JP2002277228A - Sound source position evaluating method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the evaluation precision of a device for evaluating a sound source position such as an accident point in a power plant from the outputs of noise meters 3 installed at at least three places. SOLUTION: Sampling data outputted by a noise meter 3 are divided into k/2 former-half pieces and k/2 latter-half pieces in each section, which are summed up in each section. A variation quantity is found from the absolute value of the difference between the sums to perform filter processing for extracting an abruptly generated sound. When the distance from the sound source position to respective sound absorbing means are found, correction coefficients for the reflection by a shield and its bypassing are introduced. Then the center of gravity of respective sound source positions in the section wherein the largest number of sound source positions are present is evaluated as the sound source position, by using data obtained by combining three out of the outputs of noise meters 3 installed at four positions and data on a sound source position computed including a noise meter 3 relatively distant from the sound source.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention locates an accident point such as a short circuit or a ground fault in a gas insulated switchgear filled with an inert gas called a GIS (Gas Insulated Switchgears). And a sound source position locating method that is preferably implemented to perform the above.

[0002]

2. Description of the Related Art In power plants and substations in which devices such as transformers and switches are individually installed, it is possible to find out the location of an accident by scorching or smell.
In S, the device is stored in the tank, and it is necessary to release the airtightness in order to confirm the location where the accident has occurred.
For this reason, when the airtightness of many gas compartments is sequentially released and the checking operation is performed, there is a problem that a load for returning to the airtight state is increased. The location of the accident has been identified based on deterioration and increase in gas pressure.

However, in such a method, it is necessary to provide a sensor for each gas section, and there is a problem that the configuration becomes large and the burden of maintenance increases.

[0004] For this reason, for example, Japanese Patent Application Laid-Open No. 3-13873 is disclosed.
In the publication, the timing at which an electromagnetic wave is detected by an electromagnetic wave detection antenna is defined as an accident occurrence timing, and the accident occurrence location is specified based on the arrival time of the accident sound to microphones provided at three or more known positions. I have.

[0005]

However, as in the above-mentioned prior art, in the configuration for determining the location of an accident based on the sound collected by a microphone, the influence of background noise such as a vehicle or an aircraft, or the occurrence of an accident is considered. There is a problem that the sound source position cannot be located with high accuracy due to the effect of sound reflection or detour by structures near the location.

It is an object of the present invention to provide a sound source position locating method capable of locating a sound source with high accuracy.

[0007]

According to the sound source position locating method of the present invention, the output of each of the sound collecting means installed at three or more locations is sampled at predetermined time intervals, and the difference between the arrival time of the sound and the arrival time difference of the sound is determined from the obtained sampling data. In the method of locating a sound source based on
By dividing each (even number) pieces, calculating the sum of the data of the first half k / 2 and the data of the second half k / 2 in each section, and calculating the variation from the absolute value of the difference of the sum, It is characterized in that filter processing for extracting a suddenly generated sound is performed.

[0008] According to the above configuration, a sound source such as an accident point in a substation is determined based on a difference in arrival time of sound between sound collecting means such as sound level meters provided at three or more known positions. When locating the position, filter processing is performed to remove slowly changing DC background noise components such as vehicles and aircraft, and to extract only AC sounds that suddenly occur, such as accident sounds. I do. Therefore, the influence of the background noise component can be eliminated, and the sound source position such as the accident point can be located with high accuracy.

In the filter processing, the frequency of calculation can be reduced to 1 / k by dividing the sampling data of each sound pickup means into k (even number) pieces.
By reducing the weight of one data to 1 / k, it is possible to reduce the influence of small data fluctuation. Furthermore,
In each section, a sum is obtained from the data of k / 2 in the first half and k / 2 in the second half, and a variation is obtained from the absolute value of the difference between the sums, that is, the average difference between the first half and the second half. Therefore, compared to a case where the difference is simply obtained from the preceding and succeeding data, it is possible to further reduce the influence of the small data fluctuation.

The sound source position locating method according to the present invention is a method for locating a sound source based on a relative arrival time difference of sounds to three or more sound collecting means. Is set, the distance from each sound source position to each sound pickup means is obtained, and the arrival time of sound from each sound source position to each sound pickup means is obtained based on the distance, and the arrival time difference between each sound pickup means. Is determined, the virtual sound source position where the error between the actually measured arrival time difference and the arrival time difference calculated above is reduced, is set as the sound source position, and the distance or arrival time from each sound source position to each sound pickup means is determined. In the calculation, a correction coefficient for the influence of the shield is introduced.

According to the above configuration, in locating the sound source based on the relative arrival time difference of the sound to each of the sound collecting means installed at three or more locations, a virtual sound source position is assumed, and the virtual sound source position is assumed. By changing the sound source position, a virtual sound source position in which the difference between the expected arrival time difference between the sound collecting means and the actually measured arrival time difference is reduced is estimated as the sound source position. The arrival time difference can be calculated based on the arrival time obtained by obtaining the distance from the virtual sound source position to each sound collecting means and dividing the distance by the speed of sound.
At this time, in the present invention, a correction coefficient for reflection or detour by a shielding object is introduced when obtaining the distance or the arrival time from the virtual sound source position to each sound collecting means. For example, the calculated distance is multiplied by a factor to make it seem longer than the actual linear distance.

Therefore, when locating the accident point in the substation, the accident sound bypasses the structure near the accident point, so that even if the actual arrival time is delayed, the accident point can be accurately detected. Can be determined.

Still further, a sound source position locating method according to the present invention is a method for locating a sound source based on a relative arrival time difference of sound to each sound collecting means installed at A (A is 3 or more). The sound pickup means is installed at a location B (B is A + 1 or more) so as to surround the detection symmetry area, and the detection symmetry area is divided into the B sections including the respective sound pickup means,
A sound source position is calculated for each combination of the A sound pickup means,
In a section in which the calculated sound source positions are most present, the center of gravity of each sound source position in the section is located as the sound source position.

According to the above arrangement, when locating the sound source on the basis of the relative arrival time difference of the sound to the sound collecting means installed at three or more locations A, the detection symmetry area is surrounded. Then, sound collecting means is installed at a position B (B is A + 1 or more), and a sound source position is calculated for each combination of A sound collecting means. For example, when A = 3 and B = 4,
The detection symmetry area is divided into four sections, and four combinations of three sound collection units are created, and therefore, the data of the calculated sound source position is also four. Then, the section in which the four sound source positions are most present is extracted, and the center of gravity of each sound source position in the section is located as the sound source position.

Therefore, since the data of the sound source position calculated including the sound collecting means relatively far from the sound source is excluded, the positioning accuracy can be improved.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described.
The following is a description based on FIGS. 1 to 7.

FIG. 1 is a diagram showing a schematic configuration of an in-plant accident point locating apparatus to which a sound source locating method according to an embodiment of the present invention is applied. The in-plant accident point locating device of the present invention is configured such that it roughly surrounds the locating device 1 implemented by a personal computer or the like and the GIS 2 in the site of the substation.
The sound level meter 3 installed at four or more known positions (in the example of FIG. 1, four positions, individually referred to by reference numerals 30, 31, 32, and 33, respectively; 3).

As shown in FIG. 2, the locating device 1 is realized by an input circuit 11 for sampling the output from each sound level meter 3 at a high speed of, for example, 10 kHz and performing analog / digital conversion, a microcomputer, and the like. A processing circuit 12 for holding the sampling data in the input circuit 11 and performing a filtering process and a location operation of an accident point as described later, a display device 13 realized by a cathode ray tube or the like, and realized by a printer or the like. Output device 14
It is comprised including. The input circuit 11 is realized by an overvoltage relay, an overcurrent relay, or the like, and receives an operation signal from a protection relay 20 that detects an occurrence of an accident such as a short circuit or a ground fault and drives a circuit breaker.

The filtering process in the processing circuit 12 is performed for each sound level meter 3, and first, the sampling data is divided into k (even number). Next, in each section, the data is further divided into the first half k / 2 data and the second half k / 2 data, the sum is calculated for each, and the absolute value of the difference is calculated.

That is, the data N _J used for the fault point location calculation at a certain timing j is obtained for every k pieces. When the actual sampling data is M _J , N _J = | (M _Jk−1 + M _Jk−2 +... + M _{Jk / 2} ) − (M _{Jk / 2 + 1} + M _{Jk / 2 + 2} +... + M _J ) | For example, if k = 8, the sampling frequency of the data used in the fault point location calculation, and therefore the calculation frequency of the fault point, is reduced to 1/8 to 1.25 kHz, and N _J = | (M _J−7 + M _{J-6 + M J-5} + M J-4) - (M J-3 + M J-2 + M J-1 + M J) | ... it becomes (2).

Thus, the frequency of the accident point locating operation using the data N _J can be reduced to 1 / k, and the weight of each sampled data M _J can be reduced to 1 / k to reduce the fluctuation of the fine data. The effect can be reduced. Furthermore, a sum is obtained from the data of the first half k / 2 and the second half k / 2 in each section, and the sum is calculated from the absolute value of the difference of the sum, that is, the average difference between the first half and the second half. Since the amount is obtained, the influence of small data fluctuation can be reduced as compared with the case where the difference is simply obtained from the preceding and following data.

The fluctuation amount becomes large when a sudden sound is generated. Thus, a slowly changing DC background noise component of a vehicle or an aircraft is removed, and a sudden sound such as an accident sound is generated. Can be filtered to extract only the AC-like sound generated in the above. Therefore, the influence of the background noise component can be eliminated, and the accident point can be located with high accuracy.

FIG. 3 is a diagram for explaining a method of calculating an accident point. In the calculation method of the present invention, three or more locations (FIG. 3)
Then, for the sake of simplicity, the data of the three sound level meters 3 are used, and the accident point is calculated based on the relative arrival time difference of the sound to each sound level meter 3. First, as shown in FIG. 3, the installation positions of the sound level meters 30, 31, and 32 are indicated by reference signs P0, P1, and P2, respectively, and their coordinates are (x0, y0),
(X1, y1) and (x2, y2). In the present invention, a large number of accident points are virtually set as described later. The accident point is indicated by reference symbol P, and its coordinates are (x,
y). The distances L0, L1, L2 between the accident point P and each of the installation positions P0, P1, P2 are:

[0024]

(Equation 1)

## EQU2 ##

Next, in the present invention, each of the distances L0, L
1, t2, the time required for the accident sound to propagate through L2
1, t2 is obtained from the following equation.

T0 = L0 * f0 / V (6) t1 = L1 * f1 / V (7) t2 = L2 * f2 / V (8) where V is a sound speed, and f0, f1, and f2 are Accident point P
Are correction factors for the reflection of accident noise and the influence of detours due to structures intervening from to the respective installation positions P0, P1, and P2. That is, when there is no influence by the structure, the correction coefficients f0, f1, and f2 are 1.

Therefore, each installation position P0, P1, P2
Time difference t01, t12, t20 of accident sound between
T01 = t0−t1 = L0 * f0 / V−L1 * f1 / V (9) t12 = t1−t2 = L1 * f1 / V−L2 * f2 / V (10) t20 = t2−t0 = L2 * f2 / V-L0 * f0 / V (11)

In this way, the arrival time differences t01, t12, and t20 are calculated with respect to the set virtual accident point P, and subsequently, the actually measured arrival time differences are represented by T0.
1, T12 and T20, the residual S becomes

[0030]

(Equation 2)

Is obtained as shown in FIG.

Then, by changing the coordinates (x, y) and the correction coefficients f0, f1, f2 of each of the many accident points P as appropriate, the combination of parameters that minimizes the residual S is minimized. It is determined by the square method. As shown in Table 1, the parameters of the coordinates (x, y) are changed, for example, every 0.1 m. Therefore, if the detection symmetry area is 10 m × 10 m, the number of points is 10,000.
The parameters of the correction coefficients f0, f1, f2 are as follows:
For example, the number of data is changed in the range of 0.1 to 1.9, so that the number of data is 20 ³ = 8000.

[0033]

[Table 1]

Here, depending on how the correction coefficients f0, f1, f2 are selected, the residual S at other points may be smaller than the residual S at the actual accident point P. May not be the solution representing the accident point P. However, it is conceivable that many data (approximate solutions) when the correction coefficients f0, f1, and f2 are slightly different from each other gather near the actual solution.
In the present invention, a combination of parameters in which the residual S becomes a predetermined level C, for example, 1 msec or less, is used as data indicating an accident point, and the center of gravity of the data is used as the coordinates (xL, yL) of the accident point P. Is calculated as

That is, the coordinates below the level C are represented by (x
i, yi) and the residual at that time is Si, xL = Σ (C-Si) * xi / Σ (C-Si) (13) yL = Σ (C-Si) * yi / Σ ( C-Si) (14) The coordinates (xL, yL) obtained from (14) are calculated as the coordinates of the accident point P. This is illustrated in FIG.

In this manner, the correction coefficients f0, f1, and f2 for the reflection and detour of the accident sound by the obstacle are introduced,
For example, by making the distance apparently longer than the actual linear distance, the accident point can be estimated with high accuracy even when the actual arrival time is delayed.

The accident point P estimated as described above is
In the present invention, B (B is A + 1 or more, and B = B in FIGS.
4) Among the sound level meters 30 to 33, A (A is 3 or more, A = 3 in FIG. 3) is obtained for each combination, and the final result is obtained by comparing them with each other. The accident point is located. More specifically, assuming that A = 3 and B = 4, as shown in FIG. 5, the installation positions of the sound level meters 30 to 33 are indicated by reference numerals P0 to P3, respectively. ,
The coordinates of the accident points p0 to p3 (xL0,
yL0), (xL1, yL1), (xL2, yL2),
(XL3, yL3) are compared with each other.

In the example of FIG. 5, FIG.
32, that is, the estimated values from the measurement results at the positions P0, P1, and P2, FIG. 5B shows the estimated values from the measurement results at the positions P1, P2, and P3, and FIG. P
2, P3 and P0 are estimated values from the measurement results, and FIG.
(D) is an estimated value from the measurement results at the positions P3, P0, and P1. As shown in FIGS. 5A to 5D, when the detection symmetric region is divided into the four sections Q0 to Q3 including the sound level meters 30 to 33, the accident points p0 to p3
May be biased in the section including (in FIG. 5,
Three accident points p0 to p2 are biased in section Q2). In this case, it is considered that the sound level meter of the section is closest to the accident point, and the section without the accident point is relatively far from the accident point, the sound pressure level is not so high, and it is considered that the accuracy is poor. In the present invention, as shown in FIG. 6, the center of gravity of each of the fault points p0 to p2 in the section Q2 is located at the fault point P in the section Q2 where the fault points are most present.

The method of finding the center of gravity is based on the coordinates (xLi, yL
i) (in the example of FIG. 5, i = 0, 1, 2), when the minimum values are Dx and Dy, M1 = Σ (xLi−Dx) * (yLi−Dy) (15) M0x = Σ (xLi−Dx) (16) From M0y = Σ (yLi−Dy) (17), the coordinates (x, y) of the accident point P are calculated as follows: x = M1 / M0y + Dx (18) y = M1 / M0x + Dy … (19)

With this configuration, since the data of the accident point p3 estimated including the sound level meter 33 which is relatively far from the accident point P is excluded, the localization accuracy can be improved.

Since the frequency of the accident point locating calculation is reduced to 1 / k, the data has a step-like shape.
For data N _J sought, N = α * (1- e - (t-t0) / T) + N J ... as shown in (20), using an exponential function, so as to blunt the data waveform You may. α is a coefficient, and N is data after dulling.

The arrival timing of the accident sound can be obtained, for example, as follows. First, FIG.
As shown in (a), a continuously rising sound exceeding the stationary background noise level V1 (time τ1) and for 15 msec or more (time τ2) is recognized as a sudden sounding mountain wave, The peak time τ3 and the peak value of the mountain wave are obtained.

Next, in order to remove the minute fluctuation of the waveform in the mountain wave as described above, 8 (k) is obtained from the data of 100 μsec period by the sampling of 10 kHz.
The resampled data N is obtained for each unit, and the sampling time τ1 at which the data N goes back from the peak time τ3 and is equal to or less than the background noise level V1 and equal to or less than 37% of the peak value is extracted. When the peak is roughly detected using the resampling data, an accurate timing (time τ1) exceeding the background noise level V1 by the actual sampling at 10 kHz is obtained.

Subsequently, in the resampling data, a time τ4, which becomes a minimum before the time τ1, is obtained as a tentative minimum timing, and in the actual sampling data, a time 30 msec after the time τ4 and the time τ1 are calculated. The timing (time τ5) of the data having the minimum value V2 going back from the earlier time is obtained as the true minimum timing.

Similarly, in the actual sampling data, the data from the time τ5 to before 15 msec,
The minimum value is set as the bottom value V3. Then, as shown in FIG. 7B, 15 msec from the time τ5.
From the data up to the later time or the time τ1 whichever is earlier, the time τ6 at which the second-order differential value is maximum is determined, and the linear characteristic (first-order differential expression) at the time τ6 is the bottom value. The time τ7 equal to V3 is the arrival timing of the accident sound.

By determining the arrival timing in this way, it is possible to suppress the influence of the background noise and extract only sudden sound caused by an accident.

[0047]

According to the sound source position locating method of the present invention, as described above, when locating the sound source based on the arrival time difference of sound, the sampling data of each sound collecting means is divided into k (even number) units. In each section, filter processing is performed to obtain the sum of the k / 2 data in the first half and the k / 2 data in the second half, and to calculate the variation from the absolute value of the difference between the sums.

Therefore, the influence of a slowly changing DC background noise component such as a vehicle or an aircraft can be eliminated, and a sound source position such as an accident point can be located with high accuracy. In addition, in the filter processing, the amount of variation is obtained from the difference between the average of the first half section and the second half section, so that the influence due to the fluctuation of the fine data is reduced as compared with a case where the difference is simply obtained from the preceding and following data. Can be.

The sound source position locating method of the present invention, when locating the sound source based on the relative arrival time difference of the sound to each of the sound collecting means installed at three or more locations, as described above, uses a virtual Assuming the sound source position and changing the virtual sound source position, the virtual sound source position where the error between the expected arrival time difference between each sound pickup means and the actually measured arrival time difference is reduced is determined as the sound source position. In the calculation of the arrival time difference, a correction coefficient for reflection or detour by a blocking object is introduced.

Therefore, the accident point can be located with high accuracy.

Further, in the sound source position locating method of the present invention, when locating the sound source based on the relative arrival time difference of the sound to each of the sound pickup means installed at three or more locations A, the detection symmetric region B (B is A + 1
Above) A sound pickup means is installed at a location, a sound source position is calculated for each combination of A sound pickup means, a section where the sound source position is the largest as a result of the calculation is extracted, and each section in the section is extracted. The center of gravity of the sound source position is located at the sound source position.

Therefore, the data of the sound source position calculated including the sound pickup means relatively far from the sound source is excluded.
The orientation accuracy can be improved.

[Brief description of the drawings]

FIG. 1 is a diagram showing a schematic configuration of an in-plant accident point locating device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an electrical configuration of a localization device in the in-site accident point localization device shown in FIG.

FIG. 3 is a diagram for explaining a method of calculating an accident point.

FIG. 4 is a diagram for explaining a method of estimating a true value of accident point data.

FIG. 5 is a diagram for explaining selection of accident point data obtained by a combination of sound level meters.

FIG. 6 is a diagram for explaining a selection result of FIG. 5;

FIG. 7 is a diagram for explaining how to determine the arrival timing of the accident sound.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 orientation device 2 GIS 3; 30, 31, 32, 33 sound level meter 11 input circuit 12 processing circuit 13 display device 14 output device 20 protection relay

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Toshio Sonoda 3-3-22 Nakanoshima, Kita-ku, Osaka City, Osaka Prefecture Inside Kansai Electric Power Co., Inc. Inside Shin Electric Co., Ltd. In-house (72) Inventor Naoki Okada 47F, Umezu Takaune-cho, Ukyo-ku, Kyoto, Japan F-term in Nissin Electric Co., Ltd. 5J083 AA05 AB12 AC18 AC29 AD02 AE10 BE53

Claims (3)

[Claims] 1. A method for sampling the output of each sound pickup means installed at three or more locations at predetermined time intervals and locating a sound source position based on a difference in arrival time of sound from the obtained sampled data. The sampling data of the means is divided into k (even number) pieces, and the sum is obtained from the data of the first half k / 2 and the second half k / 2 in each section, and the sum is varied from the absolute value of the difference of the sum. A sound source localization method characterized by performing a filtering process for extracting a suddenly generated sound by obtaining an amount. 2. A method of locating a sound source based on a relative arrival time difference of sounds to sound pickup means installed at three or more locations, wherein a plurality of virtual sound source positions are set. The distance to the sound pickup means is obtained, the arrival time of the sound from each sound source position to each sound pickup means is obtained based on the distance, the arrival time difference between the sound pickup means is obtained, the actually measured arrival A virtual sound source position in which an error between the time difference and the arrival time difference obtained by the above calculation is reduced is set as a sound source position. A sound source localization method characterized by introducing coefficients. 3. A method for locating a sound source based on a relative arrival time difference of sound to each sound collecting means installed at a location A (where A is 3 or more). B (B
Is located at A + 1 or more), the detection symmetric region is divided into the B sections including the respective sound pickup means, and the sound source position is calculated for each combination of the A sound pickup means. A sound source position locating method comprising: locating the center of gravity of each of the sound source positions in the section where the largest number of the respective sound source positions exist as the sound source position.