverse FFT thereto, a cross-correlation function is obtained.

FIG. 6 is a diagram for explaining a method of searching for an abnormal sound source. The noise having the highest sound pressure level for the abnormal sound should be input to the omnidirectional microphone 12 located closest to the source position S of the abnormal sound. Now
The position of the microphone is B. The positions of the two omnidirectional microphones 12 installed at adjacent positions are A and C.
Becomes The host computer 32 receives the sound pressure level data of the abnormal sound analyzed by each measurement controller 25 and specifies each omnidirectional microphone 12 at the positions B, A, and C. According to the description of FIG. 6, the line segment AB =
a, the line segment BC = b, and the angle ∠ABC = θ are all known fixed values, and since the sound pressure level I is a function of the distance d, I = f (d) ‥‥‥ (7) Conversely, the distance d is a function of the sound pressure level I.

D = g (I) ‥‥‥ (8) Therefore, the distance d is obtained from the sound pressure level at each sound receiving point.
The relative ratio of 1, d2 and d3 is obtained. _{d1 / d2 = g (I A} ) / g (I B) = α ‥‥‥ (9) d3 / d2 = g (I C) / g (I B) = β ‥‥‥ (10) Further, the adjacent From the cross-correlation function of the sound signal from the microphone 12, a distance difference from the position S of the abnormal sound is determined. d1 = d2 + Δd1 ‥‥‥ (11) d3 = d2 + Δd2 ‥‥‥ (12) By solving these equations (9) to (12) as equations, d1, d2, and d3 are obtained.

Therefore, as shown in FIG. 7, in the xyz coordinate space, the sound receiving point position B is set as the origin, and two sound receiving points A and C are taken on the xy plane, and the position S is obtained by calculation. FIG. 8 is a flowchart illustrating the control operation of the host computer 32. First, it is determined whether or not the system power is ON (step S1). If NO, the process waits until the system power is turned ON. If YES, initial settings such as data including communication settings with each measurement controller 25 are performed (step S2).

Next, in step S3, an abnormal sound detection preparation work is performed (detailed in FIG. 9), and a report of abnormal sound detection such as a sound pressure level of abnormal sound and a waveform analysis result is sent from any of the measurement controllers 25. It is determined whether or not there is (step S4)
If there is no report of abnormal sound detection in NO, the operation waits as it is,
If there is a report of abnormal sound detection in YES, the measurement controller 25 of the noise waveform analyzer 14 arranged on the left and right of the measurement controller 25 for which the report is made is specified among the measurement controllers 25 to which the address is set. (Step S
5) Instruct the two specified measurement controllers 25 to report the abnormal sound pressure level at the time of the abnormal sound detection report (step S6). Then, it is determined whether or not there is a report from each of the designated measurement controllers 25 (step S7), and if there is no report, the process waits until a report is received. Is identified, and based on the abnormal sound pressure level data at each sound receiving point, a search for an abnormal sound source is executed (step S8), and the position of the abnormal sound source and the waveform analysis result are displayed on the display 3.
6 is displayed (step S9). Then, it is determined whether or not to stop the system operation (step S10), and if the operation is still to be continued with NO, the process returns to step S3, and Y
If the operation is to be stopped at the ES, a system operation stop process is performed (step S11), and the flow ends.

FIG. 9 is a flowchart illustrating the detailed operation of the abnormal sound detection preparation work (step S3) according to the present embodiment. First, after the timer 34 is started in step S21, the power spectrum data is reported to each measurement controller 25 in order of arrangement (step S2).
2) It is determined whether data from two adjacent measurement controllers 25 has been reported (step S23), and NO
If it has not been reported yet, the process waits. If YES is reported, a correlation function between the data from the two measurement controllers 25 is obtained (step S24). Based on the obtained correlation function, the distance between each sound receiving point is corrected, that is,
Considering the influence of the actual acoustic environment of the production facility, the propagation characteristics of the sound around each sound receiving point are calculated based on the phase shift of the correlation function using the noise power spectrum at each of the two adjacent sound receiving points. The difference is obtained, and a correction coefficient is determined according to the actual distance difference. Then, the distance correction value is recorded in the memory 33. This distance correction value is used later for searching for an abnormal sound source. In order to update the correlation function data, it is determined whether or not a predetermined time has elapsed (step S27). If NO, the process waits. If YES, the process returns to step S4.

FIGS. 10 and 11 show the abnormal waveform analyzer 14.
FIG. 4 is a flowchart illustrating a control operation of measurement controller 25 of FIG. First, it is determined whether or not the system power is ON (step S31). If NO, the process waits until the system power is turned ON. If YES, the host computer 32
Initial settings such as data including communication settings with the server (step S32).

Next, in step S33, measurement by the FFT analyzer 21 is started, and the FFT analyzer 21 analyzes the noise situation during the steady operation of the production equipment (step S34).
The noise situation at the time of steady operation and the noise situation when a forklift operation sound or the like is added are distinguished and recorded in the # 1 memory 22 (step S35). Then, it is determined whether there is an instruction to report power spectrum data from the host computer 32 (step S36). If NO, the instruction is awaited if there is no report instruction. If YES, the data is reported (step S36). S37), it is determined whether or not the report has been received by the host computer 32 (step S38). If NO, the reception is not completed, and if YES, the reception is completed, the process proceeds to FIG.
This time, an autocorrelation function is obtained in real time (step S39), and the presence or absence of a certain abnormal sound is determined (step S40).
If it is determined in S that there is a possibility of abnormal sound, it is determined whether or not it is from a source other than the production equipment (step S41).
If the answer is YES, the monitoring is continued if it is not from the production equipment, and if the answer is NO, it is from within the production equipment.
Step S33 executed simultaneously with step S33 of 0
The analysis result is recorded in the # 2 memory 24 by the start of the measurement by the wavelet analyzer 23 of step 42 (step S
43), the noise data during the steady operation is stored in the # 2 memory 24;
(Step S44), it is determined whether or not an abnormal sound is detected by the wavelet analyzer 23 (step S45).
Returning to 39, the monitoring is continued, and if it is determined that the sound is abnormal, the abnormal sound waveform is extracted, the representative sound pressure level is obtained (step S46), and the sound pressure level of the abnormal sound and the waveform analysis are performed. The result is reported to the host computer 32 (step S47), and it is determined whether or not the report has been received (step S48). If NO, the process waits. If YES, the system operation is stopped. It is determined whether or not the operation is continued (step S49). If the operation is continued with NO, the process returns to step S33 and step S42 to continue. If the operation is stopped with YES, a system operation stop process is performed (step S50). ), End the flow.

The present invention is not limited to the above embodiment. The present invention may be a computer-readable recording medium on which a program for causing a computer to function as a sound analysis unit, a storage unit, and an abnormal sound detection unit of the abnormal sound detection system is recorded. -Any type of recording medium such as a ROM, an IC card, a RAM card and the like may be used.

[0054]

As described above, according to the present invention, it is possible to prevent the damage from being escalated due to any trouble occurring in a specific production facility during the operation of the production line. It is possible to reliably and promptly obtain information that can be extremely useful in avoiding the danger that the operator poses to the operator.

[Brief description of the drawings]

FIG. 1 is a diagram showing an installation position of a microphone in a production plant to which an abnormal sound detection system according to an embodiment of the present invention is applied.

FIG. 2 is a diagram illustrating an arrangement example of two types of microphones according to the present embodiment.

FIG. 3 is a diagram illustrating a configuration example of a noise waveform analyzer according to the present embodiment.

FIG. 4 is a diagram showing an example of an overall system for detecting abnormal sound according to the present embodiment.

FIG. 5 is a diagram schematically illustrating an arrangement of omnidirectional microphones in a diffuse sound field.

FIG. 6 is a diagram (part 1) illustrating a method for searching for an abnormal sound source according to the present embodiment.

FIG. 7 is a diagram (part 2) illustrating the method for searching for an abnormal sound source according to the present embodiment.

FIG. 8 is a flowchart illustrating a control operation of the host computer of the present embodiment.

FIG. 9 is a flowchart illustrating a detailed operation of an abnormal sound detection preparation operation according to the present embodiment.

FIG. 10 is a flowchart (part 1) illustrating a control operation of a measurement controller of the abnormal waveform analyzer according to the present embodiment.

FIG. 11 is a flowchart (part 2) illustrating the control operation of the measurement controller of the abnormal waveform analyzer according to the present embodiment.

FIG. 12 is a diagram illustrating a configuration example of a conventional FFT analyzer.

FIG. 13 is a diagram illustrating a waveform example of a noise signal.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Production plant 2 Power equipment installation area 3 Microphone 4,11 Noise source 12 Non-directional microphone 13 Unidirectional microphone A, B, C Microphone installation position S Abnormal sound source position

Claims (6)

[Claims] 1. A microphone that converts ambient sound into an electric signal, a sound analysis unit that detects a time change of a frequency distribution of a sound signal from the microphone, and a storage that previously stores a time change of a frequency distribution of the sound signal. Means for detecting abnormal sound by comparing the time change of the frequency distribution of the current sound signal with the time change stored in the storage means. system. 2. The abnormal sound detecting system according to claim 1, wherein said sound analyzing means performs a wavelet analysis of the sound signal. 3. The abnormal sound detection system according to claim 1, further comprising a position detecting means for detecting a sound generation position by cross-correlating sound signals from a plurality of microphones. 4. The abnormal sound detecting system according to claim 3, wherein said position detecting means compares amplitudes of sound signals from a plurality of microphones. 5. The microphone comprises: a first microphone having a sharper directivity; and a second microphone having a lower directivity than the first microphone, and the sound analyzing means includes a sound signal from the first microphone. 2. The abnormal sound detection system according to claim 1, wherein a time change of the frequency distribution of the sound signal is detected after canceling the sound signal from the second microphone. 6. A computer-readable computer having recorded thereon a program for causing a computer to function as a sound analysis unit, a storage unit, and an abnormal sound detection unit of the abnormal sound detection system according to claim 1. Recording medium.