JP6790086B2 - Leakage determination device and leak determination method - Google Patents

Description

The embodiment of the present invention relates to a water leakage determination device and a water leakage determination method.

In the current leak investigation procedure using a leak detector, an investigator such as a meter reader of a water meter first acquires data for a leak investigation of a water pipe using the leak detector at the time of meter reading. For example, the time integration rate (the ratio of signals above a certain level in a certain time) and the vibration sound (water leakage sound) by the water leakage detector are recorded as a sound wave file. The recording point (position of the water pipe) is associated with each sound wave file.

From the acquired data, first, as a primary judgment, a sound wave file having a time integration rate of a certain level or higher is extracted as a sound wave file with a possibility of water leakage.

As a secondary judgment, a file with a high possibility of water leakage is extracted from the extracted sound wave files, and the number of cases is narrowed down. This secondary judgment work is judged by a skilled engineer listening to it (by playing a sound wave file). Conduct a field survey of the recording points of the finally narrowed down sound wave files. A specialist investigator confirms the leak with a sound stick, and digs up the ground after making a hit.

The secondary judgment is a hearing judgment by a research engineer, and depends on the experience and skill of the engineer. Moreover, even if the number of sound wave files obtained in the primary judgment is enormous, it is necessary to listen to all the sound sources, which requires a great deal of labor.

JP-A-2015-43001

An object to be solved by the present invention is to provide a water leakage determination device and a water leakage determination method for more efficiently and accurately detecting water leakage.

The water leakage determination device of the embodiment includes a reception unit that receives sound wave data, an extraction unit that extracts data in a steady section with less noise from the sound wave data, a designation unit that specifies a threshold value, and a water leakage determination unit from the data in the steady section. It is provided with a determination unit for determining the presence / absence of water leakage by obtaining the ratio including the characteristic signal and comparing the ratio with the threshold value.

Further, the water leakage determination method of the embodiment is a water leakage determination method in a water leakage determination device that determines the presence or absence of water leakage from sound wave data, and extracts data of a steady section with less noise from the sound wave data to determine the presence or absence of water leakage. The threshold value for this is specified, the ratio including the signal indicating the characteristic of water leakage is obtained from the data of the steady section, and the ratio is compared with the threshold value to determine the presence or absence of water leakage.

It is a block diagram which shows the structure of the water leakage determination apparatus which concerns on 1st Embodiment. It is a figure which shows the time-series waveform which captured the water leakage sound. It is a figure which shows the time-series waveform which captured the traffic noise. An example of a time-series waveform and a waveform obtained by converting it into spectral entropy is shown. It is a figure which shows an example which extracts a steady section by spectrum entropy conversion from a time series waveform. It is a block diagram which shows the detail of the determination part of the water leakage determination apparatus which concerns on 1st Embodiment. It is a figure which shows an example of the case where an arbitrary frequency range including a frequency component due to water leakage is set from the FFT waveform. It is the figure which superposed the frequency characteristic at the time of applying 1/3, 1/6, 1/12 octave band. It is a comparison diagram of the case of smoothing by moving average and the case of smoothing by 1/12 octave band. It is a figure which shows an example which obtained the time integration rate from the time series waveform of the sound wave signal data after applying HPF. It is a figure which shows the frequency characteristic of a septic tank sound and a motor sound. It is a figure which shows the result of having performed the water leakage determination of the water pipe using the 1st determination means and the 2nd determination means. It is a figure which shows an example of the determination threshold value of the 1st determination means, the determination threshold value of the 2nd determination means, and the determination amplitude value (voltage). It is a block diagram which shows the detail of the determination part of the water leakage determination apparatus which concerns on 2nd Embodiment. It is a comparison diagram of the time series waveform and the frequency characteristic of each of the sound of water leakage and the sound of water used (sound when using water supply). It is a figure which shows the combination of the 1st to 3rd determination means. It is a block diagram which shows the detail of the determination part of the water leakage determination apparatus which concerns on 3rd Embodiment. It is a detailed block diagram of the 4th determination means. It is a figure which shows the specific example of the determination procedure when a test signal is used. It is a figure which shows an example of the determination procedure at the time of including a fluctuating sound. It is a figure which shows an example of the time series waveform when the signal of a plurality of frequency components is included. It is a figure which shows the autocorrelation function of the time series waveform of FIG. 21, each envelope line and the approximate straight line. It is a figure which shows the result of having obtained the autocorrelation function of the sound wave signal data actually measured. It is a figure which shows an example of the determination process by defining a predetermined line segment and a threshold value. It is a figure which shows the result of having compared the size of the area made by each of an envelope line and a predetermined line segment. It is a figure which shows the result when the water leakage is judged based on any of the time integration rate, the area formed by the envelope, or the slope of the envelope from the autocorrelation function. It is a figure which shows the combination of the 1st, 2nd or 4th determination means. It is a figure which shows the combination of the 2nd, 3rd or 4th determination means. It is a figure which shows the combination of the 1st, 3rd or 4th determination means. It is a figure which shows the combination of the 1st to 4th determination means.

Hereinafter, the water leakage determination device and the water leakage determination method according to the embodiment will be described with reference to the drawings. Those with the same reference numerals indicate similar ones. The drawings are schematic or conceptual, and the relationship between the thickness and width of each part and the ratio coefficient of the size between the parts are not necessarily the same as those in reality. Further, even when the same part is represented, the dimensions and ratio coefficients may be represented differently depending on the drawings.

(First Embodiment)
FIG. 1 is a block diagram showing a configuration of a water leakage determination device 1 according to the first embodiment.

The water leakage determination device 1 includes a reception unit 2 that receives sonic signal data, an extraction unit 3 that extracts a steady section with less noise from the input sonic signal data, and a sonic signal extracted by the extraction unit 3. It has a determination unit 4 that determines whether or not the data includes the characteristic of water leakage by using two determination means, and an output unit 5 that outputs the determination result. Further, the designation unit 6 is provided so that the upper and lower frequency frequencies of the frequency range, the threshold value for determining water leakage, and the like can be specified by the user. The designation unit 6 gives a threshold value or the like to the determination unit 4.

The reception unit 2 is, for example, a place where sound wave signal data such as running water flowing through a water pipe measured by a meter reader or an investigator of a water meter is input. The sound wave signal data may be data typified by the WAV format that can be handled by general personal computers and electronic devices. Further, any sound wave signal data in another format may be used as long as it can be converted into WAV format. For example, compressed data such as ADPCM, MP3, and WMA may be used. The reception unit 2 may receive the file directly by USB or the like, or may receive the data via an external medium such as an SD card. In addition, data may be accepted via wireless communication such as Wi-Fi or Bluetooth.

In addition to the vibration sound to be detected, the sound wave signal data includes a lot of noise such as traffic noise of cars and trains, mechanical noise such as outdoor unit noise of air conditioners, and human voice due to the influence of the surrounding measurement environment. Is done. The sound wave signal data is mainly given as a time series waveform.

Generally, the sound of water leakage from a water pipe is a steady sound that includes frequency components up to 3 kHz and is dominated by frequency components of 1 kHz or more and 2 kHz or less. Sound waves that are easily misjudged as water leakage sounds include the sound of water flowing out of a faucet when using water and the sound of rotation of a water meter. Other examples include sound waves caused by motors such as air conditioners. The frequency characteristics in the state where these noises are mixed deviate greatly from the original frequency characteristics of the leaking sound.

The extraction unit 3 is a location for extracting the waveform of a component having less noise from the sound wave signal data input from the reception unit 2. That is, from the time-series waveform of the sound wave signal data, the waveform of the sudden vibration component having a large amplitude or the waveform of the stationary section which is not the waveform portion having a large amplitude that is periodically superimposed on the waveform is extracted.

FIG. 2A is a diagram showing a time-series waveform when the sound of water leakage is captured. The horizontal axis is time (seconds) and the vertical axis is the signal level, which is the amplitude. FIG. 2B shows a frequency characteristic when a time series waveform of 10 seconds, which is the total time, is subjected to a fast Fourier transform (FFT). It can be seen that no clear feature has been confirmed in the range of 1 kHz or more and 2 kHz or less, which is the frequency component of the leaking sound. On the other hand, among the time-series waveforms shown in FIG. ). Clear features are confirmed in the range of 1 kHz or more and 2 kHz or less. By extracting the waveform of the stationary section and performing FFT, the characteristics of water leakage can be clearly obtained.

FIG. 3A is a diagram showing a time-series waveform when traffic noise is captured. Here, it is assumed that the water leakage sound is not superimposed. FIG. 3 (b) shows the frequency characteristics when the time-series waveform for 10 seconds, which is the total time, is FFTed.

Even in the waveform of FIG. 3B, clear features are confirmed in the range of 1 kHz or more and 2 kHz or less. This is the effect of external noise, which is characterized by the same band as the leak sound. The cause of the noise is considered to be a higher-order component of the motor sound of an automobile or a train. In general, the running noise and motor noise of automobiles and trains measured in an external environment are not steady waveforms, but the time-series waveforms suddenly change depending on the road condition and the accelerator condition.

Of the time-series waveforms of FIG. 3 (a), the frequency characteristics when FFT is performed using the waveform of the designated portion which is a steady section is shown in FIG. 3 (c). It can be seen that no clear feature was confirmed between 1 and 2 kHz as compared with the waveform of FIG. 3 (b).

By extracting a steady section with less sudden noise from the time-series waveform of the sound wave signal data by the extraction unit 3, it is possible to obtain a frequency characteristic from which noise components other than the water leakage sound are removed when FFT is performed thereafter. In addition, it is possible to prevent erroneous detection of sudden noise or the like having a frequency component close to that of water leakage sound.

Next, a method of extracting a stationary section in the extraction unit 3 will be described.

As a method for extracting a stationary section, for example, a spectral entropy method can be used. The spectral entropy method is a calculation of the deemed information entropy with the spectrum of the signal as a probability distribution, and is a feature quantity representing the whiteness of the signal. A high value is obtained when the spectrum is uniform, such as white noise, and a low value is obtained when the spectrum is non-uniform, such as an audio signal or traffic noise.

The spectral entropy method is generally expressed by the following equation (1).

S _f is an amplitude spectrum of the frequency component f obtained by FFTing the input signal. P _f is the existence probability of the frequency component f. H indicates spectral entropy. H has a high value for a white signal having a uniform spectrum, and a low value for a colored signal having a non-uniform spectrum such as an audio signal or traffic noise.

FIG. 4 shows a time series waveform and a waveform obtained by converting the time series waveform by the spectral entropy method. The stationary section of the graph shown by the spectral entropy method is flat at a high value, and the portion of external fluctuation noise is a low value. The stationary section and the external fluctuation noise part can be clearly distinguished.

As a result, only the components regarded as the stationary section are extracted while avoiding the locally fluctuating noise points. Extraction of steady-state sections is performed using a threshold value. That is, a section that exceeds the set threshold value and is regarded as a steady section is extracted. The threshold value for delineating the steady section can be set arbitrarily. In addition, the time width for extraction can be set arbitrarily.

For example, depending on the sound wave signal data, a continuous steady section may not be obtained. In that case, the dispersed parts may be extracted so that the total of the steady-state sections has an arbitrarily set time width. However, when calculating the frequency characteristics, it is preferable that the sound wave signal data is continuous. This is because if the continuity of the time-series waveform of the sound wave signal data is lost, a different spectrum is obtained when FFT is performed.

For example, when the total of the stationary sections, which is the time to be extracted, is 1 second, and if the extraction cannot be performed in a continuous time range, the fragments may be extracted for a total of 1 second. When FFT is performed with an arbitrary window length, it is performed for each fragment data. This is because the continuity of the time series waveform is lost when the FFT straddling the two fragment data is performed.

If you want to maintain the continuity of the time series waveform to be extracted, you may lower the threshold value to be set. As a result, the section determined to be equal to or higher than the threshold value becomes longer, so that it becomes easier to extract a continuous signal regarded as a steady section.

FIG. 5 is a diagram showing an example of extracting a stationary section from a target time-series waveform using the spectral entropy method.

The first stage of the figure is a target time-series waveform. A stationary section with a specified time width is extracted from this. The second row in the figure is the waveform converted by the spectral entropy method. The broken line is an arbitrarily determined threshold value. The third row in the figure is a determination result when the section above the threshold value is set to 1 and the section below the threshold value is set to 0. The “1” section, which is a section above the threshold value, is a section extracted as a steady section. From this, a section with an arbitrarily set time width is extracted. If there is no interval longer than an arbitrary time width and a continuous steady interval is required, the threshold value may be set low.

The threshold value for extracting the stationary section may be set by the total average value of the spectral entropy of the section to be extracted, the frequency distribution may be calculated and the median value thereof may be used, or an arbitrary numerical value determined by the user. May be given directly.

The extraction in the extraction unit 3 is mainly performed by a computer CPU (central processing unit) or the like.

The determination unit 4 determines whether or not the steady section of the sound wave signal data extracted by the extraction unit 3 includes features due to water leakage.

FIG. 6 is a block diagram showing details of a determination unit of the water leakage determination device according to the present embodiment.

In the determination unit of the present embodiment, the presence or absence of water leakage is determined by the first determination means 7 and the second determination means 8 different from the first determination means, respectively, using the time-series waveform of the steady section. By combining these judgment results, it is possible to easily narrow down the secondary judgment of the sound wave signal data.

First, the first determination means 7 will be described.

The first determination means performs FFT on the sound wave signal data of the stationary section extracted by the extraction unit 3 and calculates the ratio of the frequency component due to water leakage in an arbitrarily determined frequency range. Further, the presence or absence of water leakage is determined by setting a threshold value for this ratio.

FIG. 7 shows an example in which an arbitrary frequency range including a frequency component due to water leakage is set from the FFT waveform. Since the frequency component caused by water leakage is characterized in a range of about 1 kHz or more and 2 kHz or less, for example, a range of 700 Hz or more and 3 kHz or less is set as the frequency range.

In the first determination means, the Euclidean distance is calculated as the calculation of the ratio of the frequency component caused by the water leakage.

Since the frequency characteristics of the microphone or amplifier circuit are superimposed on the frequency characteristics of the sound wave signal data when there is no sound, the frequency characteristics are not always flat, and some kind of signal is often mixed. The Euclidean distance D of the frequency characteristic of the sound wave signal data of the steady section extracted by the extraction unit is obtained with respect to the frequency characteristic when the sound wave input level is the minimum, that is, the lower limit data of the dynamic range. The Euclidean distance D is calculated by the following equation (2).

X _i is the frequency characteristic data of the steady section extracted by the extraction unit. _Pi is the frequency characteristic data of the lower limit of the dynamic range. i is an array of frequencies 1 to d. The frequency characteristic data of the lower limit (minimum value) of the dynamic range is understood as the so-called background noise frequency characteristic data.

When the ratio of the Euclidean distance in the frequency range (for example, 700 Hz or more and 3 kHz or less) including the frequency component caused by the water leakage exceeds the threshold value set arbitrarily, it is determined as the water leakage. If the amplitude value is only the sound wave signal data of the steady section extracted by the extraction unit, the frequency component in which a large amplitude appears as an absolute value is large even in the lower limit data of the dynamic range, and in fact it may not be a large input. It is preferable to take the difference between the frequency characteristics of the stationary section and the lower limit data of the dynamic range.

As a method of determining the lower limit data (minimum value) of the dynamic range, a sound wave signal recorded in advance in a place where background noise is low may be used, or the lowest value of each frequency of a plurality of input sound signal data is extracted and the lowest value is extracted. It may be combined and used as the lower limit data.

When determining the minimum value of the dynamic range from a plurality of input sound wave signal data, all the files are read once before entering the block of the determination unit. Every time the file is read, FFT is performed, the dynamic range information already saved for each frequency is compared with the FFT value, and if it is small, the dynamic range information is overwritten. By doing this for the whole sound wave signal data file, the minimum value of the dynamic range is determined. The determination unit stores the frequency characteristics of the lower limit of the obtained dynamic range in a storage unit such as a memory, and calculates the Euclidean distance by the equation (2) using the stored lower limit data.

In addition to obtaining the Euclidean distance, the integrated value in an arbitrary frequency range for the frequency characteristic obtained by FFTing the stationary section of the sound wave signal data and the integrated value in an arbitrary frequency range for the frequency characteristic of the minimum dynamic range value are used. You may find the difference. The ratio (integration rate) between this and the difference between the integrated values in the entire frequency range of both frequency characteristics may be obtained and compared with an arbitrarily determined threshold value. The area of the difference between the so-called FFT frequency characteristic of the stationary section and the frequency characteristic of the minimum dynamic range is the ratio between the case where the entire frequency range is set and the case where the frequency range is set to an arbitrary range.

The determination of the threshold value can be arbitrarily set by the user from past leak determination data and empirical rules.

In the first determination means, smoothing processing is performed as an approximation method of frequency characteristics in order to reduce the amount of data processing before performing threshold determination based on the Euclidean distance. This is because the amount of data processing becomes enormous when FFT is simply performed.

Although it depends on the number of smoothing points, if steep peak noise is mixed in, the graph shape will be greatly distorted. Therefore, it is simplified by 1 / N octave band processing, the number of data points is reduced, and the processing speed is increased. In octave conversion, in the case of noise evaluation, it is common to use 1/1 octave and 1/3 octave, but it becomes more detailed by increasing the parameter of 1/6 octave, 1/12 octave band and N. .. FIG. 8 is a diagram in which the frequency characteristics in the case of 1/3, 1/6, and 1/12 octaves are overwritten. The 1/12 octave band has the largest number of data points and a detailed graph. It is good to determine the value of N by a trade-off with the data score.

FIG. 9 is a comparison diagram between the case of smoothing by moving average and the case of smoothing with a 1/12 octave band. When smoothed by a moving average, the peak portion becomes dull due to the averaging process of a plurality of points, but the smoothing in the 1/12 octave band can clearly reproduce the shape of the peak. By using 1 / N octave processed data instead of simple frequency conversion data when making a judgment based on the Euclidean distance, high-speed calculation can be performed with a small number of data points while leaving appropriate frequency information.

The designation unit 6 in the first determination means inputs the upper and lower frequency frequencies of an arbitrary frequency range. When determining the water leakage sound as described above, since the frequency component caused by the water leakage is 1 kHz or more and 2 kHz or less, the frequency range is set, for example, in the range of 700 Hz or more and 3 kHz or less. When making a judgment for sounds other than water leakage, it is better to set the frequency according to the characteristics of the sound type to be extracted. It is also possible to determine the values of the upper limit frequency and the lower limit frequency from a plurality of sound wave signal data that are certain to be leaking sounds. For example, the frequency with a large contribution may be extracted from the average value of the frequency-converted ones, or the frequency can be obtained by calculating the frequency distribution. Further, in the designation unit 6, a threshold value (%) for determining water leakage is designated with respect to the Euclidean distance D (%). Any value can be specified for the threshold value, and it can be obtained from the relationship between the frequency range and the Euclidean distance D from a plurality of sound wave signal data that are clearly known to be water leakage sounds. From the average value of the D values when the frequency range is changed, the threshold value that can be determined as water leakage can be inferred.

The parameter designation in the designation unit 6 may be performed using an external terminal such as a PC or a mobile phone, or may be directly input by attaching a monitor, a touch panel, or the like. When specified by an external terminal, data is transferred using the Internet, Wi-Fi, Bluetooth, or the like. Further, it may be specified by using a display or the like of the output unit 5 described later.

By 1 / N octave processing and Euclidean distance determination, it is possible to determine water leakage with a small number of data points.

Next, the second determination means 8 of the determination unit 4 will be described. The second determination means is a means for determining water leakage mainly in the time series region.

Specifically, a high-pass filter (HPF) is applied to the sound wave signal data of the stationary section extracted by the extraction unit 3, and the ratio of the total time of the signals having the determination amplitude E or more (time integration rate) set arbitrarily is used. Make a water leak judgment. The time interval of the sound wave signal data for determining water leakage can be arbitrarily selected.

The time integration rate (%) is expressed by the following equation (3).

F (time integration rate%) = number of times the amplitude E is exceeded / total number of sampling times x 100 ... (3)
FIG. 10 is a time series waveform of the sound wave signal data after applying the HPF. An example is shown in which the total time exceeding an arbitrarily determined amplitude ± E is calculated as the time integration rate (%).

By setting a threshold value for the time integration rate, it is determined that there is a possibility of water leakage when the threshold value is exceeded. In general, the frequency characteristics of septic tank sound and motor sound, which are easily erroneously detected by water leakage sound, are mainly peaks at the drive power supply frequency and frequencies that are multiples thereof, and have a great influence on low frequency components. FIG. 11 is a diagram showing frequency characteristics of septic tank sound and motor sound. Both frequency characteristics have a peak at low frequencies. On the other hand, the frequency characteristics of the leaking sound are different because they are characterized by a relatively high frequency of 1 kHz or more and 2 kHz or less.

Therefore, by applying the HPF, the power supply drive frequency in the low frequency range and frequency components that are multiples thereof, which are easily erroneously detected, are reduced, and the septic tank sound and the motor sound are less likely to be detected by the time integration rate. As a result, it is possible to narrow down the number of suspected leaks. It should be noted that the frequency of the power supply drive frequency to be reduced can be arbitrarily determined.

In the designation unit 6 of the second determination means, the value of the determination amplitude E described above is designated. In addition, a threshold value (%) for determining the presence or absence of water leakage is specified for the time integration rate (%). Also, input the power supply frequency. Since there are two power supply frequencies in Japan, the 50Hz region and the 60Hz region, the power supply frequency is selected according to the measurement location. With the HPF, it is possible to arbitrarily input the order component of the power supply frequency to reduce the peak sound.

In the present embodiment, two types of the first determination means and the second determination means have been described as the leak determination means of the determination unit 4, but these determination means may be applied individually or in series or in parallel. May be applied. When determining in series, for example, after determining the presence or absence of water leakage by the first determination means, the presence or absence of water leakage is determined again by the second determination means with respect to the sound wave signal data determined to have water leakage. Which determination means is performed first can be arbitrarily set. When applied in parallel, the determination results of the first determination means and the second determination means may be represented by either a logical sum or a logical product. As a judgment result, the logical product is a stricter judgment result.

As described above, the first determination means and the second determination means may be determined individually, or may be combined in series or in parallel. The judgment result in the judgment unit is output to the output unit.

Execution of the determination algorithm in the determination unit is mainly performed by the CPU of the computer or the like.

FIG. 12 shows the result when the water leakage of the water pipe is determined by using the first determination means and the second determination means.

For 103 sound wave signal data in which the investigator made the primary determination by the detector, the determination was performed by the algorithms of the first determination means and the second determination means. The leak data of the answers was 24 out of 103 cases that were judged to be water leaks by hearing.

FIG. 13 shows a determination threshold value of the first determination means, a determination threshold value of the second determination means, and a determination amplitude value (voltage).

The determination threshold value of the first determination means was set to 20%, and the determination threshold value higher than that was set to determine water leakage. The determination threshold value of the second determination means was set to 40%, and a determination threshold value higher than that was set to determine water leakage. For the threshold setting, the one with the highest judgment accuracy was selected from the past leak judgment data and empirical rules. The judgment amplitude value was determined by combining the detector data. The judgment threshold value and the judgment amplitude value can be arbitrarily selected according to the measurement target and the measurement environment.

From FIG. 12, in the case of the single determination of the first determination means, 49 out of 103 cases were determined to have water leakage. Furthermore, 20 out of 24 cases agreed with the result of hearing judgment. In the case of the single determination of the second determination means, 48 out of 103 cases were determined to have water leakage. Furthermore, 20 out of 24 cases agreed with the result of hearing judgment. In the case of the combined determination of the first determination means and the second determination means (logical product), 40 out of 103 cases were determined to have water leakage. Furthermore, 20 out of 24 cases agreed with the result of hearing judgment. In the case of the combined determination of the first determination means and the second determination means (logical sum), 57 out of 103 cases were determined to have water leakage. Furthermore, 22 out of 24 cases agreed with the result of hearing judgment.

The data suspected of leaking was narrowed down to less than half by the above judgment algorithm. In addition, it was possible to determine the data for which water leakage was determined in advance by hearing with a high accuracy of 80% or more.

In the case of the combined determination of the first determination means and the second determination means (logical product), the data having the possibility of water leakage could be narrowed down most effectively.

The output unit 5 is a place where the determination result of the determination unit 4 is output. The output unit 5 may be a display or the like that displays figures and tables. In addition, the display of a mobile terminal such as a personal computer, a laptop computer, or a mobile phone, a tablet terminal, or the like may be used. When displaying on a mobile terminal or a tablet terminal, the determination result data may be received and displayed using the Internet, Wi-Fi, Bluetooth, or the like.

The display of the output unit 5 is displayed as table data as shown in FIG. For example, each determination means may be displayed in the column direction, and the number of water leakage determinations, comparison data with the hearing determination, and the like may be displayed in the row direction. Further, instead of the table data, a pie chart, a bar graph, or the like may be displayed. Further, the threshold value of the designated unit 6 may be specified on the display or the like of the output unit 5.

By using the determination algorithm of the present embodiment, it is possible to efficiently and accurately perform the water leakage determination work, which requires extremely labor in the auditory sense determination.

(Second Embodiment)
FIG. 14 is a block diagram showing a detailed configuration of a determination unit of the water leakage determination device 1 according to the second embodiment.

The leak determination device according to the second embodiment includes a third determination unit in the determination unit. Other configurations are the same as those of the water leakage determination device of the first embodiment.

The third determination means detects the presence or absence of repeated sounds of a fixed period in the sound wave signal data of the steady section extracted by the extraction unit. This is effective in discriminating sound waves that are likely to be erroneously determined as water leakage sounds.

FIG. 15 compares the time-series waveforms and frequency characteristics of the sound of water leakage and the sound of water used (sound when using water supply).

FIG. 15A shows a time-series waveform of the water leakage sound and its frequency characteristics. FIG. 15B shows a time-series waveform of the water sound used and its frequency characteristics.

In terms of frequency characteristics, both the water leakage sound and the water sound used have a characteristic that the frequency component up to 3 kHz is included and the component of 1 kHz or more and 2 kHz or less contributes greatly, so that it is difficult to distinguish the water sound from the water sound used.

On the other hand, in the time-series waveform, the water leakage sound is a time-series waveform such as white noise (random noise), whereas the water sound used is characterized by a repetitive sound having a fixed cycle. The reason why the water used has such a characteristic is that such a sound is likely to be superimposed because the meter of the water meter, which is the recording point, rotates when the water supply is used.

To extract the repetitive sound of this fixed cycle, a method of detecting the time-series waveform of the meter rotation sound stored in advance by comparing it with the time-series waveform of the input sound signal data, or for the input sound signal data. , A method of extracting by examining the similarity of waveforms by dividing them into arbitrary window lengths. The amount of overlap of window lengths in this case can be set arbitrarily. As a result, sound types that cannot be distinguished from each other in terms of frequency characteristics can be determined by time-series waveforms. Further, it may be determined from the frequency at which repeated sounds appear from the sound wave signal data at regular time intervals. For example, when the repeated sound appears five times or more from the sound wave signal data for 5 seconds, it may be used as the water sound.

The third determination means may be applied not to the sound wave signal data of the stationary section extracted by the extraction unit but to the sound wave signal data before being extracted by the extraction unit. This is because it is conceivable that the repeated signal is removed via the extraction unit.

The third determination means of the determination unit may be used alone. Further, it may be used in combination with the first determination means or the second determination means according to the first embodiment. It is preferable that the third determination means is applied to the sound wave signal data that could not be determined as water leakage even by the first and second determination means. For example, when it is used in a combined determination with the first and second determination means, it is preferably used in the first or last flow of the series determination.

As a combination of the first to third determination means, the pattern of FIG. 16 can be considered.

As shown in FIG. 16, in the case of series determination, there are 12 combinations (in the case of a combination of two determination means + a combination of three determination means).

In the case of parallel determination, there are four combinations (in the case of a combination of two determination means + a combination of three determination means). Further, considering the judgment by the logical sum and the logical product, there are 11 ways.

There are 12 patterns in which the series determination and the parallel determination are combined.

As described above, in the determination using the first to third determination means, a total of 38 combinations can be considered including the single determination. It is preferable to select a combination of determination means according to the situation in which the leakage determination is performed. For example, when the number of data to be determined is large and it is desired to narrow down to as few as possible, the first to third determination means may be applied in parallel to output the logical product of the determination results. In the case of an area where tap water is frequently used in advance, the third determination means is applied first, and then the first and second determination means are applied in parallel, and the logical product of the result is output. You may.

Although it has been explained that the threshold values used in the first determination means and the second determination means are determined in advance by the user, they may be automatically set in the apparatus. In that case, it is better to set using a database that stores a plurality of sound wave signal data at the time of past water leakage. For example, the threshold value of each determination means is calculated back by applying the algorithm (first and second determination means) according to the present embodiment to each of the past sound wave signal data. The back-calculated threshold value, frequency characteristics, and time-series waveform after HPF are stored in the database in a state of being associated with past sound signal data. If there is a waveform similar to the data stored in the database for the sound signal data input from the reception unit, the threshold value associated with the similar waveform is applied. If there is no similar waveform, the average value of the thresholds in the database is applied. Thereby, the threshold value can be set for each sound wave signal data.

For the determination of waveform similarity, the correlation coefficient may be calculated, or general machine learning may be used. Machine learning is logic built using algorithms that solve classification problems. Examples of these algorithms include decision trees, random forests, SVMs (Support Vector Machines), neural networks, and the like. An algorithm that combines these algorithms may be used. The database is stored in a hard disk, a USB memory, a ROM, or the like. Further, the database may be stored in an external server or the like, and the sound wave signal data at the time of past water leakage may be acquired by accessing the database via the Internet or the like.

(Third Embodiment)
FIG. 17 is a block diagram showing a detailed configuration of the determination unit 4 of the water leakage determination device 1 according to the third embodiment.

The leak determination device according to the third embodiment includes a fourth determination means 10 in the determination unit 4. Other configurations are the same as those of the water leakage determination device of the first embodiment.

The fourth determination means 10 calculates an autocorrelation function for the sound signal data (also referred to as a time series waveform) of the stationary section extracted by the extraction unit 3 and compares it with the threshold value specified by the designation unit 6. To determine the presence or absence of water leakage. Specifically, the envelope of the autocorrelation function is obtained, and a predetermined line segment and the envelope are compared. The rate at which the envelope is larger than the predetermined line segment is obtained and compared with the threshold value designated by the designated unit 6. For example, when the ratio is larger than the threshold value, it is determined as sound wave signal data with water leakage. Predetermined line segments and threshold values are arbitrarily determined based on empirical rules, past leak data, and the like. The predetermined line segment is not necessarily limited to a straight line and includes a curved line and the like. Further, the comparison with the threshold value is not limited to the ratio, and may be the area formed by the envelope or the slope of the envelope. The determination unit of FIG. 17 may include a third determination means.

The autocorrelation function evaluates the degree of coincidence of the signal data of two sections cut out from the same sound wave signal data. The presence or absence of water leakage is determined by evaluating the degree of coincidence of the periodic components included in the sound wave signal data in the steady section.

An envelope is a line segment that shares a tangent with a plurality of curves, and takes the shape of a straight line or a curve depending on the type of tangent curve. For example, the envelope of a plurality of curves having vertices includes a line segment connecting the respective vertices.

The following equation (4) is an equation showing an autocorrelation function. x (n) is the input signal, r (m) is the autocorrelation function, and N is the autocorrelation data length.

The determination procedure of the fourth determination means 10 will be described in detail with reference to FIG.

FIG. 18 shows an example of a detailed block diagram of the fourth determination means 10.

As shown in FIG. 18, first, the autocorrelation function of the sound wave signal data in the steady section is obtained. Next, the absolute value of the obtained autocorrelation function is taken, and the calculation range is limited to the positive time direction. Next, we derive the envelope for the autocorrelation function. The Hilbert transform is generally used to calculate the envelope. Next, a straight line approximation is performed on the envelope. When this approximate straight line is regarded as a linear function, it is preferably obtained by the least squares method. Leakage is determined based on the obtained approximate straight line.

FIG. 19 shows an example of the determination procedure when the test signal is used.

Here, a 50 Hz sine wave was used as the signal of the sound wave signal data in the steady section. FIG. 19A shows a time-series waveform of a 50 Hz sine wave. FIG. 19B is a graph of the autocorrelation function derived based on the equation (4). FIG. 19C is a graph showing the absolute value of the autocorrelation function. Furthermore, it is limited to the graph of time in the positive direction. As shown in FIG. 19 (d), the graph of the time in the positive direction tends to have a downward slope and converge to 0 as the time increases. In addition, an envelope is derived for this graph. Further, an approximate straight line of this envelope is taken. FIG. 19 (e) is a graph showing the waveform of the autocorrelation function and the approximate straight line of the envelope. Since the graph shape is targeted in the positive and negative time directions, the graph in the negative time direction may be used as the calculation range. In that case, since the start point changes from 0 on the time axis in the negative direction, the slope of the graph is downward to the left. In the case of a single-frequency time-series waveform as shown in FIG. 19 (a), the shape of the obtained envelope and the approximate straight line are almost the same, but the actual time-series waveform has a fluctuating sound with different amplitudes. , Since it has signals of a plurality of frequency components, the shape of the envelope and the approximate straight line are generally different.

FIG. 20 shows an example of the determination procedure in the case of a time-series waveform including fluctuating sounds. As shown in FIG. 20A, the time-series waveform including the fluctuating sound is a time-series waveform in which the amplitude locally increases in a predetermined time. 20 (b) and 20 (c) are signals obtained by obtaining the autocorrelation function of this time series waveform and taking an absolute value. FIG. 20D is the result of deriving the envelope of the waveform of FIG. 20C. As shown in FIG. 20 (d), the envelope in this case has a curved shape, and therefore has a shape different from the approximate straight line of the envelope shown in FIG. 20 (e).

FIG. 21 shows an example of a time series waveform when signals of a plurality of frequency components are included.

FIG. 21 (a) shows a 50 Hz sine wave, and FIG. 21 (b) shows a time series waveform of a 1 kHz sine wave. FIG. 21 (c) is a time-series waveform in which a 50 Hz sine wave (amplitude ratio 50%) and a 1 kHz sine wave (amplitude ratio 20%) are superimposed.

Further, FIG. 21D is a time-series waveform of a random sound which is white noise. FIG. 21 (e) is a waveform in which white noise and a 50 Hz sine wave (amplitude ratio 50%) are superimposed, and FIG. 21 (f) is a waveform in which white noise and a 50 Hz sine wave (amplitude ratio 20%) are superimposed. It is a waveform.

22 (a) to 22 (f) are obtained by deriving the autocorrelation function of the time series waveforms of FIGS. 21 (a) to 21 (f), their respective envelopes, and approximate straight lines. As shown in FIGS. 22 (a) and 22 (b), the time-series waveforms in FIGS. 21 (a) and 21 (b) have a beautiful autocorrelation function, whereas the white in FIG. 22 (d). In the time series waveform of noise, the value of the autocorrelation function becomes remarkably small, and the slope of the envelope becomes almost zero. When two sine waves having different frequencies are superimposed as shown in FIG. 22 (c), the tendency is downward to the right. When the sine waves and white noise of FIGS. 21 (e) and 21 (f) are superimposed, the magnitude of the slope of the graph changes depending on the mixing ratio of the amplitudes.

As described above, the water leakage sound has a characteristic of white noise (random sound). Further, the sound of septic tanks and the sound of motors of vending machines, outdoor units, etc. are dominated by periodic components such as sine waves.

As a result, it is possible to easily determine whether or not a peak sound unrelated to the water leakage sound is included by obtaining the autocorrelation function even in the sound wave signal data of the steady section actually extracted by the extraction unit 3. In addition, the graph makes it possible to visually distinguish the peak sound.

Next, the result of calculating the autocorrelation function by the fourth determination means 10 with respect to the actually measured sound wave signal data will be described.

FIGS. 23 (a) to 23 (d) show the actually measured sound wave signal data and the result of obtaining the autocorrelation function thereof.

As shown in FIG. 23A, the autocorrelation function of the sound signal data showing a single peak sound has an envelope or an approximate straight line having a downward slope. Further, the autocorrelation function of the sound wave signal data on which the plurality of peak sounds of FIG. 23B are superimposed results in the envelope converging to zero while fluctuating. Further, the autocorrelation function of the sound wave signal data showing the water leakage sound in FIG. 23 (c) results in the slope of the envelope being close to zero. Further, in the autocorrelation function of the motor sound and the sound wave signal data indicating the use of water supply in FIG. 23 (d), the slope of the envelope becomes small. These results are substantially similar to the results of the test signal.

Summarizing the above results, in the case of single-peak sound sound signal data, the slope of the autocorrelation function envelope (or an approximate straight line of the envelope) is relatively large, and the white noise sound signal indicating water leakage sound. In the case of data, the slope of the envelope of the autocorrelation function is small. In the case of sound wave signal data in which peak sounds and random sounds are superimposed, the magnitude of the slope of the envelope varies depending on the amplitude ratio of the two signals. In the case of sound wave signal data in which the water leakage sound and the peak sound are superimposed, an arbitrary threshold value may be set and the water leakage determination process may be performed.

Next, a method of determining a predetermined line segment and a threshold value to be compared with the envelope of the autocorrelation function (or an approximate straight line of the envelope) and performing the determination process will be described in detail.

FIG. 24 is a diagram showing an example of a determination process in which a predetermined line segment and a threshold value are determined. In FIG. 24, the predetermined line segment is represented by black (solid line), and the envelope is represented by gray (solid line). Judgment is made at a rate at which the envelope is larger than the predetermined line segment.

Since the horizontal axis of the graph in FIG. 24 indicates the time for 1 second as a whole, if the slope of the predetermined line segment is X (X <0) and 0 after 1 second, the intercept is −X. In FIG. 24, the threshold value is set to 20%, and the judgment condition is set so that if the proportion of the envelope larger than this predetermined line segment (time integration rate) is 20% or less, there is a possibility of water leakage instead of a peak sound. .. The ratio at which the envelope is larger than this predetermined line segment is the time integration rate, and is represented by, for example, the following equation (5).

Time integration rate% = time when the envelope crosses a predetermined line segment / total time (time until it converges to 0) x 100 ... (5)
In the cases of FIGS. 24 (a) and 24 (b), the time integration rates are 94.5% and 69.6%, which exceeds the threshold value of 20% specified by the designated unit 6, and therefore the sound wave signal of the peak sound. It is judged as data. In the cases of FIGS. 24 (c) and 24 (d), since the time integration rate is 1.285% and 0.6%, which is lower than the threshold value of 20%, the determination result is suspected to be a water leakage sound.

The determination of water leakage in comparison with the threshold value is not limited to the time integration rate, and may be determined by time-integrating the envelope of the autocorrelation function and a predetermined line segment and comparing their magnitudes. This means comparing the area created by each of the envelope and a given line segment.

FIG. 25 shows the result of comparing the size of the area formed by each of the envelope and the predetermined line segment.

When the area formed by the envelope is larger than the area formed by the predetermined line segment, it is determined as the sound wave signal data of the peak sound instead of the water leakage sound. If the area created by the envelope is smaller than the area created by the predetermined line segment, it is determined that there is a suspicion of water leakage. In this case, the area formed by the predetermined line segment is specified as the threshold value.

In addition, the water leakage may be determined by comparing with the threshold value by comparing the slope of the envelope (or an approximate straight line of the envelope) with the slope of a predetermined line segment. In this case, the slope of a predetermined line segment is specified as the threshold value.

The above-mentioned time integration rate derived based on the envelope of the autocorrelation function, the area created by the envelope, and the slope of the envelope are also referred to as judgment values.

FIG. 26 shows the result when the water leakage is determined from the autocorrelation function based on any of the time integration rate, the area formed by the envelope, or the slope of the envelope. For 103 sound wave signal data that the investigator made the primary judgment by the detector, the judgment was made based on either the time integration rate, the area formed by the envelope, or the inclination of the envelope. The leak data of the answers was 24 out of 103 cases that were judged to be water leaks by hearing. As a result, the number of narrowing down judgments using the time integration rate was 45, the number of narrowing down using the area created by the envelope was 43, and the number of narrowing down using the slope of the envelope was 44. .. The number of correct answers included in the narrowed-down sound wave signal data was 21 (out of 24), and in each case, a high correct answer rate was shown. When the water leakage was determined using the algorithm of the fourth determination means, there was almost no difference even if the water leakage was determined based on any of the time integration rate, the area formed by the envelope, and the inclination of the envelope. It may be appropriately selected according to the situation of determining water leakage and the data acquired in the past.

By using the fourth determination means, it is possible to narrow down the data having a possibility of water leakage sound from a large number of sound wave signal data with a high probability.

Further, by combining with the first to third determination means according to the first or second embodiment, data having a possibility of water leakage can be narrowed down from a large number of sound wave signal data with a higher probability.

As the combination of the first, second or fourth determination means, the pattern of FIG. 27 can be considered.

As shown in FIG. 27, in the case of series determination, there are 12 combinations (in the case of a combination of two determination means + a combination of three determination means).

In the case of parallel determination, there are four combinations (in the case of a combination of two determination means + a combination of three determination means). Further, considering the judgment by the logical sum and the logical product, there are 11 ways.

There are 12 patterns in which the series determination and the parallel determination are combined.

As described above, in the determination using the first, second, or fourth determination means, a total of 38 combinations can be considered including the single determination.

As the combination of the second, third, or fourth determination means, the pattern of FIG. 28 can be considered.

The combination of the second, third, or fourth determination means is also a combination of patterns similar to the combination of the first, second, or fourth determination means.

As the combination of the first, third, or fourth determination means, the pattern of FIG. 29 can be considered.

Further, the first to fourth determination means may be combined. When combining the four determination means, the pattern of FIG. 30 can be considered.

As shown in FIG. 30, there are 24 combinations of series determination. In the case of parallel judgment, there are 13 combinations (excluding combinations that take a further logical product of logical products). There are 112 combinations of series determination and parallel determination.

The pattern of these combinations can be appropriately selected depending on the situation in which the water leakage is determined. For example, when it is desired to narrow down the data having a possibility of water leakage from a huge amount of sound wave signal data to the maximum, the sound wave signal data may be narrowed down by connecting the first to fourth determination means in series. Further, the logical product of the determination results of the first to fourth determination means may be obtained. Further, the logical product of the determination results of the two logical products excluded from the combination described above may be further taken.

Further, although it has been explained that the threshold value used in the fourth determination means is determined in advance by the user, it may be automatically set in the apparatus. In that case, the same method as that described in the first and second determination means can be adopted.

The above-described embodiment is not limited to the leak determination device and can be used as a system. For example, a server that includes a water leakage determination algorithm (extractor, determination unit) of the present embodiment for a plurality of sound wave signal data measured by a water leak inspection investigator of a water pipe or the like with a detector that detects sound waves (vibration sound). Transfer to via the Internet etc. The server uses the transferred sound wave signal data to determine water leakage. The server transfers the leak determination result to the investigator's detector. As a result, the investigator can determine water leakage only with a detector equipped with a communication means such as the Internet without carrying a high-performance and large-scale device.

Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other embodiments, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are also included in the scope of the invention described in the claims and the equivalent scope thereof.

1 Leakage determination device 2 Reception unit 3 Extraction unit 4 Judgment unit 5 Output unit 6 Designation unit 7 First judgment means 8 Second judgment means 9 Third judgment means 10 Fourth judgment means

Claims (10)

The reception desk that accepts sound wave data and
An extraction unit that extracts data in a steady section with less noise from the sound wave data,
A designated part that specifies the first threshold value and
It is a determination unit that determines the presence or absence of water leakage, obtains the ratio of frequency components caused by water leakage in a predetermined frequency range from the data of the steady section, and compares the ratio with the first threshold value. As a result, a determination unit provided with a first determination means for determining the presence or absence of water leakage,
Leakage determination device including. The first determination means, as the ratio, sets the entire frequency range, which is the predetermined frequency range, and an arbitrary frequency range, with respect to the difference between the frequency characteristic of the data of the stationary section and the frequency characteristic of background noise. Find the integration rate of
The water leakage determination device according to claim 1. The designation unit specifies a second threshold value and
The determination unit has a filter for reducing the low frequency component of the data in the steady section, obtains a time integration rate of a predetermined amplitude or more with respect to the data in the steady section to which the filter is applied, and obtains the time integration rate. Is further provided with a second determination means for determining the presence or absence of water leakage by comparing the above with the second threshold value.
The water leakage determination device according to claim 1 or 2. The determination unit further determines the presence or absence of water leakage by the second determination means with respect to the sound wave data determined by the first determination means that there is water leakage.
The water leakage determination device according to claim 3. The determination unit further determines the presence or absence of water leakage by the first determination means with respect to the sound wave data determined by the second determination means that there is water leakage.
The water leakage determination device according to claim 3. The designation unit specifies a third threshold value and
The determination unit detects the frequency of the repetition signal of a fixed cycle with respect to the data of the steady section, and compares the frequency with the third threshold value, so that the sound wave data includes a signal indicating the water sound used. Further provided with a third determination means for determining whether or not.
The water leakage determination device according to any one of claims 1 to 5. The designation unit specifies a fourth threshold value and
The determination unit obtains an autocorrelation function of the data of the stationary interval, compares the envelope connecting the curves of the autocorrelation function with a predetermined line segment, determines the area formed by the envelope, or the slope of the envelope. A fourth determination means for determining the presence or absence of water leakage is further provided by deriving a determination value based on the determination value and comparing the determination value with the fourth threshold value.
The water leakage determination device according to any one of claims 1 to 6. The water leakage determination device according to any one of claims 1 to 7, further comprising an output unit that outputs the determination result of the determination unit. The output unit outputs a logical sum or a logical product of the determination results of the first determination means and the second determination means.
The water leakage determination device according to claim 8, which cites claim 3. It is a leak judgment method executed by a leak judgment device.
Extract the data of the steady section with less noise from the sound wave data,
Specify a threshold for determining the presence or absence of water leakage,
For the data of the steady section, the ratio of the frequency component due to water leakage in the predetermined frequency range was calculated.
A water leakage determination method for determining the presence or absence of water leakage by comparing the ratio with the threshold value.

Images