JP4676867B2 - Anomaly detection device - Google Patents

Description

  The present invention relates to an abnormality detection device that detects abnormalities such as leakage of flammable gas and generation of combustion flame in facilities where flammable gas such as hydrogen gas is handled.

  In recent years, as environmental problems and energy problems become more serious, hydrogen energy has attracted attention as a next-generation clean energy that can replace petroleum.

Conventionally, in hydrogen handling facilities such as hydrogen production plants and hydrogen stations, hydrogen gas leaks and hydrogen flames are burned by sensors such as flame detectors, gas detectors, pressure gauges, and industrial TV (ITV). Detecting abnormalities such as the occurrence of
JP-A-9-196825 JP 2004-294423 A

  However, conventional sensors such as flame detectors used for abnormality detection have the following difficulties.

  First, it has been difficult to specify the location of a flame by monitoring with a flame detector. In addition, in order to specify the location of the flame, it was necessary to arrange a large number of fire detectors in the equipment to be monitored.

  Also, in the monitoring by the gas detector, in order to specify the gas leakage location, it is necessary to arrange many gas detectors in the equipment to be monitored. Also, hydrogen gas is lighter than air and diffuses faster. For this reason, when hydrogen gas leaks outdoors, it was difficult to detect with a gas detector.

  In the case of a pressure gauge, as long as a gas such as hydrogen continues to be supplied in a facility such as a hydrogen station, a large pressure drop does not occur even if a slight gas leak occurs. For this reason, it was difficult to detect a gas leak at an early stage using a pressure gauge.

  In the plant, monitoring is performed by ITV, but ITV has sensitivity only in the visible region. For this reason, it was impossible to detect the occurrence of a hydrogen flame having an emission peak outside the visible region by ITV.

  Further, in the above-described conventional abnormality detection, an abnormality is to be detected by each sensor. For this reason, the reliability of abnormality detection is low.

  An object of the present invention is to provide an abnormality detection device capable of early detection of gas leakage and generation of a combustion flame in which a combustible gas is burned with high reliability.

The object is to detect the sound in the monitored equipment and output the sound signal, and to process the sound signal output by the sound detection means to calculate the frequency spectrum of the sound signal, using the model, based on the frequency spectrum, the monitoring target abnormal stepwise determination to the equipment possess an acoustic processing means for detecting an abnormality in the monitored equipment, the leakage of the hydrogen gas in the monitored equipment An acoustic sensor unit that detects an abnormality, and an imaging unit that captures an image in the monitoring target facility is an imaging unit that is a near-infrared camera having spectral sensitivity in a part of a wavelength band in a wavelength range of 920 to 1100 nm. There is an image processing means for detecting an abnormality in the monitored equipment by processing an image picked up by the means Generating an image obtained by performing time difference processing for calculating a difference absolute value for each pixel between an image captured by the image capturing unit and an image captured by the image capturing unit with a predetermined time interval thereafter; Based on an image sensor unit that performs image processing for removing the influence of background light, an abnormality detection result by the acoustic sensor unit, and an abnormality detection result by the image sensor unit, an abnormality in the monitoring target facility is determined in a composite manner This is achieved by an anomaly detection device having a monitoring unit.

The abnormality detection device further includes a gas sensor unit that detects hydrogen gas in the monitored facility and detects leakage of hydrogen gas as an abnormality, or a fire detection unit that detects a fire in the monitored facility as an abnormality. In addition to the abnormality detection result by the acoustic sensor unit and the abnormality detection result by the image sensor unit, the monitoring unit has the monitoring target equipment based on the abnormality detection result by the gas sensor unit or the fire detection unit. You may make it judge the abnormality in in multiple ways.

  In the abnormality detection apparatus, the image sensor unit may detect an occurrence of a hydrogen flame in which hydrogen gas is burned in the monitoring target facility as an abnormality.

In the abnormality detection apparatus, the imaging unit includes an imaging device including an imaging device having spectral sensitivity in a part of a wavelength band longer than a wavelength of 0.7 μm, and the front surface of the imaging device or the imaging device. is arranged immediately before the low-pass filter transmits light of wavelengths longer than the wavelength 0.9μm and a wavelength selection filter for separating the light and the background light the flame emitted, or wavelength 0.9 You may make it have a band pass filter which permeate | transmits the light of a 1.2 micrometer band .

In the above abnormality detection device,
The monitoring target facility may be a hydrogen station.

  According to the present invention, an acoustic sensor unit, an image sensor unit, and a gas sensor unit are combined and combined to make use of the advantages of each sensor and make up for the disadvantages to detect abnormalities in the monitored facility. The occurrence of a combustion flame in which a combustible gas is burned can be detected early with high reliability.

[One Embodiment]
An abnormality detection apparatus according to an embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a schematic diagram illustrating the configuration of the abnormality detection device according to the present embodiment, FIG. 2 is a schematic diagram illustrating the configuration of an acoustic sensor unit in the abnormality detection device according to the present embodiment, and FIG. 3 is an image of the abnormality detection device according to the present embodiment. FIG. 4 is a schematic diagram showing the configuration of the sensor unit, FIG. 4 is a graph showing the frequency spectrum of the leakage sound of hydrogen gas, FIG. 5 is a diagram showing the imaging result of hydrogen flame by the image sensor unit in the abnormality detection device according to the present embodiment, and FIG. These are figures which show the detection result of the leak of hydrogen gas by the combination of the acoustic sensor part and gas sensor part in the abnormality detection apparatus by this embodiment.

  The abnormality detection device according to the present embodiment reports an abnormality by monitoring leakage of hydrogen gas, generation of a hydrogen flame in which hydrogen burns, etc. in a monitoring target facility that handles hydrogen, such as a hydrogen production plant or a hydrogen station. .

  As shown in FIG. 1, the abnormality detection apparatus according to the present embodiment processes an acoustic signal detected in a monitoring target facility to detect leakage of hydrogen gas in the monitoring target facility, and the monitoring target facility. An image sensor unit 12 that processes an image to be captured and detects the occurrence of hydrogen flame in the monitoring target facility; a gas sensor unit 14 that detects hydrogen gas in the monitoring target facility and detects leakage of hydrogen gas in the monitoring target facility; And a monitoring unit 16 that compositely determines an abnormality in the monitoring target facility based on the detection results of the abnormality by the sensor units of the acoustic sensor unit 10, the image sensor unit 12, and the gas sensor unit 14.

(A) Acoustic sensor unit 10
As shown in FIG. 2, the acoustic sensor unit 10 includes a microphone 18 as sound collection means for monitoring abnormal sound due to leakage of hydrogen gas, an amplification unit 20 that amplifies an acoustic signal output from the microphone 18, and An A / D converter 22 that converts the acoustic signal amplified by the amplifier 20 from an analog signal to a digital signal, and a Fourier transform unit that calculates the frequency spectrum of the acoustic signal converted into a digital signal by the A / D converter 22 24 and a determination unit 26 that determines abnormality based on a determination model based on a neural network based on the frequency spectrum calculated by the Fourier transform unit 24.

  The microphone 18 is a non-directional one having a sensitivity of, for example, 20 Hz to 20 kHz. The microphone 18 has an intrinsically safe explosion-proof structure. The microphone 18 is disposed in the monitoring target facility. In addition, you may arrange | position the several microphone 18 suitably in the monitoring object installation.

  The amplifying unit 20 amplifies an acoustic signal of 20 Hz to 20 kHz, for example, with which the microphone 18 is sensitive.

  The A / D conversion unit 22 converts the acoustic signal amplified by the amplification unit 20 from an analog signal to a digital signal. The resolution of A / D conversion by the A / D conversion unit 22 is, for example, 8 to 16 bits, and the sampling frequency is selected from, for example, 44.1 kHz, 48.0 kHz, and 51.2 kHz.

  The Fourier transform unit 24 performs a fast Fourier transform on the acoustic signal converted into a digital signal by the A / D conversion unit 22 and calculates a frequency spectrum of the acoustic signal. The number of data points of the fast Fourier transform by the Fourier transform unit 24 is selected from, for example, 200, 400, 1024, 2048, 4096, 8192, 16384, 32768, and 65536. The window function is selected from, for example, hamming, hanning, and no window function.

  Based on the frequency spectrum calculated by the Fourier transform unit 24, the determination unit 26 identifies a hydrogen gas leakage sound and a normal sound, which are abnormal sounds, using a neural network model. In the learning of the neural network, the normal sound data is output as 0 and the abnormal sound data is output as 1. As the neural network, a hierarchical neural network is used, and learning is performed by back propagation. The response function in the neural network is selected from, for example, a sigmoid function and a radial base function.

  The determination result by the determination unit using the neural network model is output as a determination value of 0 to 1. When the determination value is 0 or more and 0.2 or less, it is determined as “normal” without hydrogen gas leakage. When the determination value is larger than 0.2 and smaller than 0.8, it is determined as “caution” that there is a risk of leakage of hydrogen gas. When the determination value is 0.8 or more, it is determined as “abnormal” in which hydrogen gas leaks.

(B) Image sensor unit 12
The image sensor unit 12 captures an image of the monitored facility, and detects the occurrence of hydrogen flame in which hydrogen gas is burned based on changes in images with different imaging times.

  As shown in FIG. 3, the image sensor unit 12 processes the image captured by the image capturing device 28 and the image capturing device 28 that captures the inside of the monitoring target facility, and detects the occurrence of hydrogen flame based on the temporal change of the image. And an image processing unit 30.

  The imaging device 28 uses an ultraviolet camera that can detect the tail of the emission peak in the ultraviolet region where hydrogen flames are emitted, or a near-infrared camera that can detect water vibration peaks distributed in the wavelength range of 920 to 1100 nm. It is done.

  Specifically, as the imaging device 28, for example, an ultraviolet camera or a near infrared camera configured as follows can be used.

  For example, a broadband monochrome video camera (wavelength sensitivity band: 0.2 to 1.1 μm) represented by MC-781P (manufactured by Texas Instruments Incorporated) is equipped with an ultraviolet lens and a filter that transmits ultraviolet light of less than 300 nm. An additional ultraviolet camera can be used. Alternatively, a near-infrared camera configured by adding a general-purpose C-mount lens and a filter that transmits infrared light having a wavelength of 920 nm or more to a broadband monochrome video camera can be used. The range of the wavelength sensitivity band of the broadband monochrome video camera is not absolute, and tail sensitivity exists outside this range.

  Further, a near-infrared camera configured by adding a general-purpose C-mount lens and a filter that transmits infrared light having a wavelength of 920 nm or more to a commercially available monochrome CCD camera (wavelength sensitivity band: 0.4 to 1.0 μm) can be used. . Note that the range of the wavelength sensitivity band of the monochrome CCD camera is not absolute, and tail sensitivity exists outside this range.

  Further, a near-infrared camera configured by removing an infrared cut filter from a commercially available color CCD camera and adding a general-purpose C-mount lens and a filter that transmits infrared light having a wavelength of 920 nm or more to the color CCD camera can be used. .

  Further, an ultraviolet camera configured by removing an infrared cut filter from a commercially available CCD type web camera and adding an ultraviolet lens and a filter that transmits ultraviolet light having a wavelength of less than 300 nm to the CCD type web camera can be used. . Alternatively, a near-infrared camera configured by adding a general-purpose C-mount lens and a filter that transmits infrared light having a wavelength of 920 nm or more to a CCD type Web camera from which an infrared cut filter is removed can be used.

  As described above, in the abnormality detection device according to the present embodiment, as the imaging device 28, an ultraviolet camera capable of detecting the tail of the emission peak in the ultraviolet region where hydrogen flame is emitted, or water vibration distributed in the wavelength range of 920 to 1100 nm. Since a near-infrared camera capable of detecting a peak is used, a hydrogen flame that cannot be seen with the naked eye can be reliably detected.

  In addition, as a filter added to various cameras, it is better to use a bandpass filter that selectively transmits light in a specific wavelength range than a filter that defines an upper limit or a lower limit such as a low-pass filter or a high-pass filter. The S / N ratio of the hydrogen flame light can be increased with respect to background light such as natural light and white light which are noise components.

  Further, the ultraviolet camera and the near infrared camera used as the imaging device 28 are not limited to the above specific examples.

  When configuring an ultraviolet camera, by disposing a wavelength selection filter on the front surface of an imaging apparatus including an imaging device having spectral sensitivity in a part of a wavelength band shorter than at least a wavelength of 0.45 μm or immediately before the imaging device. Can be configured. The image pickup device included in the image pickup apparatus is a CCD image pickup device, a CMOS image pickup device, a vidicon tube image pickup device, or the like. The wavelength selective filter is for separating the light emitted from the hydrogen flame from the background light. In this case, as the wavelength selection filter, a high-pass filter that transmits light having a wavelength shorter than a wavelength of 0.4 μm or a band-pass filter that transmits light having a wavelength of 0.2 to 0.4 μm is used.

  Also, when configuring a near-infrared camera, a wavelength selection filter is arranged in front of the imaging device including an imaging device having spectral sensitivity in a part of a wavelength band longer than at least 0.7 μm, or immediately before the imaging device. This can be configured. The image pickup device included in the image pickup apparatus is a CCD image pickup device, a CMOS image pickup device, a vidicon tube image pickup device, or the like. The wavelength selective filter is for separating the light emitted from the hydrogen flame from the background light. In this case, as the wavelength selection filter, a low-pass filter that transmits light having a wavelength longer than the wavelength of 0.9 μm or a band-pass filter that transmits light having a wavelength of 0.9 to 1.2 μm is used.

  The image processing unit 30 detects the hydrogen flame with high sensitivity by performing image processing on the image captured by the imaging device 28. Specifically, the image processing unit 30 calculates, for each pixel, an absolute difference value between an image captured by the imaging device 28 at a certain moment and an image captured by the imaging device 28 at a predetermined time interval thereafter. An image subjected to time difference processing is generated. By such image processing, the influence of background light such as sunlight can be removed and the hydrogen flame can be reliably detected.

(C) Gas sensor unit 14
The gas sensor unit 14 is configured by a general-purpose gas detector or the like. The hydrogen gas in the monitoring target equipment is detected by a general-purpose gas detector or the like, and leakage of the hydrogen gas is detected.

(D) Monitoring unit 16
The monitoring unit 16 determines information sent from the sensors of the acoustic sensor unit 10, the image sensor unit 12, and the gas sensor unit 14 in a composite manner. The monitoring unit 16 is abnormal when any one of the acoustic sensor unit 10, the image sensor unit 12, and the gas sensor unit 14 reports an abnormality and when a plurality of sensor units report an abnormality. The degree is classified to determine whether there is an abnormality in the monitored facility, that is, whether hydrogen gas has leaked or hydrogen flame has occurred. Based on the determination result, the monitoring unit 16 issues an alarm notifying the leakage of hydrogen gas, the occurrence of a fire, the occurrence of abnormal noise, and the like.

  As described above, the abnormality detection device according to the present embodiment combines and combines the acoustic sensor unit 10, the image sensor unit 12, and the gas sensor unit 14 to make use of the advantages of each sensor and compensate for the disadvantages. Therefore, abnormalities such as leakage of hydrogen gas and occurrence of hydrogen flame can be detected early with high reliability.

  Next, actual examples of abnormality detection such as detection of leakage of hydrogen gas and detection of hydrogen flame by the abnormality detection apparatus according to the present embodiment will be described.

  First, detection of leakage of hydrogen gas by the acoustic sensor unit 10 will be described.

  FIG. 4 is a graph showing the frequency spectrum of the leakage sound of hydrogen gas. FIG. 4 shows the frequency spectrum of the passing sound of the vehicle (bus) as an example of the sound of the urban area together with the frequency spectrum of the hydrogen gas leakage sound. The horizontal axis of the graph shown in FIG. 4 is frequency (Hz), and the vertical axis is sound pressure (dB). In addition, the leakage sound of hydrogen gas is obtained by measuring the leakage sound of 0.9 MPa hydrogen gas leaking from a hole of Φ1 provided in a 3/8 inch pipe 3 meters away.

  As shown in FIG. 4, in the frequency spectrum of the passing sound of the vehicle, the sound pressure decreases as the frequency increases. On the other hand, the frequency spectrum of the leakage sound of hydrogen gas is almost flat with respect to the frequency. As described below, by learning such a pattern with a neural network, it is possible to distinguish and detect normal sounds without hydrogen gas leakage and abnormal sounds with hydrogen gas leakage. It was.

  For measurement of normal and abnormal sound signals, a 1/2 inch electrolet condenser microphone with an intrinsically safe explosion-proof structure was used together with a safety holder. The frequency range of the measurement sound was 20 Hz to 20 kHz which is an audible range. The measured acoustic signal data was converted from analog data to digital data by A / D conversion with a resolution of 16 bits and a sampling frequency of 51.2 kHz. The acoustic signal converted into digital data was subjected to fast Fourier transform with a frequency of 20 Hz to 20 kHz, using Hanning as a window function, and 8192 data points. Thus, the frequency spectrum was calculated for each of normal sound and abnormal sound.

  In learning the neural network model, the following data was used as normal sound and abnormal sound data.

  First, as normal sound data, 40 vehicle passing sounds were used.

  As the abnormal sound data, gas leakage sound was used. Assuming piping used in the hydrogen station, gas leakage sound was generated by drilling a hole of Φ1 in a 3/8 inch piping and leaking gas from this hole. Abnormal sound data is a frequency spectrum collected as a normal sound from a frequency spectrum of a hydrogen gas leak sound of 0.9 MPa, a frequency spectrum of a helium gas leak sound of 0.9 MPa, and a frequency spectrum of a helium gas leak sound of 2.5 MPa. Was created by artificially synthesizing sound. 120 data of abnormal sound was created.

  A radial base function was used as the response function of the neural network model. The neural network model learning was performed with 7 intermediate layers, 73 learning cycles, 0 normal and 1 abnormal.

  As a result of evaluation using the neural network model learned in this way, the average sound was 0.003 and the abnormal sound was 0.996 on average. Thereby, it was possible to identify and detect a normal sound without gas leakage and an abnormal sound in which gas leakage occurred.

  Next, detection of occurrence of hydrogen flame by the image sensor unit 12 will be described.

  FIG. 5 is a diagram illustrating the imaging result of the hydrogen flame by the image sensor unit 12.

  For imaging of hydrogen flame, a broadband monochromatic CCD camera MC-781P (manufactured by Texas Instruments Incorporated) is attached with a TU2440A1 (manufactured by Texas Instruments Incorporated), a lens for broadband UV, and a wavelength of 920 nm or more. The near-infrared camera which permeate | transmits infrared rays with a wavelength of 920 nm-1100 nm comprised by attaching the filter which permeate | transmits the infrared rays of was used.

  The hydrogen flame was imaged in an outdoor environment with direct sunlight, choosing a sunny day during the day when the ultraviolet intensity was strong. Further, the hydrogen flame to be imaged was ignited in a state where pure hydrogen was leaked from a nozzle having a diameter of 4 mm at a flow rate of about 5 L / min.

  FIG. 5A is an image of a hydrogen flame at a certain moment captured by a near-infrared camera. FIG. 5B is an image of a hydrogen flame imaged one minute after the moment when FIG. 5A is imaged. FIG. 5C is an image obtained by performing time difference processing between FIG. 5A and FIG.

  FIG. 5C shows the absolute difference between the image shown in FIG. 5A captured at a certain moment by the image processing software and the image shown in FIG. 5C captured one minute later. It is the image obtained by calculating for every pixel.

  By such image processing, the pixels in the portion where there is no change in the image are zeroized. For example, a pixel in a portion where there is no change in the image, such as a portion surrounded by a dotted line in FIGS. 5A and 5B, has a zero value as in a portion surrounded by a dotted line in FIG. It becomes. On the other hand, the amount of change between the two images is displayed on the pixel in the portion where there is a change in the image. For example, a pixel having a change on the image, such as a part surrounded by a solid line in FIGS. 5A and 5B, may be 2 as a part surrounded by a solid line in FIG. The amount of change in one image is displayed. Thereby, the influence of background lights, such as sunlight, is removed, and a hydrogen flame can be detected reliably.

  In the images shown in FIGS. 5A to 5C, the deviation of sunlight associated with the day-to-day exercise is detected together with the hydrogen flame. About such a sunlight deviation, it is possible to make it less than a detection limit by narrowing the time interval which images two images which calculate a difference absolute value. That is, only the hydrogen flame can be detected by appropriately setting the time interval for capturing two images for calculating the absolute difference value according to the imaging environment and the like.

  Next, detection of leakage of hydrogen gas by a combination of the acoustic sensor unit 10, the image sensor unit 12, and the gas sensor unit 14 will be described as an example of combining the sensor units.

  FIG. 6 is a diagram illustrating a detection result of leakage of hydrogen gas by a combination of the acoustic sensor unit 10, the image sensor unit 12, and the gas sensor unit 14. FIG. 6 is a diagram illustrating hydrogen gas generated by four representative gas sensors (CH3, CH8, CH9, CH10), one image sensor, and one acoustic sensor among the gas sensors (12) arranged in the monitoring target facility. The result of detecting the leakage of gas is shown. In the table shown in FIG. 6, the presence / absence of leakage detection by the gas sensor, the image sensor, and the acoustic sensor is indicated by ◯ etc. with respect to the time elapsed from the start of leakage of hydrogen gas. In the table, in the “gas sensor”, “image sensor”, and “acoustic sensor” columns, “x” indicates that the sensor has not detected a leak, and “○” indicates that the sensor has detected a leak and has issued a report. Show. In the “Anything is reported” column, “X” indicates that none of the gas sensor, image sensor, or acoustic sensor has detected a leak, and any of the gas sensor or acoustic sensor has detected a leak and has issued a report. The case is indicated by “Δ”. In the “acoustic and gas” column, “x” indicates that neither the gas sensor nor the acoustic sensor has detected a leak, or if only one of them has detected a leak, and both the gas sensor and the acoustic sensor have leaked. “○” indicates a case where a warning is detected and issued. In the “acoustic and image” column, “x” indicates that both the image sensor and the acoustic sensor have not detected leakage, and only one of them has detected leakage, and both the image sensor and the acoustic sensor have “◎” indicates that a leak was detected and reported.

  As shown in FIG. 6, the acoustic sensor detects hydrogen gas leakage faster than the gas sensor, so the acoustic sensor first detects hydrogen gas leakage immediately after the start of hydrogen gas leakage. Yes. Specifically, the acoustic sensor detects hydrogen gas leakage 10 seconds after the start of leakage.

  Following detection of hydrogen gas leakage by the acoustic sensor, the gas sensor detects hydrogen gas leakage. Specifically, the gas sensors CH9 and CH10 located in the leeward direction from the hydrogen gas leak location detect the hydrogen gas leak 50 to 60 seconds after the start of the leak. The gas sensors CH3 and CH8 do not detect leakage of hydrogen gas due to the wind direction or the like.

  The image sensor detects the occurrence of hydrogen flame by the ignition of leaked hydrogen gas. Specifically, the image sensor detects the occurrence of hydrogen flame 110 seconds after the start of leakage.

  Since the leakage sound continues to occur after the ignition of hydrogen gas, the acoustic sensor detects the leakage sound continuously after the ignition of hydrogen gas immediately after the leakage. On the other hand, the gas sensor no longer detects hydrogen gas due to the ignition of hydrogen gas.

  In such a combination of sensors, when one of the sensors reports an abnormality as indicated by “△”, “○”, “◎” in the table, or when multiple sensors issue an abnormality. By classifying the degree of abnormality with, it was possible to detect the abnormality early with high reliability.

  As described above, according to the present embodiment, the acoustic sensor unit 10, the image sensor unit 12, and the gas sensor unit 14 are combined and combined to make use of the advantages of each sensor and to compensate for the shortcomings. Since detection is performed, abnormalities such as hydrogen gas leakage and hydrogen flame can be detected early with high reliability.

[Modified Embodiment]
The present invention is not limited to the above embodiment, and various modifications can be made.

  For example, in the above embodiment, the present invention is applied to monitoring an abnormality in a monitoring target facility that handles hydrogen, such as a hydrogen production plant or a hydrogen station, but the scope of application of the present invention is not limited to this. The present invention can be widely applied to, for example, monitoring abnormalities such as gas leakage in facilities that handle natural gas, petroleum gas, and the like, and generation of a combustion flame in which a combustible gas is burned.

  In the above-described embodiment, the case where an ultraviolet camera or a near-infrared camera is used as the imaging device 28 in the image sensor unit 12 has been described as an example, but depending on the wavelength of light emitted from the combustion flame of the detection target gas, The imaging device 28 having sensitivity in a predetermined wavelength band can be used as appropriate. In this case, the light emitted from the combustion flame is separated from the background light in front of the imaging device including the imaging device having spectral sensitivity in the wavelength band including the wavelength of the light emitted from the combustion flame of the detection target gas or immediately before the imaging device. The imaging device 28 is configured by arranging a wavelength selection filter for this purpose.

  Moreover, although the said embodiment demonstrated the case where the acoustic sensor part 10, the image sensor part 12, and the gas sensor part 14 were compounded, the fire detection comprised by a general purpose fire detector etc. with such a sensor part. The means may be combined.

It is the schematic which shows the structure of the abnormality detection apparatus by one Embodiment of this invention. It is the schematic which shows the structure of the acoustic sensor part in the abnormality detection apparatus by one Embodiment of this invention. It is the schematic which shows the structure of the image sensor part in the abnormality detection apparatus by one Embodiment of this invention. It is a graph which shows the frequency spectrum of the leakage sound of hydrogen gas. It is a figure which shows the imaging result of the hydrogen flame by the image sensor part in the abnormality detection apparatus by one Embodiment of this invention. It is a figure which shows the detection result of the leakage of hydrogen gas by the combination of the acoustic sensor part and gas sensor part in the abnormality detection apparatus by one Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Acoustic sensor part 12 ... Image sensor part 14 ... Gas sensor part 16 ... Monitoring part 18 ... Microphone 20 ... Amplification part 22 ... A / D conversion part 24 ... Fourier transform part 26 ... Determination part 28 ... Imaging device 30 ... Image processing part

Claims (5)

A sound detection means for detecting sound in the monitored facility and outputting a sound signal; and processing the sound signal output by the sound detection means to calculate a frequency spectrum of the sound signal, using a model of a neural network, based on the frequency spectrum, the abnormality stepwise determined by the monitored equipment possess an acoustic processing means for detecting an abnormality in the monitored equipment, for detecting the leakage of the hydrogen gas in the monitored equipment as abnormal An acoustic sensor unit;
The imaging unit that captures an image in the monitoring target facility is an imaging unit that is a near-infrared camera having spectral sensitivity in a part of a wavelength band in a wavelength range of 920 to 1100 nm, and processes an image captured by the imaging unit. The image processing means for detecting an abnormality in the monitoring target equipment calculates the absolute value of the difference between the image picked up by the image pickup means at a certain moment and the image picked up by the image pickup means after a predetermined time interval. An image sensor unit that performs an image process for removing the influence of background light by generating an image that has been subjected to a time difference process that is calculated every time;
An abnormality detection apparatus comprising: a monitoring unit that collectively determines an abnormality in the monitoring target facility based on an abnormality detection result by the acoustic sensor unit and an abnormality detection result by the image sensor unit. The abnormality detection device according to claim 1,
A gas sensor unit that detects hydrogen gas in the monitored facility and detects a hydrogen gas leak as an abnormality, or a fire detection unit that detects a fire in the monitored facility as an abnormality,
In addition to the abnormality detection result by the acoustic sensor unit and the abnormality detection result by the image sensor unit, the monitoring unit detects an abnormality in the monitoring target facility based on the abnormality detection result by the gas sensor unit or the fire detection unit. An anomaly detection device characterized in that it is determined in combination. In the abnormality detection device according to claim 1 or 2 ,
The image sensor unit detects an occurrence of a hydrogen flame in which hydrogen gas is burned in the monitoring target facility as an abnormality. In the abnormality detection device according to claim 3 ,
The imaging means includes an imaging device including an imaging device having spectral sensitivity in a part of a wavelength band longer than a wavelength of 0.7 μm, a front surface of the imaging device or immediately before the imaging device, and the hydrogen flame. As a wavelength selection filter for separating the light emitted from the light and the background light, a low-pass filter that transmits light having a wavelength longer than the wavelength of 0.9 μm, or a bandpass that transmits light in the wavelength range of 0.9 to 1.2 μm An abnormality detection device comprising a filter. In the abnormality detection apparatus according to any one of claims 1 to 4 ,
The monitoring target facility is a hydrogen station.