Fire detection system

Abstract

To observe light emitted from a monitoring area to allow reliable and highly accurate detection of a fire.SOLUTION: A fire detection system comprises: a fire detection unit 18 that detects a fire in a monitoring area on the basis of a first fire determination result acquired by executing predetermined fire determination processing on observation data collected by observing an observation target in the monitoring area, and a second fire determination result calculated by executing, on the observation data, fire determination processing based on predetermined machine learning processing; an observation unit 12 that observes emission of light from the monitoring area; a first processing unit 14 that executes the predetermined fire determination processing on observation data collected through observation performed by the observation unit 12, to acquire the first fire determination result; and a second processing unit 16 that executes, on the observation data, the fire determination processing based on predetermined machine learning processing, to calculate the second fire determination result. The fire detection system obtains the first and second fire determination results on the basis of pieces of observation data in first, second, and third wavelength bands.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a fire detection system and a fire detection method for detecting a fire by observing an observation target in a monitored area. [Background technology]

[0002] An example of a fire detection system that detects fires by observing objects in a monitored area is a flame detection device that observes light emitted from inside a road tunnel and detects a fire if it determines that light characteristic of a flame has been generated.

[0003] Such flame detection devices observe the emission of light from a monitored area, and detect fires by observing the light (infrared energy) emitted from a flame when a fire breaks out (Patent Documents 1 to 3). For example, a two-wavelength flame detection device observes light emitted from a flame at around 4.5 μm, which is the CO2 resonance radiation wavelength band, and light outside the CO2 resonance radiation band, for example, at around 5.0 μm, and determines that a fire has occurred when the relative ratio of the observation data of the light in these two wavelength bands is a value indicating the occurrence of a flame and the observed waveform of the light in the CO2 resonance radiation band has a flicker frequency component specific to a flame, and for example, transmits a fire signal to a disaster prevention receiving panel to perform a fire alarm operation. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2002-246962 A [Patent Document 2] JP 2016-128796 A [Patent Document 3] JP 2018-169893 A [Patent Document 4] JP 2018-088105 A [Patent Document 5] JP 2019-191902 A Summary of the Invention [Problem to be solved by the invention]

[0005] Even under normal circumstances, flame detection devices installed in tunnels are exposed to light of various wavelengths and intensities from the headlights of vehicles passing on the road, the flashing or blinking warning lights of emergency vehicles, lighting fixtures inside the tunnel, sunlight if installed at the entrance and exit of the tunnel, and workers (human bodies) involved in maintenance and inspection. However, they are equipped with an identification function to distinguish between this external light and the light emitted from flames associated with a fire, thereby improving fire (here, fire flame) identification performance.

[0006] In addition, the fire detection system may, for example, self-diagnose whether or not it is operating normally, including this identification function, and if it is diagnosed as not operating normally, a signal indicating a malfunction or functional failure may be output and reported from the disaster prevention receiving panel.

[0007] However, even when diagnosed as normal, it was difficult to completely eliminate the possibility that disturbance light similar to the light emitted from a fire flame would be judged to be a fire and a non-fire alarm would be output from the disaster prevention receiving panel.

[0008] In such cases, until it is confirmed that the incident is not a fire, a no-entry warning must be issued using warning signs or other means to prohibit vehicles from passing through the tunnel, and a management officer must go to the site (the location of the flame detection device that has determined that there is a fire) to check the situation. This takes time and effort to reopen the tunnel, and can have a significant impact, such as causing traffic congestion.

[0009] In addition to such "non-fire alarms," ​​there are also "false alarms" that occur when an internal factor in the flame detection device causes the device to malfunction, such as misjudging there to be a fire, resulting in a fire alarm being issued even though there is no fire. False alarms have the same impact as non-fire alarms.

[0010] The cause of false alarms is, for example, spontaneous electrical noise caused by component failure or deterioration of the electrical circuit that constitutes the flame detection device. In addition, false fire signals have also been reported due to wiring defects and insulation deterioration associated with the flame detection device. During self-diagnosis of the flame detection device, such failures, deterioration, and defects are also diagnosed. However, if a circuit failure or deterioration occurs during the interval between diagnoses performed at a predetermined cycle, that is, if the previous diagnosis was normal but a circuit failure or deterioration occurs during the period until the next diagnosis, if the spontaneous noise caused by this spreads to the light receiving circuit, it may be processed as an optical input signal and be mistaken for light emitted from a flame.

[0011] In the following explanation, for convenience, no distinction will be made between non-fire alarms and false alarms, and both will be referred to as "non-fire alarms."

[0012] The present invention aims to provide a fire detection system and a fire detection method that have improved fire identification performance compared to conventional systems, and to reduce false fire alarms and false alarms. [Means for solving the problem]

[0013] (Fire Detection System 1) The present invention provides a fire detection system, comprising: A feature of this system is that it detects fires based on a first fire judgment result obtained by executing a predetermined fire judgment process on observation data of an object observed in a monitored area, and a second fire judgment result calculated by executing a fire judgment process based on a predetermined machine learning process on the observation data.

[0014] Here, "observation data" refers to time-series data obtained by sampling observation signals (analog signals) obtained by observing the object in the monitoring area at a predetermined frequency and A / D converting the samples. Also, "fire determination process" refers to an algorithm (essentially a computer program) for determining whether or not a predetermined condition for determining a fire is met, and "first fire determination result" refers to the result of the determination.

[0015] Further, the "fire judgment process based on machine learning processing" is an algorithm (substantially a computer program) for predicting whether the degree of fire likelihood is high or low, and the "second fire judgment result" is a result of quantitatively expressing and evaluating the prediction result and judging a fire based on this. Here, "machine learning" means inputting unknown data into a mathematical model constructed from a data set including both input and output, and predicting the output. When applied to the "machine learning processing" of the present invention, it means inputting new observation data into a mathematical model (learning model section) constructed (learned) from learning data including both previously collected observation data and the second fire judgment result based on the previously collected observation data, deriving a prediction result of the degree of fire likelihood corresponding to the new observation data, and obtaining a second fire judgment result based on this. Machine learning in this embodiment will be described in detail later.

[0016] (Fire Detection System 2) The present invention also provides a fire detection system, comprising: A feature of this system is that it detects fires based on a first fire judgment result obtained by applying a predetermined fire judgment process to observation data collected by observing the emission of light from a monitored area, and a second fire judgment result calculated by applying a fire judgment process based on a predetermined machine learning process to the observation data.

[0017] (Fire Detection System 3) The present invention also provides a fire detection system, comprising: an observation unit for observing light emission from a monitoring area; a first processing unit that executes a predetermined fire judgment process on the observation data observed and collected by the observation unit to obtain a first fire judgment result; A second processing unit that executes a fire judgment process based on a predetermined machine learning process on the observation data to calculate a second fire judgment result; a fire detection unit that detects a fire in a monitored area based on the first fire judgment result and the second fire judgment result; The present invention is characterized by comprising:

[0018] (Second fire judgment result) The second fire judgment result is a value ranging from 0 when there is no fire to 1 when there is a fire. In other words, the second fire judgment result based on the observation data when there is a fire is 1 (or 100%), and the second fire judgment result based on the observation data when there is no fire is 0 (or 0%), and the second fire judgment result ranges from 0 to 1 (or 0% to 100%) depending on the observation data.

[0019] (Machine Learning) The second processing unit is Equipped with a machine learning unit that inputs observation data and outputs the second fire judgment result, The Machine Learning Department The correct answer (teacher) is at least when the second fire judgment result is 0. Observation data from normal monitoring is input as learning data in advance, and machine learning is performed. After machine learning, newly observed observation data is input and the second fire judgment result is output.

[0020] (Fire judgment based on the first and second observation data 1) The observation section is First observation data of light in a first wavelength band emitted from a flame associated with a fire; Second observation data of light outside the first wavelength band; and Collect and The first processing unit acquires a first fire determination result based on the first observation data and the second observation data, The second processing unit calculates a second fire determination result based on the first observation data and the second observation data.

[0021] (Fire judgment based on the first and second observation data 2) The observation section is First observation data of light in a first wavelength band emitted from a flame associated with a fire; Second observation data of light outside the first wavelength band; and Collect and the first processing unit acquires a first fire judgment result based on a first integral value of the first observation data for each predetermined period and a ratio between the first integral value and a second integral value of the second observation data for each predetermined period; The second processing unit calculates a second fire determination result based on the first integral value and the ratio generated by the first processing unit.

[0022] (Fire detection) The fire detection unit determines that a fire is detected when the first fire judgment result satisfies a predetermined condition and the second fire judgment result is equal to or greater than a predetermined value or exceeds a predetermined value.

[0023] (Fire warning detection) The fire detection section is When the first fire judgment result satisfies a predetermined condition and the second fire judgment result is less than a predetermined value or equal to or less than a predetermined value, or When the first fire judgment result does not satisfy the specified condition and the second fire judgment result is equal to or exceeds the specified value, It is used to detect signs of fire.

[0024] (Fire detection system placement) In addition, the fire detection system of the present invention includes: A flame detection device provided with an observation unit, a first processing unit, and a fire detection unit; a disaster prevention receiving panel to which a flame detection device is connected and to which a second processing unit is provided; The present invention is characterized by comprising:

[0025] (Fire detection method 1) The present invention provides a fire detection method, comprising: A feature of this system is that it detects fires based on a first fire judgment result obtained by executing a predetermined fire judgment process on observation data of an object observed in a monitored area, and a second fire judgment result calculated by executing a fire judgment process based on a predetermined machine learning process on the observation data.

[0026] (Fire detection method 2) The present invention provides a fire detection method, comprising: A feature of this system is that it detects fires based on a first fire judgment result obtained by applying a predetermined fire judgment process to observation data collected by observing the emission of light from a monitored area, and a second fire judgment result calculated by applying a fire judgment process based on a predetermined machine learning process to the observation data.

[0027] (Fire detection method 3) The present invention also provides a fire detection method, comprising: The observation unit observes the light emitted from the monitored area; A first processing unit executes a predetermined fire judgment process on the observation data observed and collected by the observation unit to obtain a first fire judgment result; A second processing unit executes a fire judgment process based on a predetermined machine learning process on the observation data to calculate a second fire judgment result; A fire in the monitored area is detected by the fire detection unit based on the first fire judgment result and the second fire judgment result. The other features are the same as those of the fire detection device. Effect of the Invention

[0028] (Effectiveness of fire detection systems) According to the fire detection system of the present invention, fires are detected based on a first fire judgment result obtained by executing a predetermined fire judgment process on observation data collected by observing the emission of light from a monitored area, and a second fire judgment result calculated by executing a fire judgment process based on a predetermined machine learning process on the observation data.This makes it possible to improve fire identification performance compared to conventional systems, and to prevent the problem of false fire alarms, where a fire is judged to be a fire when there is no fire.

[0029] (Effect of the second fire judgment result) In addition, the second fire judgment result is a value ranging from 0 if there is no fire to 1 if there is a fire, making it possible to grasp the degree of fire-likelihood (the probability of a fire occurring) as a constant and ensure reliable fire detection.

[0030] (Effects of machine learning) The second processing unit includes a machine learning unit that inputs observation data and outputs a second fire judgment result, and the machine learning unit inputs observation data during normal monitoring as learning data in advance, with the correct answer (teacher) being at least the case where the second fire judgment result is 0, and performs machine learning. By inputting newly observed observation data after the machine learning and outputting the second fire judgment result, it is not necessary to artificially come up with and determine conditions, etc. for calculating the second fire judgment result, and the second fire judgment result can be output with high accuracy even for unknown observation data by machine learning. Since this learning data is observation data in the environment where the flame detection device is actually installed, if there is a phenomenon specific to the installation location that is different from a general phenomenon, the second fire judgment result can be calculated with even higher accuracy specialized for the installation location.

[0031] In addition, machine learning is performed by inputting, as learning data, "observation data during fires" in which the correct answer is when the second fire judgment result is 1, and "observation data during normal monitoring" in which the correct answer is when the second fire judgment result is 0. However, since fires occur very rarely and are exceptional, it is difficult to collect "observation data during fires" as learning data. On the other hand, "observation data during normal monitoring" is observation data observed during normal monitoring when no fires have occurred, and can be collected easily and simply as a large amount of learning data through the operation of the fire detection system. Therefore, by using machine learning on "observation data during normal monitoring", which can be collected in large quantities, as learning data, it is possible to improve the accuracy of the learning machine unit that inputs observation data and outputs the second fire judgment result.

[0032] (Effect of fire judgment 1 based on the first and second observation data) In addition, the fire detection system collects first observation data of light in a first wavelength band emitted from a flame in the event of a fire, and second observation data of light outside the first wavelength band, and obtains a first fire judgment result and calculates a second fire judgment result based on the first observation data and the second observation data.Therefore, when the system is installed in a tunnel to detect a fire, it will not determine that a fire has been detected even if it observes light emitted from various sources other than flames, such as the headlights of vehicles passing on the road, the flashing or blinking warning lights of emergency vehicles, lighting fixtures inside the tunnel, sunlight when installed at the entrance and exit of a tunnel, and workers (human bodies) engaged in maintenance and inspection.This improves fire identification performance compared to conventional systems, and makes it possible to prevent situations from occurring such as a fire alarm being issued when there is no fire, and passage through the tunnel being prohibited until it is confirmed that there is no fire alarm.

[0033] (Effect of fire judgment 2 based on the first and second observation data) In addition, the fire detection system collects first observation data of light in a first wavelength band emitted from a flame in the event of a fire, and second observation data of light outside the first wavelength band, obtains a first fire judgment result based on a first integral value of the first observation data for each specified period and a ratio between the first integral value and a second integral value of the second observation data for each specified period, and calculates a second fire judgment result based on the first integral value and ratio generated by the first processing unit.In this way, in addition to the above-mentioned ``effect of fire judgment 1 based on first observation data and second observation data,'' the system reduces the number of inputs to the second processing unit that performs machine learning (reduces the number of vector dimensions), simplifies the configuration of the machine learning unit, and makes it possible to reduce the burden on calculation processing.

[0034] (Fire detection effect) In addition, the fire detection unit detects a fire when the first fire judgment result satisfies a specified condition and the second fire judgment result is greater than or exceeds a specified value, thereby enabling more reliable and accurate fire detection.

[0035] (Effect of fire warning detection) If the first fire judgment result satisfies the specified conditions and the second fire judgment result is less than or equal to a specified value, or if the first fire judgment result does not satisfy the specified conditions and the second fire judgment result is greater than or equal to a specified value, the fire detection unit cannot determine whether there is a fire or not.In this case, the fire detection unit, for example, detects a fire precursor and issues a fire warning, for example by outputting a fire warning alarm from a disaster prevention receiving panel, encouraging relevant parties to check the site, thereby enabling action to be taken.

[0036] (Effect of fire detection system placement) Furthermore, the fire detection system of the present invention includes a flame detection device provided with an observation section, a first processing section, and a fire detection section, and a disaster prevention receiving panel to which the flame detection device is connected and provided with a second processing section, thereby simplifying the configuration of the flame detection device. That is, since the learning control section and learning data storage section of the second processing section are not essential for normal monitoring after learning (the learning model section is essential), the present invention can be implemented without significantly changing the configuration of a conventional flame detection device by arranging them in the disaster prevention receiving panel, which has ample installation space.

[0037] (Effectiveness of fire detection methods) Furthermore, the fire detection method of the present invention can provide the same effects as those of the above-mentioned fire detection system. [Brief description of the drawings]

[0038] [Figure 1] FIG. 1 is an explanatory diagram showing specific contents of an embodiment of a fire detection system. [Diagram 2] 2 is an explanatory diagram showing a more detailed functional configuration of an observation unit and a second processing unit in FIG. 1; [Diagram 3] 1 is a time chart showing an example of a comparison between first and second observed values ​​observed during normal monitoring. [Figure 4] 1 is a time chart showing an example of a comparison between first and second observation values ​​observed during a fire. [Diagram 5] FIG. 11 is an explanatory diagram showing specific contents of another embodiment of the fire detection system. [Figure 6]6 is an explanatory diagram showing a more detailed functional configuration of an observation unit and a second processing unit in FIG. 5. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0039] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of a fire detection system and a fire detection method according to the present invention will be described in detail with reference to the accompanying drawings. Note that the present invention is not limited to the following preferred embodiments.

[0040] [Basic Concept of the Embodiment] The embodiment generally relates to a fire detection system and a fire detection method. Here, the "fire detection system" is a means for observing an observation target (described later) in a monitoring area, detecting a fire based on the observation, and outputting a fire detection signal to, for example, a disaster prevention receiving panel, and is a concept including, for example, a fire detector, a fire detector, a fire alarm, etc. Those included in the fire detection system include a flame detection device (for example, a flame detector) that observes light from a flame to detect a fire, a smoke detection device (for example, a smoke detector) that observes smoke to detect a fire, a heat detection device (for example, a heat detector) that observes heat (specifically, for example, temperature) to detect a fire, and a gas detection device (for example, a gas detector) that observes gas (specifically, for example, gas concentration) to detect a fire.

[0041] Here, the "monitored area" refers to the area to be monitored by the fire detection system, and is an outdoor or indoor space with a certain extent, and is a concept that includes areas such as the inside of a structure such as a tunnel, a room, a corridor, or a staircase of a building.

[0042] Moreover, the "observation target" is an object observed by the fire detection system in the monitoring area, and the fire detection system detects a fire based on this observation data. The observation target is, for example, something that occurs with the physical and chemical phenomenon of a fire, and specific examples include light, smoke, temperature, gas, etc. A fire is detected based on, for example, a change in light observation data due to light being emitted with the flames of a fire (specifically, for example, a change in the infrared intensity emitted from the monitoring area), a change in smoke observation data due to smoke being generated with the burning of a fire (specifically, for example, a change in smoke concentration), a change in heat observation data due to heat being generated by burning (specifically, for example, a change in temperature), and a change in gas observation data due to gas being generated with burning (specifically, for example, a change in gas concentration).

[0043] As an example, the fire detection system is a disaster prevention system composed of a flame detection device and a disaster prevention receiving panel connected to the flame detection device, and includes an observation unit, a first processing unit, a second processing unit, and a fire detection unit. For example, the observation unit is disposed in the flame detection device, and the first processing unit and the second processing unit may be disposed anywhere in the system, but in the embodiment, the first processing unit is disposed in the flame detection device, and the second processing unit is disposed in the disaster prevention receiving panel.

[0044] Moreover, the fire detection system detects a fire in a monitoring area in a road tunnel, and is a component of a tunnel disaster prevention system. In other words, the tunnel disaster prevention system (tunnel disaster prevention equipment) is a system including the fire detection system of the present application.

[0045] The "observation unit" observes the emission of light from the monitored area. Here, "emission of light from the monitored area" is a concept that includes, for example, light emitted from flames in the event of a fire, and also includes emission of light of various wavelengths and intensities from sources other than flames, such as headlights of passing vehicles, flashing or blinking warning lights of emergency vehicles, lighting fixtures in tunnels, sunlight when installed at the entrance and exit of a tunnel, and workers (human bodies) involved in maintenance and inspection.

[0046] Furthermore, "observing light radiation" means, for example, observing light emitted from a flame in a first wavelength band specific to the flame to collect a first observation value, and observing light in a second wavelength band other than the first wavelength band specific to the flame to collect a second observation value, and then sampling each at a predetermined frequency and A / D converting the samples to generate first observation data and second observation data that are time-series data for each predetermined time. In the following description, the first observation value and the second observation value that are analog signals are indicated as lowercase x1 and x2, and the first observation data and the second observation data are indicated as uppercase X1 and X2.

[0047] In addition, the "first processing unit" judges whether or not there is a fire based on the observation data observed and collected by the observation unit, and in this embodiment, judges whether or not there is a fire based on the first fire judgment result obtained by executing a predetermined fire judgment process on the observation data.

[0048] The "second processing unit" quantitatively calculates the degree of fire-likeness based on the observation data observed and collected by the observation unit, and in this embodiment, quantitatively determines the degree of fire-likeness from the second fire judgment result calculated by executing a fire judgment process based on a predetermined machine learning process on the observation data. Here, the second fire judgment result based on the observation data when there is a fire is 1 (or 100%), and the second fire judgment result based on the observation data when there is no fire is 0 (or 0%), and the second fire judgment result ranges from 0 to 1 (or 0% to 100%) depending on the observation data.

[0049] In this embodiment, the second processing unit is a machine learning unit that inputs an observed value and performs binary classification to calculate and output a second fire judgment result, and is a learning machine (computer, etc.). Here, the "machine learning unit" is an algorithm that generates a learning model unit that learns the relationship between input and output from supervised learning data given in advance (data in which the correct output for the input is known), and after learning, inputs unknown data into the learning model unit to estimate the output, and is a concept that includes, for example, known multi-layer (deep) neural networks (DNN), random forests, support vector machines (SVC), etc.

[0050] In this embodiment, the machine learning unit collects and performs machine learning on the learning data in advance, namely, "observation data during a fire", where the correct answer (teacher) is when the second fire judgment result is 1, and "observation data during normal monitoring", where the correct answer (teacher) is when the second fire judgment result is 0, and inputs newly observed observation data after the machine learning to output the second fire judgment result.

[0051] Here, the "observation data during a fire", which is one type of learning data, is observation data observed when a fire occurs in the monitored area and the first processing unit judges that "it is a fire", and since the frequency of fire occurrence is extremely low, almost zero, it can be said that it is difficult or impossible to collect it as learning data. In contrast, the other type of learning data, "observation data during normal monitoring", is observation data observed when the first processing unit judges that "it is not a fire" during normal monitoring when no fire has occurred, and can be collected in large quantities simply and easily, and this embodiment improves the accuracy of the machine learning unit that inputs the observation data and outputs the second fire judgment result by performing machine learning using this "observation data during normal monitoring".

[0052] The fire detection system detects a fire based on the first fire judgment result by the first processing unit and the second fire judgment result by the second processing unit, and in this embodiment, a fire is detected when the first fire judgment result satisfies a predetermined condition (judgment that there is a fire) and the second fire judgment result is equal to or greater than a predetermined value. Also, when the first fire judgment result satisfies a predetermined condition (judgment that there is a fire) and the second fire judgment result is less than or equal to a predetermined value, or when the first fire judgment result does not satisfy a predetermined condition (judgment that there is no fire) and the second fire judgment result is equal to or greater than a predetermined value, a fire precursor is detected, for example.

[0053] In another embodiment of the fire detection system, the first processing unit obtains a first fire judgment result based on a first integral value for each specified period of the first observation data and a ratio between the first integral value and a second integral value for each specified period of the second observation data, and the second processing unit calculates a second fire judgment result based on the first integral value generated by the first processing unit and the ratio.

[0054] A specific embodiment will be described below. In the embodiment shown below, a case will be described in which the "monitoring area" is the "internal space of a tunnel", the observation unit, the first processing unit, and the fire detection unit of the "fire detection system" are provided in a "flame detection device", the second processing unit is provided in a disaster prevention receiving panel, the "observation data of light emission from the monitoring area" is "first observation data of light in a first wavelength band emitted from a flame due to a fire" and "second observation data of light in a second wavelength band other than the first wavelength band", and the "machine learning unit of the second processing unit" is a "multi-layer neural network".

[0055] [Specific details of the embodiment] Next, the specific contents of the embodiment will be described. The contents will be described separately as follows. Fire detection system b. Observation section c. First processing section d. Second processing section d1. Machine Learning Department d2. Learning model section d3. Learning control section d4. Collecting learning data d5. Machine Learning d6. Fire detection section e. Other embodiments of the fire detection system e1. Second processing section e2. Machine Learning Department e3. Learning model section e4. Learning control unit e5. Collection of learning data e6. Machine learning E7. Effect of reducing the number of inputs e8. Fire detection section f. Modifications of the present invention

[0056] [a. Fire detection system] First, an embodiment of the fire detection system will be described in more detail. As shown in Fig. 1, the fire detection system 10 (10-1) includes an observation unit 12, a first processing unit 14, a second processing unit 16, and a fire detection unit 18.

[0057] [b. Observation section] The observation unit 12 observes the emission of light from a monitored area, for example, the internal space of a tunnel, and outputs observation data. Although the configuration and functions thereof are arbitrary, in this embodiment, it observes light in a first wavelength band around 4.5 μm emitted from flames associated with a fire, and outputs first observation data X1, and also observes light in a second wavelength band other than the flames, for example, around 5.0 μm, and outputs second observation data X2.

[0058] More specifically, in this embodiment, the observation unit 12 is provided in a flame detection device 20 installed in a tunnel, as shown in Fig. 2. The observation unit 12 includes a light-transmitting window 25, a first observation unit 12a, a second observation unit 12b, and an observation data generation unit 13.

[0059] The first observation unit 12a receives light in a first wavelength band (around 4.5 μm) specific to flames from the light emitted from the tunnel space, which is selectively transmitted (passed) by an optical wavelength bandpass filter using the sensor unit 22a, converts it into an electrical signal, and performs predetermined processing such as amplification using the amplification processing unit 24a to obtain the first observation value x1, which is then output to the observation data generation unit 13.

[0060] The second observation unit 12b receives light in a second wavelength band (for example, near 5.0 μm) different from the first wavelength band, which is selectively transmitted (passed) by the optical wavelength bandpass filter from the light emitted from the tunnel space, by the sensor unit 22b, converts it into an electric signal, and performs predetermined processing such as amplification by the amplification processing unit 24b to obtain a second observation value x2, which is output to the observation data generation unit 13. The amplification processing units 24a and 24b are provided with a preamplifier, a frequency filter that selectively passes a predetermined frequency band including the flame flicker frequency, a main amplifier, and the like.

[0061] FIG. 3 is an example comparing a first observation value x1 and a second observation value x2 observed during normal monitoring, each showing a waveform change of the observation value with respect to the midpoint level for a predetermined unit time T, for example, T=2 seconds, observed as time series data. During normal monitoring when no fire is occurring, the amount of flame-specific light passing through the optical wavelength bandpass filter of the first wavelength band is small, so the amplitude change of the first observation value x1 is small. On the other hand, the amount of light due to factors other than flame passing through the optical wavelength bandpass filter of the second wavelength band is relatively large, so the amplitude change of the second observation value x2 is larger than that of the first observation value x1.

[0062] Figure 4 shows an example comparing the first observation value x1 and the second observation value x2 observed during a fire, in which the amount of flame-specific light that passes through the optical wavelength bandpass filter of the first wavelength band is large, so the amplitude change of the first observation value x1 is large. On the other hand, the amount of light due to factors other than flame that passes through the optical wavelength bandpass filter of the second wavelength band is relatively small, so the amplitude change of the second observation value x2 is smaller than that of the first observation value x1.

[0063] The observation data generation unit 13 samples the first observation value x1 and the second observation value x2, which are analog signals output from the first observation unit 12a and the second observation unit 12b, at a predetermined unit time, for example, every T=2 seconds as shown in Figure 3 or Figure 4, at a predetermined sampling frequency, for example, 128 Hz, and performs A / D conversion to generate 256 points of time series data, which are then read and stored as the first observation data X1 and the second observation data X2. In the following description, "the first observation data X1" and "the second observation data X2" refer to this "256 points of time series data."

[0064] [c. First Processing Section] The first processing unit 14 executes a predetermined fire judgment process on the first observation data X1 and the second observation data X2, which are time-series data every T = 2 seconds stored in the observation data generation unit 13, obtains the first fire judgment result E1, and outputs it to the fire detection unit 18. Although the first processing unit 14 has any configuration and function, in this embodiment, as shown in Figure 2, it is provided in the flame detection device 20 and is composed of a computer circuit equipped with a CPU, memory, and various input / output ports.

[0065] The observation data generating unit 13 provided in the observation unit 12 may be provided in the first processing unit 14, in which case the second processing unit 16 will acquire the first observation data X1 and the second observation data X2 from the first processing unit 14. Any fire judgment condition may be set in the fire judgment process in the first processing unit 14, but for example, the following two-stage fire judgment condition may be set.

[0066] The first stage fire judgment condition is, for example, when the integral value σX1 of the first observation data X1 is equal to or exceeds a predetermined threshold, a relative ratio (σX1 / σX2) between the integral value σX1 of the first observation data X1 and the integral value σX2 of the second observation data X2 is calculated, and when this relative ratio exceeds a predetermined threshold, the first stage fire judgment condition is deemed to be satisfied, and the process proceeds to the next second stage.

[0067] The second-stage fire judgment condition is to analyze the first observation data X1 by performing a fast Fourier transform (FFT) and calculate the relative ratio (σFL / σFH) between the integral value σFL of the low-frequency component (relative intensity) of 4 Hz or less and the integral value σFH of the high-frequency component (relative intensity) of more than 4 Hz and less than 8 Hz. If this relativization is equal to or exceeds a predetermined threshold, it is determined that the second-stage fire judgment condition is met, a fire is judged, and the first fire judgment result E1=1 is output. Moreover, if the first or second-stage fire judgment condition is not met, the first processing unit 14 judges that there is no fire and outputs the first fire judgment result E1=0.

[0068] [d. Second Processing Section] The second processing unit 16 executes a fire judgment process based on a predetermined machine learning process on the first observation data X1 and the second observation data X2, which are time series data every T = 2 seconds stored in the observation data generation unit 13, calculates a second fire judgment result E2, and outputs it to the fire detection unit 18. Although the configuration and function of the second processing unit 16 are arbitrary, in this embodiment, the second processing unit 16 is composed of a computer circuit equipped with a CPU, memory, and various input / output ports, and is provided, for example, on the disaster prevention receiving panel 40 side rather than the flame detection device 20 side, as shown in the second processing unit 16 of Figure 2, and is equipped with a machine learning unit 26.

[0069] (d1. Machine Learning Department) The machine learning unit 26 inputs observation data (here, the first observation data X1 and the second observation data X2) and outputs the second fire judgment result E2. Although its function, configuration, and type are arbitrary, in this embodiment, it is composed of a learning model unit 28, a learning control unit 30, and a learning data memory unit 32.

[0070] (d2. Learning model section) The learning model unit 28 is machine-learned using supervised learning data (data for which the output is the correct answer for the input), and the configuration and function thereof are arbitrary, but for example, it is a multi-layered neural network that performs binary classification. As is well known, a multi-layered neural network for binary classification is composed of an input layer, multiple intermediate layers, and an output layer, and a sigmoid function, for example, is used as the activation function of the output layer. In this embodiment, it is a multi-layered neural network for binary classification that classifies the degree Y1 of fire and the degree Y2 of non-fire.

[0071] The trained learning model unit 28 receives in parallel the first observation data X1 (X11, X12, . . . X1n) and the second observation data X2 (X21, X22, . . . X2n) for each unit time observed by the observation unit 12, calculates the degree Y1 of a fire and the degree Y2 of a non-fire, and outputs the degree Y1 of a fire as the second fire judgment result E2. The second fire judgment result E2 takes a value in the range of 0 to 1, with a minimum value of 0 when no fire has occurred (normal monitoring) and a maximum value of 1 when a fire has occurred.

[0072] (d3. Learning control unit) The learning control unit 30 performs machine learning on the learning model unit 28 using supervised learning data, and while its functions and configurations are arbitrary, in this embodiment it is configured as a computer circuit equipped with a CPU, memory, and various input / output ports, and has the function of collecting learning data and the function of performing machine learning on the learning model unit 28.

[0073] (d4. Collection of learning data) The learning control unit 30 collects first observation data X1 and second observation data X2 during a fire, and generates first learning data by combining teacher data (Y1, Y2) = (1, 0) in which the correct answer to the degree of a fire during a fire is Y1 = 1 and the correct answer to the degree of a non-fire is Y2 = 0, for example: (Y1,Y2)(X1,X2)=(1,0)(X1,X2) and stores the first learning data in the learning data storage unit 32. However, the occurrence of fires in tunnels is extremely rare and can be said to be virtually zero, and therefore, in reality, it is often difficult or impossible to collect the first learning data.

[0074] The learning control unit 30 also collects the first observation data X1 and the second observation data X2 during normal monitoring when no fire has occurred, and generates second learning data by combining teacher data (Y1, Y2) = (0, 1) in which the correct answer for the degree of fire during normal monitoring is Y1 = 0 and the correct answer for the degree of non-fire is Y2 = 1, for example: (Y1,Y2)(X1,X2)=(0,1)(X1,X2) and stores the second learning data in the learning data storage unit 32. This second learning data can be easily collected in large quantities during normal monitoring, and in this embodiment, the learning data essentially means the second learning data.

[0075] The learning control unit 30 collects the learning data (second learning data) for a predetermined period of time, for example, one to three months, preferably one year, after the start of operation of the fire detection system 10 (10-1). If the second processing unit 16 samples and reads the first observation data X1 and the second observation data X2 for a unit time T of 2 seconds at 128 Hz, as in the first processing unit 14, it can collect a large amount of learning data (second learning data), such as 43,200 pieces of learning data per day (24 hours) and 1,296,000 pieces of learning data per month (30 days).

[0076] (d5. Machine Learning) After the learning data collection period has elapsed, the learning control unit 30 reads out the learning data (second learning data) from the learning data storage unit 32 and performs machine learning of the learning model unit 28. Any method of machine learning may be used, but for example, the learning control unit 30 sets the number of nodes (e.g., 256 nodes) in the input layer of the multilayer neural network constituting the learning model unit 28, the number of stages and nodes in the intermediate layer, the number of nodes (e.g., 2 nodes) in the output layer, and the weights of each node, based on the user's operation. Next, the learning data (second learning data) is read out, the first and second observation data (X1, X2) are set as the input of the learning model unit 28, and the correct answer (0, 1) is set as the output (Y1, Y2), and machine learning is performed by a known backpropagation method.

[0077] The machine learning procedure is arbitrary, but may be, for example, the following procedure. (A) Divide the learning data into training data and validation data. 70-80% of the data is training data, and the remaining 20-30% is validation data. (a) Divide the training data into multiple blocks, for example, 10,000 blocks of data. (c) Divide the verification data into multiple blocks, for example, into blocks of 1,000 data. (E) The training data is read out in blocks and set in the learning model unit 28 for machine learning. (E) Verification data is read out in units of blocks, and input to the trained learning model unit 28 to perform verification to find the accuracy rate. The accuracy rate is the average value of the non-fire degree Y2 output by inputting the verification data. (F) As long as the accuracy rate increases, (D) and (E) are repeated. If the accuracy rate saturates or starts to decrease during the repetition, overlearning is considered and learning is terminated. Note that if the accuracy rate is less than a predetermined value, for example 80%, the accuracy is too low, so the learning model unit 28 is reconstructed and the process is started over from the beginning.

[0078] In addition, the function of the learning control unit 30 shown in Figure 2 may be provided as a function of a dedicated machine learning tool in a separate computer device or server etc. separated from the second processing unit 16, and machine learning of the learning model unit 28 may be performed in the machine learning tool in the same manner as described above, and the learned learning model unit 28 may be implemented (placed) in the second processing unit 16.

[0079] (d6. Fire detection section) The fire detection unit 18 detects a fire based on the first fire judgment result E1 by the first processing unit 14 and the second fire judgment result E2 by the second processing unit 16. For example, when the first fire judgment result E1 by the first processing unit 14 satisfies a predetermined condition for judging a fire (judged as a fire) and the second fire judgment result E2 by the second processing unit 16 is a predetermined value, for example, 0.6 or more, it judges that a fire is detected and transmits a fire detection signal E3 to the disaster prevention receiving panel 40 (for example, the control unit 42). The disaster prevention receiving panel 40 (for example, the alarm unit 44) that receives the fire detection signal E3 outputs a fire alarm to notify the public, and issues a no-entry alarm by an alarm display board or the like to prohibit vehicles from passing through the tunnel, and a management person goes to the site to confirm the fire and deal with it.

[0080] In addition, the fire detection unit 18 includes (1) When the first fire judgment result E1 by the first processing unit 14 satisfies a predetermined condition and the second fire judgment result E2 by the second processing unit 16 is less than a predetermined value, for example, 0.6, or (2) When the first fire judgment result E1 by the first processing unit 14 satisfies a predetermined condition and the second fire judgment result E2 by the second processing unit 16 is a predetermined value, for example, 0.6 or more, This is regarded as a fire sign detection, and a fire sign detection signal E4 is sent to a disaster prevention receiving panel (e.g., the control unit 42). The disaster prevention receiving panel 40 (e.g., the alarm unit 44) that receives the fire sign detection signal E4 outputs a fire warning alarm to call attention, enabling a management person to go to the site, confirm the fire, and deal with it.

[0081] [e. Other embodiments of the fire detection system] FIG. 5 is an explanatory diagram showing another embodiment of the fire detection system of the present invention, and FIG. 6 is an explanatory diagram showing a more detailed functional configuration of the observation unit and second processing unit of FIG.

[0082] The fire detection system 10 (10-2) of this embodiment is composed of an observation unit 12, a first processing unit 14, a second processing unit 16, and a fire detection unit 18. The observation unit 12, the first processing unit 14, and the fire detection unit 18 are basically the same as those in the embodiment of Fig. 1, but differs from the embodiment of Fig. 1 in that the second processing unit 16 receives an integral value (first integral value) σX1 of the first observation data X1 calculated by the first processing unit 14 and a relative ratio (σX1 / σX2) between this integral value σX1 and an integral value (second integral value) σX2 of the second observation data X2.

[0083] (e1. Second processing section) The second processing unit 16 inputs the integral value σX1 of the first observation data X1 generated by the first processing unit 14 and the relative ratio (σX1 / σX2) between the integral value σX1 and the integral value σX2 of the second observation data X2, executes a fire judgment process based on a predetermined machine learning process, calculates a second fire judgment result E2, and outputs it to the fire detection unit 18. For example, as shown in the second processing unit 16 in FIG. 6, the second processing unit 16 is equipped with a machine learning unit 26.

[0084] (e2. Machine Learning Department) The machine learning unit 26 inputs an integral value σX1 and a relative ratio (σX1 / σX2) based on the first and second observation data X1, X2, and outputs a second fire judgment result E2. The function, configuration, and type of the machine learning unit 26 are arbitrary, but it is composed of a learning model unit 28, a learning control unit 30, and a learning data storage unit 32, as in FIG. 2.

[0085] (e3. Learning model section) The learning model unit 28 is machine-learned using supervised learning data (data for which the output is the correct answer for the input), and the configuration and function thereof are arbitrary, but for example, it is a multi-layered neural network that performs binary classification. As is well known, a multi-layered neural network for binary classification is composed of an input layer, multiple intermediate layers, and an output layer, and a sigmoid function, for example, is used as the activation function of the output layer. In this embodiment, it is a multi-layered neural network for binary classification that classifies the degree Y1 of fire and the degree Y2 of non-fire.

[0086] The trained learning model unit 28 receives in parallel the integral value σX1 and the relative ratio (σX1 / σX2) calculated from the first and second observation data X1, X2 observed by the observation unit 12 for each unit time, calculates the degree of fire Y1 and the degree of non-fire Y2, and outputs the degree of fire Y1 as the second fire judgment result E2. The second fire judgment result E2 takes a value in the range of 0 to 1, with the minimum value 0 when no fire has occurred (normal monitoring) and the maximum value 1 when a fire has occurred.

[0087] (e4. Learning control unit) The learning control unit 30 performs machine learning on the learning model unit 28 using supervised learning data, and while its functions and configurations are arbitrary, in this embodiment it is configured as a computer circuit equipped with a CPU, memory, and various input / output ports, and has the function of collecting learning data and the function of performing machine learning on the learning model unit 28.

[0088] (e5. Collection of learning data) The learning control unit 30 collects an integral value σX1 and a relative ratio (σX1 / σX2) calculated based on the first and second observation data X1 and X2 during a fire, and generates first learning data by combining teacher data (Y1, Y2) = (1, 0) in which the correct answer to the degree of fire during a fire is Y1 = 1 and the correct answer to the degree of non-fire is Y2 = 0, for example: (Y1,Y2)(σX1,σX1 / σX2)=(1,0)(σX1,σX1 / σX2) and stores the first learning data in the learning data storage unit 32. However, the occurrence of fires in tunnels is extremely rare and can be said to be virtually zero, and therefore, in reality, it is often difficult or impossible to collect the first learning data.

[0089] The learning control unit 30 also collects the integral value σX1 and the relative ratio (σX1 / σX2) calculated based on the first and second observation data X1 and X2 during normal monitoring when no fire has occurred, and generates second learning data by combining teacher data (Y1, Y2) = (0, 1) in which the correct answer for the degree of fire during normal monitoring is Y1 = 0 and the correct answer for the degree of non-fire is Y2 = 1, for example: (Y1,Y2)(σX1,σX1 / σX2)=(0,1)(σX1,σX1 / σX2) and stores the second learning data in the learning data storage unit 32. This second learning data can be easily collected in large quantities during normal monitoring, and in this embodiment, the learning data essentially means the second learning data.

[0090] (e6. Machine Learning) After the learning data collection period has elapsed, the learning control unit 30 reads out the learning data (second learning data) from the learning data storage unit 32 and performs machine learning of the learning model unit 28. The method of this machine learning is arbitrary, but for example, the learning control unit 30 sets the structure of the multilayer neural network constituting the learning model unit 28, such as the number of nodes in the input layer (e.g., 256 nodes), the number of stages and nodes in the intermediate layer, and the number of nodes in the output layer (e.g., 2 nodes), and the weight of each node, based on the user's operation. Next, the learning data (second learning data) is read out, and the integral value σX1 and the relative ratio (σX1 / σX2) are set to the input of the learning model unit 28, and the correct answer (0, 1) is set to the output (Y1, Y2), and machine learning is performed by a known backpropagation method. The machine learning procedure is arbitrary, but is the same as that shown in the embodiment of FIG. 2.

[0091] (e7: Effect of reducing the number of inputs) Here, in the embodiment of FIG. 2, 256 observation values ​​consisting of the first observation data X1 and the second observation data X2 are input in parallel to the learning model unit 28 to process a 256-dimensional vector input, whereas in the embodiment of FIG. 6, two feature values ​​consisting of the integral value σX1 and the relative ratio (σX1 / σX2) are input in parallel to the learning model unit 28 to reduce it to a two-dimensional vector input, thereby significantly reducing the configuration and calculation processing of the machine learning unit 26, including the calculation processing of the learning control unit 30.

[0092] (e8. Fire detection section) The fire detection unit 18 detects a fire based on the first fire judgment result E1 by the first processing unit 14 and the second fire judgment result E2 by the second processing unit 16, in a manner similar to the embodiment of Figure 2.

[0093] [f. Modifications of the present invention] A detailed description will now be given of modified embodiments of the fire detection system and fire detection method according to the present invention.

[0094] (First processing section of the three-wavelength method) The above observation unit 12 uses a so-called two-wavelength system as an example, but may use a so-called three-wavelength system. In this case, the observation unit 12 detects a first observation value x1 of light in a first wavelength band (near 4.5 μm) specific to a flame, a second observation value x2 of light in a second wavelength band (for example, near 5.0 μm) other than the first wavelength band, and observes a third observation value x3 of light in a third wavelength band (for example, near 2.3 μm) other than the first and second wavelength bands.

[0095] When the observation unit 12 is a three-wavelength system, the fire detection process in the first processing unit 14 sets the fire detection conditions in three stages to detect a fire. The fire detection conditions in the first stage are the same as the fire detection conditions in the first stage of the two-wavelength system described above. The fire detection conditions in the second stage are calculated by calculating the relative ratio (σX1 / σX3) of the integral values ​​of the first observation data X1 and the third observation data X3, and if this relative ratio exceeds a predetermined threshold, it is determined that the fire detection conditions in the second stage are satisfied and the process proceeds to the third stage. The fire detection conditions in the third stage are the same as the fire detection conditions in the second stage of the two-wavelength system described above.

[0096] Furthermore, when the observation unit 12 is a three-wavelength type, the machine learning processing in the second processing unit 16 is configured so that the learning machine unit 26 inputs the first observation data X1, the second data X2, and the third observation data X3 to calculate the degree of fire Y1 and the degree of non-fire Y2, collects the first learning data during a fire and the second learning data during normal monitoring and performs machine learning, and then inputs the newly observed observation data X1, X2, and X3 to classify the degree of fire Y1 and the degree of non-fire Y2, and outputs the degree of fire Y1 as the second fire judgment result E2.

[0097] As another embodiment in which the observation unit 12 is of the three-wavelength type, the machine learning processing in the second processing unit 16 is configured to input the learning machine unit 26 with the integral value σX1 of the first observation data X1, the relative ratio (σX1 / σX2) between the integral value σx1 and the integral value σX2 of the second data X2, and the relative ratio (σX1 / σX3) between the integral value σx1 and the integral value σX3 of the third data X3 to calculate the degree of fire Y1 and the degree of non-fire Y2, and to collect and machine-learn the first learning data during a fire and the second learning data during normal monitoring, and then input newly observed observation data X1, X2, and X3 to classify the degree of fire Y1 and the degree of non-fire Y2, and output the degree of fire Y1 as the second fire judgment result E2.

[0098] (Machine Learning Department) In the above embodiment, the machine learning unit 26 of the second processing unit 16 outputs the degree of fire Y1 as the second fire judgment result E2 to the fire detection unit 18, but instead may output the degree of non-fire Y2 as the second fire judgment result E2. In this case, the fire detection unit 18 determines that a fire has been detected when the first processing unit 14 satisfies a predetermined condition for determining that there is a fire (determines that there is a fire) and the second fire judgment result E2 by the second processing unit 16 is equal to or less than a predetermined value, for example, 0.4.

[0099] (Fire alarm system) The above embodiment has been described with reference to a tunnel disaster prevention system, but is not limited thereto and may be applied to any system, for example, a fire alarm system that monitors fires by connecting a sensor to a receiver. In the fire detection system of the present embodiment in a fire alarm system, for example, a flame detection device may be installed in a room of a building that is a monitoring area, or may be installed outdoors to monitor arson, etc.

[0100] (others) Furthermore, the present invention includes appropriate modifications that do not impair the objects and advantages of the present invention, and is not limited to the numerical values ​​shown in the above embodiment. [Explanation of symbols]

[0101] 10(10-1), 10(10-2): Fire detection systems 12: Observation section 12a: First observation section 12b: Second Observation Section 13: Observation data generation unit 14: First processing section 16: Second processing section 18: Fire detection section 20: Flame detection device 22a, 22b: Sensor section 24a, 24b: Amplification processing section 25: Translucent window 26: Machine Learning Department 28: Learning model section 30: Learning control unit 32: Learning data storage unit 40: Disaster prevention receiving panel 42: Control unit 44: Alarm section

Claims

1. a fire detection unit that detects a fire in the monitored area based on a first fire judgment result obtained by executing a predetermined fire judgment process on observation data collected by observing an observation target in the monitored area and a second fire judgment result calculated by executing a fire judgment process based on a predetermined machine learning process on the observation data; an observation unit that observes radiation of light from the monitoring area; a first processing unit that executes a predetermined fire determination process on the observation data observed and collected by the observation unit to obtain the first fire determination result; A fire detection device comprising: A second processing unit that executes a fire determination process based on a predetermined machine learning process on the observation data to calculate the second fire determination result; A fire detection system comprising:

2. 2. The fire detection system of claim 1, A fire detection system, characterized in that the second processing unit is disposed in a location different from the monitored area.

3. 3. The fire detection system according to claim 1, The observation unit includes: First observation data of light in a first wavelength band emitted from a flame associated with the fire; Second observation data of light in a second wavelength band other than the first wavelength band; and Third observation data of light in a third wavelength band other than the first wavelength band and the second wavelength band; and Collect The first processing unit acquires the first fire judgment result based on the first observation data, the second observation data, and the third observation data, The second processing unit calculates the second fire determination result based on the first observation data, the second observation data, and the third observation data. A fire detection system comprising:

4. 3. The fire detection system according to claim 1, The observation unit includes: First observation data of light in a first wavelength band emitted from a flame associated with the fire; Second observation data of light in a second wavelength band other than the first wavelength band; and Third observation data of light in a third wavelength band other than the first wavelength band and the second wavelength band; and Collect the first processing unit acquires the first fire judgment result based on a first integral value of the first observation data for each predetermined period, a ratio between the first integral value and a second integral value of the second observation data for each predetermined period, and a ratio between the first integral value and a third integral value of the third observation data for each predetermined period; The second processing unit calculates the second fire judgment result based on the first integral value generated by the first processing unit, the ratio between the first integral value and the second integral value, and the ratio between the first integral value and the third integral value. A fire detection system comprising:

Images