KR20220105039A - Apparatus And Method For Remotely Detecting Pipe Leakage In A Non-contact and Non-destructive Manner Through Ultrasonic Sound - Google Patents

Abstract

The apparatus for remotely detecting a pipe leakage comprises: an acoustic sensor collecting an acoustic signal; a signal processing unit calculating a power spectrum having FFT lines by frequency from the acoustic signal; and a control unit detecting a pipe leakage by analyzing the power spectrum having the FFT lines by frequency. The apparatus remotely detects a pipe leakage in a non-contact and non-destructive method through acoustic sound.

Description

Apparatus And Method For Remotely Detecting Pipe Leakage In A Non-contact and Non-destructive Manner Through Ultrasonic Sound

Pipe leak remote detection technology is provided. More particularly, an apparatus and method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound are provided.

Plumbing fixtures are blood vessels, for example in humans. Bleeding, that is, leakage of the internal medium of the pipe, has a great influence on the operation of the equipment. Considering the loss of leakage fluid or gas, environmental destruction, etc., the effect is very large. For example, the leakage of highly corrosive media (fluids, gases) and very hazardous substances such as Na has a huge impact on the environment. In addition, a leaking accident of flammable high-pressure gas has a risk of causing a secondary disaster, so a detection system that can detect the leak early is required.

Korean Patent Publication No. 2009-0122131 published on November 26, 2009 (Title: Leak monitoring device using image and sound information)

One embodiment of the present invention is to provide an apparatus and method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound.

An apparatus for remotely detecting a pipe leak according to an embodiment of the present invention includes an acoustic sensor for collecting an acoustic signal, a signal processing unit for deriving a power spectrum having a FFT line for each frequency from the acoustic signal, and an FFT line for each frequency and a control unit for detecting a pipe leak by analyzing a power spectrum having

The signal processing unit includes a filter that removes signals in a band other than a predetermined ultrasound band from among the acoustic signals and outputs an acoustic signal in a predetermined ultrasound band, a waveform detection unit that converts an acoustic signal in the ultrasound band into a digital acoustic signal, and a digital acoustic signal It may include a digital signal processing unit for deriving a power spectrum having an FFT line for each frequency by performing FFT transformation on the .

When a power spectrum having an FFT line for each frequency is input, the control unit may include an artificial neural network that outputs a probability of whether the pipe is leaked in response to the power spectrum, and a detection module that determines whether the pipe is leaked according to the probability. have.

The control unit inputs training data, which is a power spectrum including a label indicating whether leakage, into the neural network, and when the neural network calculates an output value for the training data, the binary cross entropy loss between the output value and the label is minimized. It may include a learning module for optimizing the weight of .

A method for remotely detecting a pipe leak according to an embodiment of the present invention includes the steps of: an acoustic sensor collecting an acoustic signal; a signal processing unit deriving a power spectrum having an FFT line for each frequency from the acoustic signal; and detecting a pipe leak by analyzing a power spectrum having an FFT line for each frequency, and remotely detecting the pipe leak in a non-contact and non-destructive manner through ultrasonic sound.

The step of deriving the power spectrum includes, by the filter of the signal processing unit, removing a signal in a band other than a predetermined ultrasound band from among the acoustic signals and outputting an acoustic signal in a predetermined ultrasound band, and a waveform detecting unit of the signal processing unit in the ultrasound band. It may include converting the sound signal into a digital sound signal, and the digital signal processing unit of the signal processing unit performing FFT conversion on the digital sound signal to derive a power spectrum having FFT lines for each frequency.

The step of analyzing the power spectrum to detect the pipe leakage includes: when the artificial neural network of the control unit receives a power spectrum having an FFT line for each frequency, outputting a probability of whether the pipe leaks in response to the power spectrum; The detection module may include determining whether the pipe is leaking according to the probability.

A method for remotely detecting a pipe leak includes, before the acoustic sensor collects an acoustic signal, inputting, by a learning module of the control unit, learning data, which is a power spectrum including a label indicating leakage, into an artificial neural network; Calculating an output value for the training data, and the learning module may include optimizing the weight of the artificial neural network so that the binary cross entropy loss between the output value and the label is minimized.

A method for remotely detecting a pipe leak according to an embodiment of the present invention includes the steps of: mapping a background signal measured by a plurality of detection devices installed at a plurality of different locations with respect to a pipe facility and an installation location; calculating the ratio of S/N of the leakage signal to the background signal measured by each detection device of remotely detecting pipe leaks in a non-contact and non-destructive manner via ultrasonic sound.

According to an embodiment of the present invention, it is possible to provide an apparatus and a method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound.

1 is a view for explaining the configuration of an apparatus for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention.
2 is a graph for explaining a power spectrum for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention.
3 is a diagram for explaining the configuration of an artificial neural network according to an embodiment of the present invention.
4 is a flowchart illustrating a method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention.
5 is a flowchart illustrating a learning method of an artificial neural network according to an embodiment of the present invention.
6 is a flowchart illustrating a method of detecting a pipe leakage by analyzing a power spectrum through an artificial neural network according to an embodiment of the present invention.
7 and 8 are diagrams for explaining a method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention.
9 and 10 are graphs for explaining a method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention.

Prior to the detailed description of the present invention, the terms or words used in the present specification and claims described below should not be construed as being limited to their ordinary or dictionary meanings, and the inventors should develop their own inventions in the best way. It should be interpreted as meaning and concept consistent with the technical idea of the present invention based on the principle that it can be appropriately defined as a concept of a term for explanation. Accordingly, the embodiments described in this specification and the configurations shown in the drawings are only the most preferred embodiments of the present invention, and do not represent all the technical ideas of the present invention, so various equivalents that can replace them at the time of the present application It should be understood that there may be water and variations.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In this case, it should be noted that the same components in the accompanying drawings are denoted by the same reference numerals as much as possible. In addition, detailed descriptions of well-known functions and configurations that may obscure the gist of the embodiments of the present invention will be omitted. For the same reason, some components are exaggerated, omitted, or schematically illustrated in the accompanying drawings, and the size of each component does not fully reflect the actual size.

First, a configuration of an apparatus for remotely detecting pipe leakage in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention will be described. 1 is a view for explaining the configuration of an apparatus for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention. 2 is a graph for explaining a power spectrum for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention. 3 is a diagram for explaining the configuration of an artificial neural network according to an embodiment of the present invention.

Referring to FIG. 1 , an apparatus 100 for remotely detecting a pipe leak according to an embodiment of the present invention remotely detects a pipe leak through ultrasonic waves. When a leak occurs due to a defect in the high-pressure pipe, for example, a perforation, crack, or the like, ultrasonic waves are generated. When the fluid or gas inside the high-pressure pipe flows out into the air, the pressure decreases and expands to generate an elastic wave. The leaked gas becomes turbulent at the outlet to generate ultrasonic waves. Accordingly, the detection device 100 is a gas, for example, a leak rate of the gas is 1x10 ^-3 std. When it is more than cc/sec, ultrasonic waves generated due to turbulence can be detected. The detection device 100 detects the ultrasonic sound signal as described above, and remotely detects whether the pipe is leaking in a non-contact and non-destructive manner.

To this end, the detection device 100 includes an acoustic sensor 110 , a signal processing unit 120 , and a control unit 130 . Also, the control unit 130 of the detection device 100 may further include a communication interface 140 , and may be connected to the data logger 200 through the communication interface 140 .

The acoustic sensor 110 collects acoustic signals. For example, the acoustic sensor 110 may be implemented through a micro-microphone of a MEMS (Micro-Electro Mechanical Systems) structure.

The signal processing unit 120 derives a power spectrum having an FFT line for each frequency of the ultrasound band from the acoustic signal collected by the acoustic sensor 110 . To this end, the signal processing unit 120 may include an amplifier 121 , a filter 123 , a waveform detector 125 , and a digital signal processor 127 .

The amplifier 121 amplifies the acoustic signal collected by the acoustic sensor 110 .

The filter 123 filters the amplified sound signal to output only a sound signal of a predetermined ultrasonic band, for example, 20 kHz to 100 kHz. For example, the filter 123 performs filtering to remove a signal in a band other than a predetermined ultrasonic band among the amplified sound signals, for example, an audible frequency band of 20 kHz or less. According to this filtering, the filter 123 outputs an acoustic signal of a predetermined ultrasonic band.

The waveform detector 125 converts an acoustic signal of an ultrasonic band into a digital acoustic signal. For example, the sound signal in the ultrasonic band filtered by the filter 123 is an analog sound signal, and the waveform detector 125 converts this analog sound signal in the 20 kHz to 100 kHz band into a 256 kHz sampling frequency fs to convert the digital sound signal. create And the waveform detector 125 outputs the generated digital sound signal.

The digital signal processor 127 derives a power spectrum having FFT lines for each frequency by performing Fast Fourier Transform (FFT) on the digital sound signal. In this case, the block length (the number of sampling points) N for the FFT is 1,024. In addition, the FFT calculation time τ and the frequency resolution Δf are calculated through Equations 1 and 2 below.

[Equation 1]

τ≤ N/f _s ≒ 4 ms

[Equation 2]

Δf ≤ f _s /N = 250 Hz

In this way, if a power spectrum is obtained by frequency-converting a digital sound signal to a frequency resolution of 250Hz through an ultrasonic band (20kHz to 100kHz) through FFT, multiple An FFT line is formed. For example, as shown in FIG. 2 , a power spectrum having 320 FFT lines may be derived because the band of the digital sound signal is 20 kHz to 100 kHz and the frequency resolution is 250 Hz. Here, the X axis represents the frequency (kHz), and the Y axis represents the power spectrum.

The control unit 130 analyzes a power spectrum having an FFT line for each frequency of the ultrasonic band to detect a pipe leak. For example, the control unit 160 analyzes 320 FFT lines (power spectrum) in the 20 kHz to 100 kHz band generated by the signal processing unit 150 to detect whether the pipe is leaking.

When high-pressure gas leaks from a pipe, it is almost impossible to detect pipe leakage with a sound level meter in the audible frequency range under a mechanical noise environment. Therefore, according to one embodiment, the audible sound region of 20 kHz or less among the leakage acoustic frequency region is removed, and the power spectrum is obtained in the ultrasonic band in the range of 20 kHz to 100 kHz, where it is relatively easy to distinguish the leak sound from the background noise including the mechanical noise. Analyze to detect pipe leaks. To this end, the controller 130 may include an artificial neural network 131 , a learning module 133 , and a detection module 135 .

The artificial neural network 131 includes a plurality of layers, and each of the plurality of layers includes a module for performing at least one operation. For example, the artificial neural network 131 has a plurality of layers composed of a plurality of operations, and results of operations between the plurality of layers are transmitted. At this time, the transmission of the result of the operation is transmitted by weighting.

Referring to FIG. 3 , the artificial neural network 131 includes an input layer 10, first and second convolution layers 20 and 40, and first and second pooling layers 30 and 50: pooling. layer), a fully-connected layer (60) and an output layer (70: output layer). The first and second convolutional layers 20 and 40 perform a convolution operation, the first and second pooling layers 30 and 50 perform a pooling operation, and the fully connected layer 60 operates by an activation function. carry out

When a power spectrum having an FFT line for each frequency is input, the artificial neural network 131 according to an embodiment of the present invention calculates the probability of leakage of the pipe through a plurality of calculations in which a plurality of layer weights are applied to the power spectrum. It is learned to calculate and output the calculated probability as an output value. For example, when a power spectrum is input, the artificial neural network 131 that has been trained calculates the probability of whether the pipe is leaked through a plurality of calculations to which the weights of a plurality of layers are applied, and determines whether the calculated pipe is leaked. Output the probability for

More specifically, when a power spectrum is input to the input layer 10, a convolution operation using the filter (kernel: W) of the first convolution layer 20 is performed, and the result is the first pooling layer 30 . is input to, and the pooling operation of the first pooling layer 30 using the filter W is performed on the inputted result value, and the result is input to the second convolutional layer 40 . Subsequently, a convolution operation using the filter W of the second convolution layer 40 is performed on the input result value, the result is input to the second pooling layer 50, and the second The pooling operation of the pooling layer 50 is performed and the result is input to the fully connected layer 60 . Then, an operation by the activation function of the fully connected layer 60 is performed, and the result is input to the output layer 70 and output as an output value. The above-described operation of each layer is weighted and transmitted to the next layer. The output layer 70 has two nodes O1 and O2, the first node O1 corresponds to the probability that the pipe has leaked, and the second node O2 corresponds to the probability that the pipe has not leaked.

The learning module 133 is for learning the artificial neural network 131 . The learning module 133 learns the artificial neural network 131 using the learning data. The training data is a power spectrum that includes a label indicating whether or not there is a leak. The learning module 133 inputs the training data, which is the power spectrum, to the artificial neural network 131 . Accordingly, the artificial neural network 131 will calculate an output value through a plurality of calculations to which the weights of a plurality of layers are applied to the training data. Then, the learning module 133 learns the artificial neural network 131 by optimizing the weights of the artificial neural network 131 so that the binary cross entropy loss between the output value of the artificial neural network 131 and the label is minimized. The learning method of the learning module 133 will be described in more detail below.

The detection module 135 detects pipe leakage according to the output of the learned artificial neural network 131 . The detection module 135 inputs the power spectrum generated by the signal processing unit 120 to the artificial neural network 131 , and then obtains an output value of the artificial neural network 131 for the input power spectrum. The output value of the artificial neural network 131 is the probability of pipe leakage, and the detection module 135 determines that the pipe leakage has occurred when the probability is greater than or equal to a preset threshold. For example, when the value of the first node O1 of the artificial neural network 131 is greater than the value of the second node O2 and greater than or equal to the threshold, the detection module 135 may determine that the pipe is leaking. For example, assuming that the threshold is 75%, for example, if the values of the first and second nodes (O1, O2) are (O1, O2) = (0.8176, 0.1824), the probability of leakage is 82%, and there is no leakage. Since there is an 18% probability that it was not, it can be judged that it was leaked. As another example, if the values of the first and second nodes (O1, O2) are (O1, O2) = (0.7211, 0.2789), the probability of leakage is 72%, and the probability of not leaking is 28%. Although it is high, since it has a probability below the threshold, it can be determined that it is not leaked. The pipe leak detection method of the detection module 135 will be described in more detail below.

Next, a method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention will be described. 4 is a flowchart illustrating a method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention.

The acoustic sensor 110 receives the acoustic signal in step S110. Then, the amplifier 121 of the signal processing unit 120 amplifies the acoustic signal collected by the acoustic sensor 110 in step S120. Next, the filter 123 of the signal processing unit 120 filters the sound signal amplified in step S130 to output only a sound signal in a predetermined ultrasonic band, for example, 20 kHz to 100 kHz. For example, the filter 123 performs filtering to remove a signal in a band other than a predetermined ultrasonic band among the amplified sound signals, for example, an audible frequency band of 20 kHz or less. According to this filtering, the filter 123 outputs an acoustic signal of a predetermined ultrasonic band.

Next, the waveform detector 125 of the signal processing unit 120 converts the ultrasonic band sound signal into a digital sound signal in step S140 . For example, the sound signal in the ultrasonic band filtered by the filter 123 is an analog sound signal, and the waveform detector 125 converts this analog sound signal in the 20 kHz to 100 kHz band into a 256 kHz sampling frequency fs to convert the digital sound signal. create And the waveform detector 125 outputs the generated digital sound signal.

Next, the digital signal processor 127 of the signal processing unit 120 derives a power spectrum having an FFT line for each frequency by performing FFT on the digital sound signal in step S150 . For example, the block length (the number of sampling points) N for the FFT may be 1,024, and the FFT operation time τ and the frequency resolution Δf are calculated according to Equations 1 and 2 described above. In this way, when a power spectrum is obtained by frequency-converting a digital sound signal to an ultrasonic band (20 kHz to 100 kHz) with a frequency resolution of 250 Hz through FFT, a plurality of FFT lines are formed for each frequency according to the frequency resolution within the ultrasonic band. For example, as shown in FIG. 2 , a power spectrum having 320 FFT lines may be derived because the band of the digital sound signal is 20 kHz to 100 kHz and the frequency resolution is 250 Hz.

Next, the control unit 130 analyzes the power spectrum having the FFT line for each frequency of the ultrasonic band in step S160 to detect the pipe leakage. For example, the control unit 160 analyzes 320 FFT lines in the 20 kHz to 100 kHz band generated by the signal processing unit 150 to detect whether the pipe is leaking.

According to an embodiment, the controller 130 detects a pipe leak by analyzing a power spectrum having an FFT line for each frequency of the ultrasonic band using the artificial neural network 131 . This method will be described in more detail.

In order to use the artificial neural network 131, learning of the artificial neural network 131 should be preceded. Then, such a learning method will be described. 5 is a flowchart illustrating a learning method of an artificial neural network according to an embodiment of the present invention.

Referring to FIG. 5 , the learning module 133 prepares learning data in step S210 . The training data is a power spectrum that includes a label indicating whether or not there is a leak. For example, in a situation in which a pipe leaks, an acoustic signal may be received through the acoustic sensor 110 , and a signal obtained by extracting a power spectrum through the signal processing unit 120 may be used as learning data. In addition, as a control, an acoustic signal is received through the acoustic sensor 110 in a situation in which the pipe does not leak, and a signal obtained by extracting a power spectrum through the signal processing unit 120 may be used as learning data. The label for the learning data in the leaked situation is set as (O1, O2) = (1, 0) as the values of the first and second nodes, and the label for the learning data (control) in the non-leaked situation is the first The values of 1 and the second node will be set as (O1, O2)=(0, 1).

Next, the learning module 133 inputs the learning data to the artificial neural network 131 in step S220. Then, the artificial neural network 131 will calculate an output value for the training data by performing a plurality of operations of a plurality of layers to which a weight is applied on the training data. Then, the learning module 133 optimizes the weight of the artificial neural network 131 using an optimization algorithm so that the binary cross entropy loss between the output value of the artificial neural network 131 and the label is minimized in step S230. do. Learning may be completed by repeating steps S210 to S230 using a plurality of learning data.

As described above, when the learning of the artificial neural network 131 is completed, the pipe leakage may be detected using the artificial neural network 131 . This method will be described in more detail. 6 is a flowchart illustrating a method of detecting a pipe leak by analyzing a power spectrum through an artificial neural network according to an embodiment of the present invention.

Referring to FIG. 6 , the detection module 135 may receive diagnostic data in step S310 . The diagnostic data is a power spectrum having an FFT line for each frequency of an acoustic signal derived through the process shown in FIG. 4 above.

The detection module 135 inputs the power spectrum to the artificial neural network 131 in step S320. Then, the artificial neural network 131 will calculate the probability of whether the pipe leaks through a plurality of calculations to which the weights of a plurality of layers are applied corresponding to the power spectrum.

Then, the detection module 135 determines whether the pipe is leaked according to the probability calculated by the artificial neural network 131 in step S330. For example, assuming that the threshold is 75%, for example, if the values of the first and second nodes (O1, O2) are (O1, O2) = (0.8176, 0.1824), the probability of leakage is 82%, and there is no leakage. Since there is an 18% probability that it was not, it can be judged that it was leaked. As another example, if the values of the first and second nodes (O1, O2) are (O1, O2) = (0.7211, 0.2789), the probability of leakage is 72%, and the probability of not leaking is 28%. Although it is high, since it has a probability below the threshold, it can be determined that it is not leaked.

When gas leaks from a high-pressure pipe, the leakage sound is uniformly distributed in the entire frequency range including the audible frequency range. At the nearest distance, you can hear the leaking sound of the high-temperature pipe even in the audible frequency range of 20 kHz or less. However, since ambient noise such as mechanical noise such as a plant or living noise is very large, it is almost impossible to hear a leak sound at a distance out of a certain range. Therefore, according to an embodiment, the leak sound may be detected from a long distance in a non-contact manner using a signal of the ultrasonic band.

7 and 8 are diagrams for explaining a method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention. 9 and 10 are graphs for explaining a method for remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound according to an embodiment of the present invention.

According to an embodiment, as shown in FIG. 7 , when the data server cabinet NS is spaced apart from one side of the piping facility PS and the piping facility PS, a plurality of detection devices 100: 101 , 102 , 104 , 105 are arranged in four different orientations with respect to the plumbing fixture PS.

The data recording apparatus 200 includes an interface module 210 , a storage module 220 , and a control module 230 . The interface module 210 is a device for communicating with a plurality of detection devices (100: 101, 102, 104, 105). This communication may be a wired or wireless protocol. The storage module 220 maps and stores a background signal measured by a plurality of detection devices installed at a plurality of different locations with respect to a piping facility and an installation location. The control module 230 receives a leak signal from each of the plurality of detection devices 100: 101, 102, 104, and 105 through the interface module 210 and compares it with a pre-stored background signal to identify an area where the leak occurred. do. As described above, a method for specifying a region in which leakage occurs will be described in more detail.

The detection device 100: 101, 102, 104, 105 has different background (B/G, background) signals, as shown in the graph shown in FIG. 9 , according to the installation position or direction. Therefore, if the background signal and the installation location of the ultrasonic leak detector are mapped, and the S/N ratio difference of the signals detected by the detection devices (100: 101, 102, 104, 105) of each installed point is compared as shown in FIG. , it is possible to specify the installation area of the detection device (100: 101, 102, 104, 105) where the S/N ratio shows the largest value as the leakage occurring.

For example, referring to FIGS. 7 and 8 , it is assumed that air leaked to the 0.2mm nozzle corresponding to the first leak point #1 and the second leak point #2 under a pressure condition of 70 kPa. do. 9 shows background signals detected by the detection devices 100 ( 101 , 102 , 104 , 105 ) installed at four different locations. As such, it can be seen that the background signals detected by the detection devices 100 ( 101 , 102 , 104 , 105 ) installed at four different positions are all different.

In addition, as shown in the graph shown in FIG. 9, the background signal of the ultrasonic leak sound signal according to the installation position of the detection device 100: 101, 102, 104, 105 is different depending on the frequency (FFT line). It is possible to map the characteristics of the signal according to the frequency (FFT line) and the installation location.

10 is a second leak point (Leak point #2) when the air leaks from the detection device (100: 101, 102, 104, 105) installed at four different positions detected at each installation position detected. It represents the leakage signal obtained by measuring the acoustic signal in the ultrasonic band.

When comparing the background signal (AVG (B/G)) and the leakage signal (AVG (0.2 mm leak)) in FIG. 10, if the S/N ratio of the leakage signal to the background signal according to the installation location is obtained, the first and second 4 The S/N ratio of the leakage signal to the background signal detected by the detection devices 101 and 104 is 0.1 or less or a negative value, indicating that the background signal and the leakage signal cannot be distinguished, and the second and fifth detection devices ( 102, 105) detected the leakage signal to the S/N ratio of the background signal is calculated as 2.56dB, 0.27dB. As described above, from the comparison of the S/N ratio, it can be seen that leakage has occurred in the monitoring area of the second detection device 102 , which is a detection device where the S/N ratio has the largest value.

The storage module 220 of the data recording device 200 may store the respective installation positions of the plurality of detection devices 101 , 102 , 104 , 105 installed at a plurality of different positions with respect to the piping system PS. In this case, the control module 230 of the data recording device 200 may receive the background signal measured by each of the plurality of detection devices 101 , 102 , 104 , 105 through the interface module 210 . Then, the control module 230, as shown in the graph shown in FIG. 9, the background signal of the ultrasonic leakage acoustic signal according to the installation position of each of the plurality of detection devices (101, 102, 104, 105) in the frequency (FFT line) Since it is different from each other, the characteristics of the signal according to the frequency (FFT line) of the background signal and the installation location may be mapped, and this may be stored in the storage module 220 . In addition, the control module 230 of the data recording device 200 may receive the leakage signal measured by each of the plurality of detection devices 101 , 102 , 104 , 105 through the interface module 210 . Then, the control module 230 may calculate the S/N ratio of the leak signal measured by each of the plurality of detection devices 101 , 102 , 104 , and the stored background signal. Then, the control module 230 may specify the area monitored by the detection device 101 , 102 , 104 , or 105 having the largest calculated S/N ratio as the leaked area.

Meanwhile, the method according to the embodiment of the present invention described above may be implemented in the form of a program readable by various computer means and recorded in a computer readable recording medium. Here, the recording medium may include a program command, a data file, a data structure, etc. alone or in combination. The program instructions recorded on the recording medium may be specially designed and configured for the embodiment of the present invention, or may be known and available to those skilled in the art of computer software. For example, the recording medium includes magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floppy disks ( magneto-optical media), and hardware devices specially configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions may include high-level languages that can be executed by a computer using an interpreter or the like, as well as machine language such as generated by a compiler. Such hardware devices may be configured to operate as one or more software modules to perform the operations of the embodiments of the present invention, and vice versa.

The present invention has been described above using several examples, but these examples are illustrative and not restrictive. As such, those of ordinary skill in the art to which the present invention pertains will understand that various changes and modifications can be made in accordance with the doctrine of equivalents without departing from the spirit of the present invention and the scope of rights set forth in the appended claims.

110: sound sensor
120: signal processing unit
121: amplifier
123: filter
125: waveform detector
127: digital signal processor
130: control unit
131: artificial neural network
133: learning module
135: detection module
140: communication interface
200: data recording device

Claims (9)

an acoustic sensor for collecting acoustic signals;
a signal processing unit for deriving a power spectrum having an FFT line for each frequency from the sound signal; and
a control unit for detecting a pipe leak by analyzing a power spectrum having the FFT line for each frequency;
A device for remotely detecting a pipe leak, comprising: remotely detecting a pipe leak in a non-contact and non-destructive manner through ultrasonic sound. In claim 1,
The signal processing unit,
a filter for removing a signal in a band other than a predetermined ultrasound band from among the acoustic signals and outputting an acoustic signal in the predetermined ultrasound band;
a waveform detection unit converting the sound signal of the ultrasonic band into a digital sound signal; and
a digital signal processing unit for deriving a power spectrum having an FFT line for each frequency by performing FFT conversion on the digital sound signal;
A device for remotely detecting a pipe leak, comprising a. In claim 1,
The control unit is
an artificial neural network for outputting a probability of whether the pipe is leaking in response to the power spectrum having the FFT line for each frequency being input; and
a detection module for determining whether the pipe is leaking according to the probability;
A device for remotely detecting a pipe leak, comprising a. In claim 3,
The control unit is
Input the training data, which is a power spectrum including a label indicating whether leakage, to the artificial neural network,
When the artificial neural network calculates an output value for the learning data,
and a learning module for optimizing the weight of the artificial neural network so that the binary cross entropy loss between the output value and the label is minimized. Acoustic sensor collecting an acoustic signal;
deriving, by a signal processing unit, a power spectrum having an FFT line for each frequency from the sound signal; and
detecting, by a control unit, a power spectrum having the FFT line for each frequency to detect a pipe leak;
including,
A method for remotely detecting pipe leaks, comprising remotely detecting pipe leaks in a non-contact and non-destructive manner via ultrasonic acoustics. In claim 5,
The step of deriving the power spectrum is
removing, by a filter of the signal processing unit, a signal in a band other than a predetermined ultrasound band from among the acoustic signals and outputting an acoustic signal in the predetermined ultrasound band;
converting, by the waveform detecting unit of the signal processing unit, the sound signal of the ultrasonic band into a digital sound signal; and
deriving, by the digital signal processing unit of the signal processing unit, a power spectrum having an FFT line for each frequency by performing FFT conversion on the digital sound signal;
A method for remotely detecting a pipe leak, comprising: In claim 5,
Detecting the pipe leakage by analyzing the power spectrum
when the artificial neural network of the controller receives the power spectrum having the FFT line for each frequency, outputting a probability of whether the pipe is leaked in response to the power spectrum; and
determining, by the detection module of the control unit, whether the pipe is leaking according to the probability;
A method for remotely detecting a pipe leak, comprising: In claim 7,
Before the sound sensor collects the sound signal,
inputting, by the learning module of the control unit, learning data that is a power spectrum including a label indicating whether leakage is present to the artificial neural network;
calculating, by the artificial neural network, an output value for the training data; and
optimizing, by the learning module, the weights of the artificial neural network such that the binary cross entropy loss between the output value and the label is minimized;
Further comprising, a method for remotely detecting a pipe leak. mapping a background signal measured by a plurality of detection devices installed at a plurality of locations different from each other based on a piping facility and an installation location;
calculating a ratio of S/N of the leakage signal to the background signal measured by each of the plurality of detection devices; and
specifying an area monitored by a detection device having the largest value of the calculated S/N ratio as a leakage area;
including,
A method for remotely detecting pipe leaks, comprising remotely detecting pipe leaks in a non-contact and non-destructive manner via ultrasonic acoustics.

Images