Fire precursor detection system, fire precursor detection method, and fire precursor detection program

Abstract

This invention provides a fire precursor detection system, a fire precursor detection method, and a fire precursor detection program that can detect signs of fire at an early stage. [Solution] The fire precursor detection system is a fire precursor detection system that is installed in an area to be monitored and is capable of communicating with multiple types of sensors that measure the environment of the area and detects signs of fire in the area, and is characterized by comprising: a measurement data acquisition unit that acquires measurement data for each of multiple indicators of the environment of the area acquired by multiple types of sensors; a precursor presence / absence determination unit that determines whether or not there are signs of fire in the area based on a combination of measurement data of predetermined indicators selected from the measurement data acquired by the measurement data acquisition unit; and a notification unit that notifies the area monitor that there are signs of fire in the area when the precursor presence / absence determination unit determines that there are signs of fire in the area.

Description

Technical Field

[0001] The present invention relates to a fire omen detection system, a fire omen detection method, and a fire omen detection program, and particularly to a fire omen detection system, a fire omen detection method, and a fire omen detection program capable of early detecting a fire prediction.

Background Art

[0002] Buildings such as buildings are equipped with fire alarm equipment and fire fighting equipment such as sprinklers, but in the construction site of a building under construction, the installation of fire alarm equipment and fire fighting equipment may not be installed, and as a countermeasure against fire, there may be only patrol by security guards and a fire fighting water bucket.

[0003] For example, the construction site of a building under construction may catch fire due to a fire of a cigarette, sparks generated during welding, or scattering of molten metal. These careless fires are difficult to notice because they are smoldering in the time zone when people are present and are not conspicuous, and may flare up and spread several hours later when people are not present. From the above, it is particularly desirable to detect a fire early at such a construction site, and equipment that can detect early signs of a fire such as a smoldering state of the fire is required. In addition, there may be no equipment for detecting a fire in outdoor and field facilities, and due to carelessness of a fire during the time when people are present during the day, a fire may break out several hours later when people are not present. Therefore, for outdoor and field facilities as well, equipment that can detect early signs of a fire is required.

[0004] In Patent Document 1, a fire detection system that senses the presence or absence of smoke contained in the air has been proposed. Since the amount of smoke is small in a smoldering state of a fire indicating a fire omen, in the fire detection system disclosed in Patent Document 1, it may be difficult to sense a smoldering state of a fire, and there is a risk that a fire omen cannot be discovered. There is still a need for a proposal of a fire omen detection system that can detect early signs of a fire such as a smoldering state of a fire. [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2023-075361 [Overview of the Initiative] [Problems that the invention aims to solve]

[0006] Therefore, the present invention aims to provide a fire precursor detection system, a fire precursor detection method, and a fire precursor detection program that can detect signs of fire at an early stage. [Means for solving the problem]

[0007] In other words, the fire precursor detection system according to the first embodiment is a fire precursor detection system that is installed in an area to be monitored and is capable of communicating with a plurality of types of sensors that measure the environment of the area and detects signs of fire in the area, and is characterized by comprising: a measurement data acquisition unit that acquires measurement data for each of a plurality of indicators of the environment of the area acquired by the plurality of types of sensors; a precursor presence / absence determination unit that determines whether or not there are signs of fire in the area based on a combination of measurement data of a predetermined indicator selected from the measurement data acquired by the measurement data acquisition unit; and a notification unit that notifies the area monitor that there are signs of fire in the area when the precursor presence / absence determination unit determines that there are signs of fire in the area.

[0008] In a second embodiment, the fire precursor detection system according to the first embodiment may use a precursor learning model that has previously learned the correspondence between a combination of measurement data of a predetermined indicator and the presence or absence of fire precursors in an area, and input a combination of measurement data of a predetermined indicator selected from the measurement data acquired by the measurement data acquisition unit into the precursor learning model to determine whether or not there are fire precursors in an area.

[0009] A third embodiment may be a fire precursor detection system according to the first embodiment, further comprising a reception unit that receives the type of area to be monitored, wherein the precursor determination unit refers to a table that defines the correspondence between the type of area to be monitored and the combination of measurement data for an indicator, and determines the combination of measurement data for an indicator based on the type of area to be monitored received by the reception unit.

[0010] A fourth embodiment is a fire precursor detection system according to the first embodiment, further comprising a risk determination unit that determines the degree of danger of a fire precursor based on a combination of measurement data of predetermined indicators that formed the basis of the determination by the precursor presence determination unit when the precursor presence determination unit determines that there is a fire precursor in the area, and a notification unit that, when the precursor presence determination unit determines that there is a fire precursor in the area, notifies the area monitor of the presence of a fire precursor and the degree of danger of the precursor.

[0011] A fifth embodiment is a fire precursor detection system according to the fourth embodiment, in which the risk determination unit uses a risk learning model that has been pre-learned the correspondence between a combination of measurement data of a predetermined indicator and the risk of fire precursors in a region, and inputs the combination of measurement data of a predetermined indicator that formed the basis for the determination by the precursor presence / absence determination unit that there are fire precursors in the region into the risk learning model, thereby determining the risk of fire precursors in the region.

[0012] The sixth embodiment is a fire prediction detection system according to the first embodiment, wherein one or more thresholds are set for the measurement data of each of a plurality of indicators, and measurement data that exceeds the threshold for each of the indicators is acquired from each of a plurality of types of sensors.

[0013] A seventh aspect is a fire prediction detection system according to the first aspect, in which multiple types of sensors may include odor sensors that measure the smell of an area.

[0014] The eighth aspect is a fire precursor detection system according to the first aspect, wherein the multiple types of sensors may include a two-dimensional thermopile radiation temperature sensor that measures a two-dimensional temperature distribution by dividing a two-dimensional plane of an area into predetermined sections and indicating the temperature of each section as a numerical value.

[0015] The ninth aspect is a fire prediction detection system according to the first aspect, in which the multiple types of sensors may include a barometric pressure sensor that measures the atmospheric pressure in a region.

[0016] A tenth embodiment is a fire precursor detection system according to the first embodiment, in which the multiple types of sensors may include a temperature and humidity sensor that measures the temperature and humidity of the area in a steady state.

[0017] An eleventh embodiment is a fire precursor detection system according to the first embodiment, in which multiple types of sensors may include an air pollution sensor that measures the level of air pollution in a given area.

[0018] A twelfth embodiment is a fire prediction detection system according to the first embodiment, in which the multiple types of sensors may include smoke sensors that detect the concentration of smoke in an area.

[0019] A thirteenth embodiment is a fire precursor detection system according to the first embodiment, in which the multiple types of sensors may include a carbon monoxide sensor that detects the concentration of carbon monoxide in a given area.

[0020] A fourteenth aspect is a fire prediction detection system according to the first aspect, in which the multiple types of sensors may include a carbon dioxide sensor that detects the concentration of carbon dioxide in a given area.

[0021] The 15th embodiment is a fire precursor detection system according to the first embodiment, in which the multiple types of sensors may include an illuminance sensor for measuring the illuminance of an area.

[0022] The 16th aspect may be such that, in the fire omen detection system according to the 1st aspect, the plurality of types of sensors include a noise sensor that measures the magnitude of the noise in the area.

[0023] The 17th aspect may be such that, in the fire omen inspection system according to the 1st aspect, the plurality of types of sensors include an oxygen sensor that measures the concentration of oxygen in the area.

[0024] The 18th aspect may be such that, in the fire omen inspection system according to the 1st aspect, the plurality of types of sensors are mounted on an unmanned aircraft or a robot, and the unmanned aircraft or the robot flies or travels in the area by remote operation or autonomous control.

[0025] The 19th aspect may be such that, in the fire omen inspection system according to the 1st aspect, by performing unsupervised clustering on the measurement data acquired by the measurement data acquisition unit, an outlier of the measurement data of each of the plurality of indicators of the environment of the area included in the measurement data is extracted as an abnormal value of the measurement data of each of the plurality of indicators, and further includes an abnormal value extraction unit, and an abnormal value omen determination unit that determines the presence or absence of a fire omen in the area based on the abnormal values.

[0026] The 20th aspect may be such that, in the fire omen inspection system according to the 6th aspect, by performing unsupervised clustering on the measurement data acquired by the measurement data acquisition unit, an outlier of the measurement data of each of the plurality of indicators of the environment of the area included in the measurement data is extracted as an abnormal value of the measurement data of each of the plurality of indicators, and further includes an abnormal value extraction unit and a threshold resetting unit that resets the threshold of the measurement data of each of the plurality of indicators based on the abnormal values. The measurement data acquisition unit may acquire, from each of the plurality of types of sensors, measurement data that exceeds the threshold reset by the threshold resetting unit for each of the indicators.

[0027] The fire precursor detection method according to the 21st embodiment is a fire precursor detection method that can communicate with multiple types of sensors installed in an area to be monitored and measure the environment of the area, and detects signs of fire in the area, characterized in that the computer performs: a measurement data acquisition step of acquiring measurement data for each of multiple indicators of the environment of the area acquired by multiple types of sensors; a precursor presence / absence determination step of determining whether or not there are signs of fire in the area based on a combination of measurement data of predetermined indicators selected from the measurement data acquired in the measurement data acquisition step; and a notification step of notifying the area monitor that there are signs of fire if it is determined in the precursor presence / absence determination step that there are signs of fire in the area.

[0028] The fire precursor detection program according to the 22nd embodiment is a fire precursor detection program that is installed in an area to be monitored and is capable of communicating with multiple types of sensors that measure the environment of the area and detects signs of fire in the area, characterized in that it causes a computer to perform: a measurement data acquisition function that acquires measurement data for each of multiple indicators of the environment of the area acquired by multiple types of sensors; a precursor presence / absence determination function that determines whether or not there are signs of fire in the area based on a combination of measurement data of predetermined indicators selected from the measurement data acquired by the measurement data acquisition function; and a notification function that notifies the area monitor that there are signs of fire in the area when the precursor presence / absence determination function determines that there are signs of fire in the area. [Effects of the Invention]

[0029] The fire precursor detection system according to the present invention is a fire precursor detection system that is installed in an area to be monitored and is capable of communicating with multiple types of sensors that measure the environment of the area and detects signs of fire in the area, and is characterized by comprising: a measurement data acquisition unit that acquires measurement data for each of multiple indicators of the environment of the area acquired by multiple types of sensors; a precursor presence / absence determination unit that determines whether or not there are signs of fire in the area based on a combination of measurement data of predetermined indicators selected from the measurement data acquired by the measurement data acquisition unit; and a notification unit that notifies the area monitor that there are signs of fire in the area when the precursor presence / absence determination unit determines that there are signs of fire in the area, so that signs of fire can be detected early. [Brief explanation of the drawing]

[0030] [Figure 1] Figure 1 is a diagram illustrating the overview of the fire prediction detection system according to this embodiment. [Figure 2] Figure 2 is an example of an infrastructure configuration diagram including the fire prediction detection system according to this embodiment. [Figure 3] Figure 3 is a diagram illustrating an example of how the environmental measurement terminal is installed according to this embodiment. [Figure 4] Figure 4 is a block diagram illustrating an example of the hardware configuration of the fire prediction detection system according to this embodiment. [Figure 5] Figure 5 is a block diagram illustrating an example of the hardware configuration of the environmental measurement terminal according to this embodiment. [Figure 6a] Figure 6a is a diagram illustrating an example of the circuit configuration of the environmental measurement terminal according to this embodiment. [Figure 6b] Figure 6b is a diagram illustrating an example of the circuit configuration of the environmental measurement terminal according to this embodiment. [Figure 6c] Figure 6c is a diagram illustrating an example of the circuit configuration of the environmental measurement terminal according to this embodiment. [Figure 6d] Figure 6d is a diagram illustrating an example of the circuit configuration of the environmental measurement terminal according to this embodiment. [Figure 7]Figure 7 is a block diagram illustrating an example of the functional configuration of the fire prediction detection system according to this embodiment. [Figure 8] Figure 8 shows an example of a table that defines the correspondence between the types of areas to be monitored according to this embodiment and combinations of measurement data for multiple environmental indicators of those areas. [Figure 9] Figure 9 is a diagram illustrating an example of a combination of measurement data for multiple environmental indicators of the area to be monitored according to this embodiment. [Figure 10] Figure 10 is a diagram illustrating an example of experimental results for the fire prediction detection system according to this embodiment. [Figure 11] Figure 11 is an example of a flowchart for a fire precursor detection program according to this embodiment. [Figure 12] Figure 12 is an example of a flowchart for a fire precursor detection program according to another embodiment. [Modes for carrying out the invention]

[0031] (Overview of Fire Prediction Detection System 10) An embodiment of the fire prediction detection system 10 according to this disclosure will be described with reference to Figures 1 to 10. First, with reference to Figure 1, an overview of the fire prediction detection system 10 according to this embodiment will be described.

[0032] The fire precursor detection system 10 is installed in the area to be monitored and is capable of communicating with multiple types of sensors that measure the environment of that area. It measures the environment of the area using multiple types of sensors and detects signs of fire. Multiple types of sensors are implemented in the environmental measurement terminal 20, which will be described later. The multiple types of sensors are controlled by the CPU 20e of the environmental measurement terminal 20, and the measurement data from the multiple types of sensors is transmitted to the fire prediction detection system 10.

[0033] The environmental measurement terminal 20 is installed in the area to be monitored, and multiple types of sensors are installed in that area. It should be noted that, even if multiple types of sensors are not built into the environmental measurement terminal 20, they may be installed separately in the area to be monitored, away from the environmental measurement terminal 20, and the environmental measurement terminal 20 acquires their measurement data and transmits it to the fire prediction detection system 10. The areas to be monitored are those where signs of fire need to be detected, and are envisioned to include buildings, logistics warehouses, and construction sites of various factories. However, the areas to be monitored are not limited to construction sites; any location where fire needs to be monitored and signs of fire need to be detected can be included, and a wide range of areas can be included. For example, this could include houses, nursing homes, childcare facilities, power plants, forests, dams, fishing ports, railway tracks, and airport runways, and it can be during or after construction work is completed. Large outdoor facilities that do not involve buildings, such as fields, parks, grounds, and parking lots, can also be included in the areas to be monitored. Furthermore, the interiors of moving objects such as buses, trains, ships, airplanes, submarines, and spacecraft can also be included in the areas to be monitored. If the area to be monitored exceeds the measurable range of the multiple types of sensors implemented in a single environmental measurement terminal 20, multiple environmental measurement terminals 20 are installed in the area to be monitored to cover the entire area.

[0034] The multiple types of sensors in this embodiment, as shown in Figure 5, include an odor sensor 21, a two-dimensional thermopile radiation temperature sensor 22, a pressure sensor 23, a temperature and humidity sensor 24, an air pollution sensor 25, a smoke sensor 26, a carbon monoxide sensor 27, a carbon dioxide sensor 28, an illuminance sensor 29, a noise sensor 30, a night vision camera 31, and an oxygen sensor 32. These sensors are used to measure various indicators of the environment in the area to be monitored. Furthermore, due to improvements to the fire prediction detection system 10 or the environmental measurement terminal 20, some of these various types of sensors may become obsolete and no longer be installed in the environmental measurement terminal 20, or they may be replaced with other sensors.

[0035] The odor sensor 21 measures the odor of the area being monitored. The odor sensor 21 detects a "smoky smell" among many other smells, which is a smell that occurs when a fire is smoldering (smoking state) as a precursor to a fire. It measures the intensity of this smell, quantifies it, and outputs it. Odor refers to a state in which odor substances are in the form of particulate matter or gas. The principles of the odor sensor 21 include semiconductor methods, quartz crystal oscillator methods, FET biosensor methods, and MSS methods. When a pulse voltage is applied to the odor sensor 21, a sensor resistance value corresponding to the concentration of odor molecules is detected. The analog voltage generated across the sensor resistance of the odor sensor 21 is input to the CPU 20e of the environmental measurement terminal 20, which will be described later.

[0036] The semiconductor method derives the concentration of odor molecules by measuring the change in resistance when odor molecules adsorb to oxygen on the semiconductor surface. While metal oxide semiconductors are the most common type of semiconductor, methods using organic semiconductors also exist. The quartz crystal oscillator method works by measuring the decrease in the resonant frequency of a quartz crystal oscillator when a sensitive film attached to it captures odor molecules. Natural or synthetic lipids are used for the sensitive film. By using multiple sensors with sensitive films simultaneously, it is possible to identify multiple odor molecules.

[0037] The FET biosensor method uses probe molecules fixed on the insulating film of a transistor to detect the charge of adsorbed gas molecules. By analyzing the electrical signals resulting from these charges, the components and concentrations of the gas molecules can be determined. FET stands for Field Effect Transistor. The MSS method works by analyzing the change in electrical resistance caused by the adsorption of gas molecules onto a sensitive membrane, thereby determining the composition and concentration of the gas molecules. MSS is an abbreviation for Membrane-type Surface Stress Sensor.

[0038] The 2D thermopile radiation temperature sensor 22 measures a 2D temperature distribution by dividing the 2D plane of the area to be monitored into predetermined sections and displaying the temperature of each section as a numerical value. The 2D thermopile radiation temperature sensor 22 is capable of measuring the 2D temperature distribution at a distance. The 2D thermopile radiation temperature sensor 22 measures the 2D heat distribution by collecting infrared radiation energy emitted from a distant material using a lens and irradiating a total of 64 thermopiles arranged in an 8x8 grid. Therefore, the 2D thermopile radiation temperature sensor 22 divides the area to be monitored into 64 sections arranged in an 8x8 grid, and measures the 2D temperature distribution by associating the detected temperature of each of the 64 thermopiles with the temperature of each section.

[0039] The 2D thermopile radiation temperature sensor 22 digitally outputs the measured temperature data via I2C communication in a format that responds to data requests from the CPU 20e of the environmental measurement terminal 20 (described later), transmitting the temperatures detected by the 64 thermopiles as 64 temperature data points at predetermined intermittent intervals. I2C (Inter-Integrated Circuit) communication is one of the communication interfaces for synchronous serial communication, which transmits data in synchronization with a clock signal. Note that the communication method of the 2D thermopile radiation temperature sensor 22 is not limited to I2C communication, and may be changed to other communication methods through specification changes or improvements to the 2D thermopile radiation temperature sensor 22.

[0040] This intermittent time can be switched by a rotary switch incorporated into the sensor circuit, which includes the 2D thermopile radiation temperature sensor 22, allowing the intermittent time to be switched to intervals of 4 seconds, 10 seconds, 30 seconds, or 60 seconds. However, the intermittent time is not limited to these and may be other intervals. Furthermore, the intermittent time can be synchronized with the timing of detecting a LOW level of the TX signal of communication via the TX interface, and the 2D thermopile radiation temperature sensor 22 can output and transmit 64 temperature data points at the timing of detecting a LOW level of the TX signal.

[0041] The two-dimensional thermopile radiation temperature sensor 22 starts measuring when a measurement start push button switch, which is incorporated into the sensor circuit including the two-dimensional thermopile radiation temperature sensor 22, is pressed.

[0042] The pressure sensor 23 measures the atmospheric pressure in the area being monitored. Since the output values ​​of various sensors fluctuate depending on the atmospheric pressure, the output values ​​of the various sensors are corrected based on the measured values ​​of the atmospheric pressure in the monitored area, which are measured by the pressure sensor 23. The pressure sensor 23 is an 8-pin DIP module. The measured pressure is internally corrected and converted, and the pressure (hPa) is obtained by dividing the 24-bit reading by 4096. The measurement data from the pressure sensor 23 is output digitally via I2C communication in a format that responds to data requests from the CPU 20e of the environmental measurement terminal 20, which will be described later. Furthermore, the communication method of the pressure sensor 23 is not limited to I2C communication, and may be changed to other communication methods through changes or improvements to the specifications of the pressure sensor 23.

[0043] The temperature and humidity sensor 24 measures the temperature and humidity of the area being monitored. The temperature and humidity sensor 24 measures the temperature and humidity of the region in a steady state, enabling monitoring of changes in temperature and humidity due to smoldering from a fire. The temperature and humidity sensor 24 is a composite sensor module that simultaneously measures temperature and humidity, and outputs digital data in a format that responds to data requests from the CPU 20e of the environmental measurement terminal 20 (described later) via I2C communication. Furthermore, the communication method of the temperature and humidity sensor 24 is not limited to I2C communication, and may be changed to other communication methods through specification changes or improvements to the temperature and humidity sensor 24.

[0044] The air pollution sensor 25 measures the level of air pollution in the area being monitored. The air pollution sensor 25 can detect carbon monoxide, hydrogen, ammonia, methane, ethanol, propane, and isobutane, which are considered to be causes of air pollution, but it cannot identify the type of gas detected and therefore lacks gas selectivity. Although the air pollution sensor 25 of this embodiment does not have the ability to individually identify and detect multiple types of gases, a gas sensor that can individually identify and detect multiple types of gases to be detected may be used as the air pollution sensor. Alternatively, different types of gas sensors may be used for each of the multiple types of gases to be detected.

[0045] The air pollution sensor 25 operates with a 5V DC voltage and outputs the measured value as an analog voltage. When the air pollution sensor 25 detects gas, it amplifies the output voltage according to the proportion of gas present. The heater current consumption of the air pollution sensor 25 is 25-35mA, and its operation can be stopped using the EN pin to reduce current consumption. The output terminal of the air pollution sensor 25 is connected to a 10x differential amplifier circuit with offset via a buffer circuit. The output terminal of this 10x differential amplifier circuit is connected to the CPU 20e of the environmental measurement terminal 20 described later, and the analog voltage of the air pollution sensor 25 is input to the CPU 20e of the environmental measurement terminal 20 described later. The buffer circuit is used to correct the output voltage and signal strength of the air pollution sensor 25.

[0046] The smoke sensor 26 detects the concentration of smoke in the area being monitored. The smoke sensor 26 is an optical particulate sensor that detects fine particulate matter (PM2.5). Fine particulate matter (PM2.5) refers to very small particles suspended in the atmosphere, with a size of 2.5 micrometers or less. Its components include carbon components, nitrates, sulfates, ammonium salts, as well as inorganic elements such as silicon, sodium, and aluminum.

[0047] The smoke sensor 26 detects carbon monoxide, carbon dioxide, hydrogen cyanide (prussic acid gas), and chlorine gas, which are the main components of fire smoke. The smoke sensor 26 emits light from a light-emitting element (LED) into the air flowing in through a through-hole in its central part. The light scattered by minute particles (smoke components) is detected by a photodetector (photodiode), and the output signal of the photodetector is amplified and output. The light from the light-emitting element is focused by a lens and a slit. The photodetector is also focused by a lens and a slit, similar to the light-emitting part, to cut out unwanted light as much as possible and efficiently receive the detected light. The point where these optical axes intersect becomes the detection area.

[0048] Even when no fine particles are present, the light-receiving element receives unwanted light and generates an output voltage. The light-receiving element provides an output current proportional to the amount of light received, which is amplified by an amplification circuit and output as an analog voltage (pulse output) proportional to the dust concentration. The output terminal of the smoke sensor 26 is connected to the input terminal of a peak hold circuit, and the output terminal of the peak hold circuit is connected to the input terminal of an LPF (Low Pass Filter) circuit. The output terminal of the LPF circuit is connected to the CPU 20e of the environmental measurement terminal 20, which will be described later. A peak hold circuit is a circuit that holds the maximum value for a certain period of time or until it is reset.

[0049] The carbon monoxide sensor 27 detects the concentration of carbon monoxide in the area being monitored. The carbon monoxide sensor 27 is selective and capable of measuring the absolute value of carbon monoxide. The carbon monoxide sensor 27 is an electrochemical gas sensor. Electrochemical gas sensors utilize the principle of converting chemical reactions into electric current and voltage to generate an electrical signal proportional to the amount of the substance being measured. The carbon monoxide sensor 27 exhibits a linear relationship between carbon monoxide concentration and sensor output current, and can be used for quantitative measurement of carbon monoxide concentration by calibrating the sensor output current with gas of known carbon monoxide concentration.

[0050] The voltage across the load resistor connected to the carbon monoxide sensor 27 is connected to the input terminal of the differential amplifier circuit. The output terminal of the differential amplifier circuit is connected to the input terminal of the LPF circuit, and the output terminal of the LPF circuit is connected to the CPU 20e of the environmental measurement terminal 20, which will be described later.

[0051] The carbon dioxide sensor 28 detects the concentration of carbon dioxide in the area being monitored. The carbon dioxide sensor 28 is an NDIR (non-dispersive infrared) type carbon dioxide sensor, integrated as a module, operating on a 5V DC power supply and outputting an analog voltage PWM. NDIR (non-dispersive infrared) gas sensors detect gases by utilizing the phenomenon where emitted infrared radiation causes molecular vibrations in the target gas (carbon dioxide), resulting in the absorption of infrared radiation at a specific wavelength. The infrared transmittance (ratio of transmitted light intensity to radiation intensity from the source) is determined by the concentration of carbon dioxide, which is the target gas. Carbon dioxide molecules have the property of absorbing the 4.26 μm wavelength in the infrared region, and the higher the carbon dioxide concentration, the more infrared radiation at 4.26 μm is absorbed. Therefore, the carbon dioxide concentration can be measured by measuring the intensity of the infrared radiation. The output terminal of the PWM output of the carbon dioxide sensor 28 is connected to the CPU 20e of the environmental measurement terminal 20, which will be described later.

[0052] The illuminance sensor 29 measures the illuminance of the area to be monitored. The illuminance sensor 29 enables early, highly sensitive detection of fires at night. The illuminance sensor 29 is a phototransistor that amplifies and outputs the photocurrent generated by the irradiated light. The illuminance sensor 29 is connected to a temperature compensation circuit, which corrects output characteristics that change with the operating temperature (temperature drift). The output terminal of the illuminance sensor 29 is connected to the CPU 20e of the environmental measurement terminal 20, which will be described later. Temperature drift refers to the variation in output characteristics in response to changes in ambient temperature.

[0053] The noise sensor 30 measures the level of noise in the area being monitored. The noise sensor 30 is an ECM (electret condenser microphone). The output voltage of the noise sensor 30 is amplified by an amplification circuit and then converted to an RMS (root mean square) value, which is one of the indicators of the magnitude of an AC electrical signal, by an RMS (root mean square) conversion circuit. After that, it passes through a 10Hz LPF circuit and a level conversion circuit and is input to the CPU 20e of the environmental measurement terminal 20 described later. The RMS value is a value that indicates the effective magnitude of an AC electrical signal and is used to compare the magnitude of the output voltage of the noise sensor 30.

[0054] The night vision camera 31 captures images of the area being monitored. The night vision camera 31 is equipped with either a night vision correction function or an infrared function to enable shooting even in dark areas. The night vision correction function is a function that automatically amplifies sensitivity even in dimly lit places with few light sources. The infrared function is a function that visualizes infrared radiation emitted from objects. The output terminal of the night vision camera 31 is connected to the CPU 20e of the environmental measurement terminal 20, which will be described later, and the captured data from the night vision camera 31 is input to the CPU 20e. Furthermore, the environmental measurement terminal 20 may be equipped with a thermal camera instead of the night vision camera 31, or it may be equipped with a thermal camera in addition to the night vision camera 31. A thermal camera is a type of infrared camera that captures infrared radiation emitted by a subject, enabling the measurement of the subject's temperature, and is also called a thermographic camera. All objects above absolute zero have the characteristic of emitting infrared radiation corresponding to their temperature, and infrared cameras are a general term for cameras that capture and photograph the infrared radiation emitted by a subject. Therefore, the environmental measurement terminal 20 can measure the temperature of the area to be monitored and can detect abnormalities such as high temperatures in that area. In addition, the environmental measurement terminal 20 can detect specific gases, such as hydrocarbon gases like methane, carbon monoxide, carbon dioxide, and ammonia, and can detect the generation of these specific gases in the area to be monitored.

[0055] The oxygen sensor 32 measures the oxygen concentration in the area being monitored. The oxygen sensor 32 is used for detecting oxygen deficiency in the area being monitored and for managing the oxygen concentration. Oxygen deficiency refers to a decrease in the proportion of oxygen in the air, which can have serious consequences for human health and survival. Because oxygen aids combustion, an increase in oxygen concentration in the monitored area increases the risk of fire. Specifically, as the oxygen concentration increases, the ignition temperature of substances decreases, making them easier to ignite. Therefore, the oxygen concentration measured by the oxygen sensor 32 is taken into account when determining whether or not there are signs of a fire in the area monitored by the fire warning determination unit 14. Furthermore, the predictive learning model incorporates the oxygen concentration measurement data of the monitored area, measured by the oxygen sensor 32, into a combination of measurement data for a predetermined indicator, and learns in advance the correspondence between these combinations of measurement data and the presence or absence of fire signs in the monitored area. Furthermore, as the oxygen concentration increases, the flame temperature rises and combustion speeds up, expanding the combustion area and increasing the possibility of explosion, thus increasing the danger. Therefore, the oxygen concentration measured by the oxygen sensor 32 is taken into account when the danger determination unit 15 determines the danger level of the fire precursor in the monitored area. Furthermore, the risk learning model incorporates the oxygen concentration measurement data of the monitored area, measured by the oxygen sensor 32, into a combination of measurement data for a predetermined indicator, and learns in advance the correspondence between these combinations of measurement data and the risk level of fire precursors in the monitored area. The output terminal of the oxygen sensor 32 is connected to the CPU 20e of the environmental measurement terminal 20, which will be described later, and the measurement data from the oxygen sensor 32 is input to the CPU 20e.

[0056] Furthermore, the measurement data from the multiple types of sensors described above is expected to be affected by various types of noise. For this reason, the multiple types of sensors may be driven by a power supply that has been boosted and stabilized by a DC-DC converter. In addition, the measurement data from the multiple types of sensors may be passed through an interference suppression filter and an outlier removal filter immediately before being input to the CPU 20e of the environmental measurement terminal 20 described later. Alternatively, noise contained in the measurement data from the multiple types of sensors may be separated from the measurement data and trained in a deep learning algorithm, and the trained deep learning algorithm may be used to remove the noise contained in the measurement data from the multiple types of sensors.

[0057] (Regarding the infrastructure configuration, including the fire prediction detection system 10) Next, with reference to Figure 2, the infrastructure configuration including the fire prediction detection system 10 will be described. Figure 2 is an example of an infrastructure configuration diagram including the fire prediction detection system 10 according to this embodiment.

[0058] As shown in Figure 2, the fire prediction detection system 10 and the environmental measurement terminals 20 (environmental measurement terminal A20a, environmental measurement terminal B20b) are connected to the Internet 11. The fire prediction detection system 10 and the environmental measurement terminals 20 (environmental measurement terminal A20a, environmental measurement terminal B20b) can communicate bidirectionally via the Internet 11. Alternatively, the fire prediction detection system 10 and the environmental measurement terminals 20 (environmental measurement terminal A20a, environmental measurement terminal B20b) may be directly connected by wired or wireless communication without going through the Internet 11. Wired communication refers to, for example, communication via a wired LAN (Local Area Network). Wireless communication refers to, for example, short-range wireless communication such as wireless LAN or Bluetooth (registered trademark).

[0059] For convenience, Figure 2 shows two environmental measurement terminals 20 (environmental measurement terminal A20a, environmental measurement terminal B20b), but it is not limited to this, and one environmental measurement terminal 20, or three or more environmental measurement terminals 20, may be connected to and used with the fire prediction detection system 10. Furthermore, although Figure 2 shows one fire precursor detection system 10 for convenience, it is not limited to this, and two or more fire precursor detection systems 10 may be connected to one or more environmental measurement terminals 20 and used.

[0060] (Regarding the installation method of the environmental measurement terminal 20) Next, with reference to Figure 3, the installation configuration of the environmental measurement terminal 20 will be described. Figure 3 is a diagram illustrating an example of the installation configuration of the environmental measurement terminal 20 according to this embodiment. The environmental measurement terminal 20 may be installed suspended from the ceiling 33 in the area to be monitored, mounted on a stand 34 erected on the floor, or fixed to a worker's helmet 37. Furthermore, the environmental measurement terminal 20 may be mounted on the ceiling, wall, or floor, or it may be attached to equipment capable of moving in the air or on the ground, such as drones, robots, and nanobots, allowing it to be used while moving.

[0061] Multiple types of sensors may be mounted on an unmanned aerial vehicle or robot, and the unmanned aerial vehicle or robot may fly or drive through the area to be monitored by remote control or autonomous control. As shown in Figure 3, an environmental measurement terminal 20 is mounted on an unmanned aerial vehicle (UAV) 35 such as a drone, and multiple types of sensors implemented on the environmental measurement terminal 20 measure the environment of the area to be monitored. While the UAV 35 is circling or hovering within the area to be monitored, multiple types of sensors implemented on the environmental measurement terminal 20 measure the environment of the area to be monitored. Unmanned aerial vehicles (UAVs) are also called quadrotors or multicopters, in addition to drones. The unmanned aerial vehicle 35 may be remotely controlled by radio waves, or it may fly using an autonomous control automatic flight function. Remote control by radio waves means that a person operating the unmanned aerial vehicle 35 uses a controller (not shown) to control the unmanned aerial vehicle 35 from a remote location via radio communication. Autonomous control automatic flight function means that the unmanned aerial vehicle 35 perceives its surrounding environment on its own and flies according to a pre-set route or instructions, and flies according to a flight plan that has been created and implemented in advance. Measurement data from multiple sensors implemented in the environmental measurement terminal 20 installed on the unmanned aerial vehicle 35 is transmitted to the fire prediction detection system 10 via wireless communication such as Wi-Fi (registered trademark) or Bluetooth.

[0062] As shown in Figure 3, an environmental measurement terminal 20 is installed on a four-legged robot 36 as an example of a robot, and multiple types of sensors implemented on the environmental measurement terminal 20 measure the environment of the area to be monitored. While the four-legged robot 36 is stopped or moving at a low speed within the area to be monitored, multiple types of sensors implemented on the environmental measurement terminal 20 measure the environment of the area to be monitored. The four-legged robot 36 may be remotely controlled wirelessly, similar to the unmanned aerial vehicle 35, or it may move using an autonomous driving function controlled by an autonomous system. Measurement data from multiple sensors implemented in the environmental measurement terminal 20 installed on the four-legged robot 36 is transmitted to the fire prediction detection system 10 via wireless communication such as Wi-Fi (registered trademark) or Bluetooth, similar to the unmanned aerial vehicle 35. Furthermore, the robot is not limited to the four-legged robot 36; it may also be a two-legged robot, or a robot with three or more legs, or a robot equipped with tires or wheels instead of legs, or a robot that does not have legs, tires, or wheels and moves using other means of locomotion (e.g., rails, ropes, etc.).

[0063] (Regarding the hardware configuration of the fire prediction detection system 10) Next, the hardware configuration of the fire prediction detection system 10 will be described with reference to Figure 4. Figure 4 is a block diagram illustrating an example of the hardware configuration of the fire prediction detection system according to this embodiment.

[0064] The fire prediction detection system 10 is a so-called computer and is equipped with a communication interface 10a, ROM (Read Only Memory) 10b, RAM (Random Access Memory) 10c, storage unit 10d, CPU (Central Processing Unit) 10e, and input / output interface 10f, etc.

[0065] The communication interface 10a has the function of sending and receiving data handled by the fire prediction detection system 10 to and from other devices via the internet 11. Other devices include the environmental measurement terminal 20 and other fire prediction detection systems 10. When the fire prediction detection system 10 connects to other devices via Bluetooth, the communication interface 10a becomes a Bluetooth interface. When the fire prediction detection system 10 connects to other devices via wireless LAN, the communication interface 10a becomes a wireless LAN interface.

[0066] The storage unit 10d can be used as a storage device for the fire prediction detection system 10 and can be composed of, for example, a hard disk drive, a solid state drive, and flash memory. Furthermore, the storage unit 10d can also be configured using cloud storage. The storage unit 10d of the fire prediction detection system 10 may be configured as a database. By configuring the storage unit 10d as a database, a large amount of data can be managed efficiently, providing excellent searchability and accessibility, and allowing necessary information to be quickly retrieved.

[0067] Furthermore, the storage unit 10d of the fire precursor detection system 10 stores the fire precursor detection program described later, the OS (Operating System) necessary for the operation of the fire precursor detection system 10, various other applications, and various data used by those applications.

[0068] The fire prediction detection system 10 stores the fire prediction detection program, described later, in ROM 10b or storage unit 10d, and loads the fire prediction detection program into the main memory, which is composed of RAM 10c or the like. The CPU 10e accesses the main memory containing the fire prediction detection program and executes the fire prediction detection program. The fire prediction detection program may be stored in the ROM 20b or storage unit 20d of the environmental measurement terminal 20 and executed by the CPU 20e.

[0069] The input / output interface 10f transmits and receives data to and from external devices of the fire prediction detection system 10. External devices refer to input devices 10g and output devices 10h that input and output data to and from the fire prediction detection system 10. Input devices 10g include, for example, a keyboard and mouse, while output devices 10h include, for example, a monitor, printer, and speaker.

[0070] (Regarding the hardware configuration of the environmental measurement terminal 20) Next, the hardware configuration of the environmental measurement terminal 20 will be described with reference to Figure 5. Figure 5 is a block diagram illustrating an example of the hardware configuration of the environmental measurement terminal 20.

[0071] The environmental measurement terminal 20 includes a communication interface 20a, ROM 20b, RAM 20c, storage unit 20d, CPU 20e, input / output interface 20f, and multiple types of sensors. The multiple types of sensors are directly connected to the CPU 20e. The multiple types of sensors are an odor sensor 21, a two-dimensional thermopile radiation temperature sensor 22, a pressure sensor 23, a temperature and humidity sensor 24, an air pollution sensor 25, a smoke sensor 26, a carbon monoxide sensor 27, a carbon dioxide sensor 28, an illuminance sensor 29, a noise sensor 30, and a night vision camera 31. Of the multiple types of sensors, the measurement data from the two-dimensional thermopile radiation temperature sensor 22, the pressure sensor 23, the temperature and humidity sensor 24, the night vision camera 31, and the oxygen sensor 32 are digitally output via I2C communication in a format that responds to data requests from the CPU 20e of the environmental measurement terminal 20, as described later. In addition, the measurement data from the carbon dioxide sensor 28 is input to the CPU 20e of the environmental measurement terminal 20 as a PWM signal. Furthermore, the measurement data from the air pollution sensor 25, carbon monoxide sensor 27, illuminance sensor 29, and noise sensor 30 are converted from analog signals to digital signals by a 12-bit ADC (Analog to Digital Converter) after passing through a differential amplifier circuit, and then input to the CPU 20e of the environmental measurement terminal 20. Furthermore, the measurement data from the smoke sensor 26 is converted from an analog signal to a digital signal by a 12-bit ADC after the peak value has been held for a certain period of time, and then input to the CPU 20e of the environmental measurement terminal 20. The communication methods for the odor sensor 21, the two-dimensional thermopile radiation temperature sensor 22, the pressure sensor 23, the temperature and humidity sensor 24, the air pollution sensor 25, the smoke sensor 26, the carbon monoxide sensor 27, the carbon dioxide sensor 28, the illuminance sensor 29, the noise sensor 30, the night vision camera 31, and the oxygen sensor 32 may be changed to other communication methods in accordance with changes or improvements to the specifications of these sensors.

[0072] The CPU 20e controls the odor sensor 21, the two-dimensional thermopile radiation temperature sensor 22, the pressure sensor 23, the temperature and humidity sensor 24, the air pollution sensor 25, the smoke sensor 26, the carbon monoxide sensor 27, the carbon dioxide sensor 28, the illuminance sensor 29, the noise sensor 30, the night vision camera 31, and the oxygen sensor 32 to acquire measurement data from these multiple types of sensors and transmit it to the fire prediction detection system 10.

[0073] The communication interface 20a has the function of sending and receiving data handled by the environmental measurement terminal 20 to and from other devices via the internet 11. The other devices are mainly the fire prediction detection system 10. When the environmental measurement terminal 20 connects to other devices via Bluetooth, the communication interface 10a becomes a Bluetooth interface. When the environmental measurement terminal 20 connects to other devices via wireless LAN, the communication interface 20a becomes a wireless LAN interface.

[0074] The storage unit 10d can be used as a storage device for the environmental measurement terminal 20 and consists of, for example, a hard disk drive, a solid state drive, and flash memory. The storage unit 20d of the environmental measurement terminal 20 stores firmware and various data necessary for the operation of the environmental measurement terminal 20.

[0075] The input / output interface 20f transmits and receives data to and from external devices of the environmental measurement terminal 20. External devices refer to input and output devices (not shown) that input and output data to and from the environmental measurement terminal 20.

[0076] Furthermore, the CPU 20e of the environmental measurement terminal 20 may be a PIC microcontroller (Peripheral Interface Controller microcomputer). While PIC microcontrollers are not suitable for general-purpose use like personal computers, they are well-suited for controlling specific applications such as peripheral devices like sensors. Alternatively, the CPU 20e may be a microcontroller such as M5stack, Raspberry Pi, Arduino, or ESP, or various semiconductors equipped with device control functions may be used.

[0077] (Regarding the circuit configuration of the environmental measurement terminal 20) Next, the circuit configuration of the environmental measurement terminal 20 will be explained with reference to Figures 6a to 6d. Figures 6a to 6d are diagrams illustrating an example of the circuit configuration of the environmental measurement terminal 20.

[0078] Figure 6a shows the configuration of the analog circuits of the environmental measurement terminal 20, mainly relating to the air pollution sensor 25, smoke sensor 26, and carbon dioxide sensor 28. Figure 6b shows the configuration of the analog circuit, mainly related to the carbon monoxide sensor 27, within the circuit configuration of the environmental measurement terminal 20. Figure 6c shows the configuration of the digital circuitry in the environmental measurement terminal 20, mainly relating to the connections of the terminals of the PIC microcontroller used as the CPU 20e. Figure 6d shows the configuration of the digital circuits of the environmental measurement terminal 20, mainly relating to the two-dimensional thermopile radiation temperature sensor 22, the atmospheric pressure sensor 23, the temperature and humidity sensor 24, and the illuminance sensor 29.

[0079] (Regarding the functional configuration of the fire prediction detection system 10) The functional configuration of the fire prediction detection system 10 will be described with reference to Figure 7. Figure 7 is a block diagram illustrating an example of the functional configuration of the fire prediction detection system 10.

[0080] The fire precursor detection system 10, by executing the fire precursor detection program described later, has a CPU 10e equipped with functional units such as a reception unit 12, a measurement data acquisition unit 13, a precursor presence / absence determination unit 14, a risk level determination unit 15, a notification unit 16, an abnormal value extraction unit 17, an abnormal value precursor determination unit 18, and a threshold reset unit 19.

[0081] The reception unit 12 receives the type of area to be monitored. The reception unit 12 receives the type of area to be monitored, which is entered by the user of the fire prediction detection system 10, or the owner or supervisor of the area to be monitored. The types of areas subject to monitoring include, for example, buildings, logistics warehouses, petrochemical plants, machinery factories, metal factories, food factories, pharmaceutical factories, paper mills, and construction sites of semiconductor factories, which are classified according to the types of materials they contain. In addition, other types of areas subject to monitoring include residences, nursing homes, childcare facilities, power plants, forests, dams, fishing ports, train stations, and airports. Furthermore, areas that do not involve buildings, such as fields, parks, grounds, and parking lots, can also be subject to monitoring, whether they are under construction or have been completed. Moreover, the interiors of moving objects such as buses, trains, ships, airplanes, submarines, and spacecraft can also be subject to monitoring.

[0082] The measurement data acquisition unit 13 acquires measurement data for each of several indicators of the environment of the area to be monitored, which are acquired by multiple types of sensors. The measurement data acquisition unit 13 may directly acquire measurement data from multiple types of sensors, or it may acquire measurement data from multiple types of sensors that have been temporarily stored on a server on the Internet 11 from that server. This server refers to, for example, an IoT server, and can be an on-premise server or a cloud server. Multiple indicators refer to factors such as odor, atmospheric pressure, temperature, humidity, air pollution, smoke concentration, carbon monoxide concentration, carbon dioxide concentration, illuminance, and noise level in the area being monitored. The measurement data acquisition unit 13 stores the acquired measurement data from multiple types of sensors in the storage unit 10d.

[0083] The measurement data acquisition unit 13 may have one or more thresholds set for the measurement data of each of the multiple indicators, and may acquire measurement data that exceeds the threshold for each of the multiple types of sensors.

[0084] In the environmental measurement terminal 20, one or more thresholds are set for each of the measurement data from multiple types of sensors, and the CPU 20e transmits measurement data that it determines has exceeded a threshold to the fire prediction detection system 10. The measurement data acquisition unit 13 acquires the measurement data transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10. Since the measurement data from multiple types of sensors is expected to be enormous in volume, transmitting only the measurement data deemed useful to the fire prediction detection system 10 can reduce the amount of measurement data transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10.

[0085] A threshold is set to define the lower or upper limit of the measurement data to be transmitted to the fire prediction detection system 10. For example, if a lower threshold is set, measurement data equal to or greater than the set threshold is transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10. If an upper threshold is set, measurement data equal to or less than the set threshold is transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10.

[0086] Multiple thresholds may be set for a single type of measurement data. These thresholds define the lower and upper limits of the measurement data to be transmitted to the fire prediction detection system 10. If only measurement data for one bandwidth is transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10, two thresholds are set. For example, if a lower threshold and an upper threshold are set, measurement data ranging from the set lower threshold to the upper threshold is transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10. If measurement data for two bandwidths is transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10, a total of four thresholds are defined for the measurement data of one indicator.

[0087] The fire precursor determination unit 14 determines whether or not there are signs of a fire in the monitored area based on a combination of measurement data of a predetermined indicator selected from the measurement data acquired by the measurement data acquisition unit 13. The indicator detection unit 14 refers to a table that defines the correspondence between the type of area to be monitored and the combination of measurement data for environmental indicators of that area, and determines the combination of measurement data for indicators based on the type of area to be monitored received by the reception unit 12.

[0088] Refer to Figure 8 to explain the table. Figure 8 is an example of a table that defines the correspondence between the type of area to be monitored and the combination of measurement data for multiple environmental indicators of that area. The table specifies the combination of measurement data for indicators that the fire precursor determination unit 14 uses to determine whether or not there are signs of fire in the monitored area, for each type of area being monitored. The combinations specified in the table are merely examples and are not limited to them; they can be changed.

[0089] If the area to be monitored is a construction site of a building, the fire warning determination unit 14 determines whether or not there are signs of a fire in the area to be monitored based on a combination of measurement data for odor, temperature, carbon dioxide concentration, and oxygen concentration. If the area to be monitored is a construction site for a logistics warehouse, the fire warning determination unit 14 determines whether or not there are signs of a fire in the monitored area based on a combination of measurement data for odor, temperature, carbon dioxide concentration, and oxygen concentration.

[0090] If the area to be monitored is a construction site of a petrochemical plant, the fire warning determination unit 14 determines whether or not there are signs of a fire in the monitored area based on a combination of measurement data for odor, temperature, air pollution, carbon monoxide concentration, and oxygen concentration. If the area to be monitored is a construction site of a machine shop, the fire warning determination unit 14 determines whether or not there are signs of a fire in the area to be monitored based on a combination of measurement data for odor, temperature, smoke concentration, carbon dioxide concentration, and oxygen concentration.

[0091] If the area to be monitored is a construction site of a metal factory, the fire warning determination unit 14 determines whether or not there are signs of a fire in the monitored area based on a combination of measurement data for odor, temperature, smoke concentration, carbon dioxide concentration, and oxygen concentration. If the area to be monitored is a construction site of a food factory, the fire warning determination unit 14 determines whether or not there are signs of a fire in the area to be monitored based on a combination of measurement data for odor, humidity, carbon monoxide concentration, and oxygen concentration.

[0092] If the area to be monitored is a construction site of a pharmaceutical factory, the fire warning determination unit 14 determines whether or not there are signs of a fire in the area to be monitored based on a combination of measurement data for odor, atmospheric pressure, carbon monoxide concentration, carbon dioxide concentration, and oxygen concentration. If the area to be monitored is a construction site of a paper mill, the fire warning determination unit 14 determines whether or not there are signs of a fire in the monitored area based on a combination of measurement data for odor, atmospheric pressure, temperature, carbon dioxide concentration, and oxygen concentration.

[0093] If the area to be monitored is a construction site of a semiconductor factory, the fire warning determination unit 14 determines whether or not there are signs of a fire in the area to be monitored based on a combination of measurement data for odor, temperature, carbon monoxide concentration, carbon dioxide concentration, and oxygen concentration.

[0094] Referring to Figure 9, the determination of whether or not there are signs of fire in the area monitored by the fire precursor determination unit 14 will be explained. Figure 9 is a diagram illustrating an example of a combination of measurement data for multiple environmental indicators in the area monitored.

[0095] In the example shown in Figure 9, the fire precursor determination unit 14 determines whether or not there are signs of a fire in the monitored area based on a combination of measurement data of carbon monoxide concentration, temperature, and carbon dioxide concentration, which are multiple indicators of the environment in the monitored area. The graph lines 41 for carbon monoxide concentration, 42 for temperature, and 43 for carbon dioxide concentration each show changes over time. The fire warning determination unit 14 determines whether or not there are signs of a fire in the monitored area by looking at the patterns of change over time in the measured data of carbon monoxide concentration, temperature, and carbon dioxide concentration from a comprehensive perspective.

[0096] Therefore, without setting thresholds for each measurement data, the presence or absence of signs of fire in the monitored area is determined based on the correlation of measurement data for carbon monoxide concentration, temperature, and carbon dioxide concentration. Correlation refers to a relationship in which two or more things are closely related, and when one changes, the others also change.

[0097] If the warning unit 14 determines that there are signs of a fire in the area being monitored, the notification unit 16 notifies the monitor of that area that there are signs of a fire. The notification unit 16 notifies the monitor of the area via email, push notification, phone call to a designated telephone number, and alarm sound. Push notification refers to notification using the functions of an application on a smartphone or similar device.

[0098] The fire precursor determination unit 14 uses a fire precursor learning model that has been pre-learned to determine the correspondence between combinations of measurement data for predetermined indicators of the environment of the monitored area and the presence or absence of fire precursors in that area. By inputting combinations of measurement data for predetermined indicators selected from the measurement data acquired by the measurement data acquisition unit 13 into the fire precursor learning model, it determines whether or not there are fire precursors in the monitored area.

[0099] Supervised training data for predictive learning models is created as follows: Training data annotated with "signs of fire" are generated as a set of datasets by adding metadata indicating that there were signs of fire in a particular area to a combination of measurement data for environmental indicators of the monitored area that show signs of fire.

[0100] Training data annotated as "no signs of fire" is generated as a set of datasets by adding metadata to combinations of measurement data for environmental indicators of a monitored area that show no signs of fire, indicating that there were no signs of fire in that area. By creating large amounts of training data annotated with "signs of fire present" and training data annotated with "no signs of fire present," and then training the predictive learning model with this large amount of training data, the predictive accuracy of the predictive learning model can be improved. Furthermore, training data for the predictive learning model needs to be created for each type of region to be monitored.

[0101] Furthermore, the training data for the predictive learning model is newly created using measurement data of the environment of the monitored area acquired through the operation of the fire prediction detection system 10. Each time a predetermined amount of such measurement data is accumulated, new training data is created, and by applying machine learning to the predictive learning model, the amount of learning increases, and the prediction accuracy of the predictive learning model can be improved.

[0102] When the Precursor Detection Unit 14 determines that there are signs of a fire in the area being monitored, the Risk Determination Unit 15 determines the risk level of the fire signs based on a combination of measurement data of multiple predetermined environmental indicators for that area that formed the basis of the Precursor Detection Unit 14's determination. If the Precursor Detection Unit 14 determines that there are signs of a fire in the area being monitored, the notification unit 16 notifies the monitor of the area that there are signs of a fire and the degree of danger of the signs as determined by the Danger Determination Unit 15.

[0103] The degree of risk refers to the fire damage that is expected to occur based on the detected signs of a fire. Fire damage refers to the combined damage from burning and firefighting. Burning damage refers to damage caused by flames and high temperatures, such as burning, breaking, or alteration of materials, while firefighting damage refers to damage such as water damage incurred during firefighting, and damage or soiling that occurs during the firefighting process. However, the degree of risk is not limited to these, and other types of damage may also be considered. For example, estimated costs for firefighting may be included, or human casualties, which involve loss of life, may also be considered. The risk level (grade) classification may be divided into multiple categories, for example, into 2, 3, or 5 categories. In this embodiment, the risk level is divided into three stages, from most to least damaging: Level 1, Level 2, and Level 3.

[0104] The risk assessment unit 15 uses a risk assessment learning model that has been pre-learned to determine the correspondence between a combination of measurement data for a predetermined indicator of the environment of the area to be monitored and the risk level of fire precursors in that area. By inputting the combination of measurement data for the predetermined indicator that formed the basis for the fire precursor determination unit 14's determination that there are fire precursors in that area into the risk assessment learning model, the unit determines the risk level of fire precursors in that area.

[0105] The supervised training data for the risk assessment learning model is created as follows: Training data annotated with a risk level of 1 is generated as a set of datasets by adding metadata to combinations of measurement data for environmental indicators of a monitored area where the risk level is 1, indicating that the risk level of fire precursors in that area was 1. Training data annotated with a risk level of 2 is generated as a set of datasets by adding metadata to combinations of measurement data for environmental indicators of monitored areas where the risk level is 2, indicating that the risk level of fire precursors in that area was 2. Training data annotated with a risk level of 3 is generated as a set of datasets by adding metadata to combinations of measurement data for environmental indicators of a monitored area that are in a risk level of 3, indicating that the risk level of fire precursors in that area was level 3.

[0106] By creating numerous training datasets annotated with risk levels 1, 2, and 3, and then using these large datasets to train the risk assessment model, the prediction accuracy of the risk assessment model can be improved. Furthermore, the training data for the risk assessment learning model, like the training data for the predictive learning model, needs to be created for each type of area to be monitored.

[0107] Furthermore, the training data for the risk assessment learning model is newly created using measurement data of the environment of the monitored area acquired through the operation of the fire prediction detection system 10, similar to the training data for the predictive learning model. By creating new training data each time a predetermined amount of such measurement data is accumulated, machine learning is applied to the risk assessment learning model, increasing the amount of learning and improving the prediction accuracy of the risk assessment learning model.

[0108] The anomaly extraction unit 17 extracts outliers from the measurement data of multiple environmental indicators of the monitored area, which are included in the measurement data, by performing unsupervised clustering of the measurement data acquired by the measurement data acquisition unit 13, as anomalies in the measurement data of each of the multiple indicators. The anomaly detection unit 17 uses the measurement data acquired by the measurement data acquisition unit 13 as unsupervised training data and performs clustering on this unsupervised training data. An outlier is a value that is extremely small or extremely large compared to others. The anomaly detection unit 17 uses algorithms such as K-means or hierarchical clustering to create groups (clusters) within the measurement data of multiple environmental indicators of the monitored area, acquired by the measurement data acquisition unit 13, based on the internal similarity of these measurement data. It also extracts outliers that do not belong to any of these groups as anomalies. This clustering is performed on each of the measurement data of multiple environmental indicators of the monitored area, and anomalies are extracted for each set of measurement data for each indicator. Clustering is a type of unsupervised learning that uses unlabeled (annotated) data to calculate the similarity (distance) between data points and classify the data into several groups (clusters). Clustering calculates the distance between all data points. Clustering groups data, making it easier to find patterns or structures within the data, and also making it easier to identify outliers that don't fit into the groups. Therefore, clustering is useful for extracting abnormal values ​​from data. Clustering methods include agglomerative and divisible algorithms. Agglomerative methods initially treat each data point as a single cluster and gradually merge these clusters. Divisive methods initially treat all data points as a single cluster and then divide them until a predetermined number of clusters (groups) are formed. The anomaly detection unit 17 uses clustering, a type of unsupervised learning, to extract anomalies from the vast amount of measurement data acquired by the measurement data acquisition unit 13 without the time-consuming task of creating supervised learning data. Furthermore, the unsupervised learning performed by the fire prediction detection system 10 is not limited to clustering, and other forms of unsupervised learning may be performed. For example, the fire prediction detection system 10 may use unsupervised learning to analyze the structure and characteristics of the acquired data itself, to discover frequently occurring patterns in the acquired data, to group the acquired data, or to simplify the acquired data. Furthermore, the fire prediction detection system 10 may use, for example, a generative adversarial network, which is a type of unsupervised learning, to generate non-existent data or transform data according to the characteristics of the acquired data by learning features from the acquired data. Furthermore, the fire prediction detection system 10 may use, for example, association analysis, a type of unsupervised learning, to discover relationships between acquired data. Furthermore, the fire prediction detection system 10 may also use, for example, principal component analysis (PCA), a type of unsupervised learning, to summarize (reduce dimensionality) the multiple explanatory variables in the acquired data, find new explanatory variables, and make the data easier to understand by using these new explanatory variables as the principal components of the data.

[0109] The abnormal value prediction unit 18 determines whether or not there are signs of a fire in the area based on the abnormal values ​​extracted by the abnormal value extraction unit 17. The abnormal value prediction unit 18 may determine that there is a fire precursor in the monitored area if the abnormal value extracted by the abnormal value extraction unit 17 exceeds a preset threshold. This threshold is set for each of several indicators of the environment of the monitored area, and the abnormal value prediction unit 18 may determine that there is a fire precursor in the monitored area if an abnormal value exceeds the preset threshold in any of the measurement data of each of the several indicators, or it may determine that there is a fire precursor in the monitored area if an abnormal value exceeds the threshold in a predetermined number of indicators. Alternatively, the abnormal value prediction unit 18 may use a learning model that has been pre-learned to determine whether or not there are signs of fire in the monitored area, based on the abnormal values ​​extracted by the abnormal value extraction unit 17, by learning the correspondence between the abnormal values ​​extracted by the abnormal value extraction unit 17 and the presence or absence of signs of fire in the monitored area. The abnormal value prediction unit 18 determines whether there are signs of a fire in the area based on the abnormal values ​​extracted by the abnormal value extraction unit 17, and the sign presence / absence determination unit 14 determines whether there are signs of a fire in the monitored area based on a combination of measurement data of a predetermined indicator selected from the measurement data acquired by the measurement data acquisition unit 13. Therefore, the abnormal value prediction unit 18 and the sign presence / absence determination unit 14 can determine whether there are signs of a fire in the monitored area based on two different criteria (one being abnormal values, and the other being a combination of measurement data).

[0110] The threshold resetting unit 19 resets the thresholds for each of the measurement data of multiple indicators based on the abnormal values ​​extracted by the abnormal value extraction unit 17. The threshold resetting unit 19 resets the threshold used by the measurement data acquisition unit 13 when acquiring measurement data from each of the multiple types of sensors, based on the abnormal values ​​extracted by the abnormal value extraction unit 17 for each of the multiple indicators of the environment of the area to be monitored. The threshold resetting unit 19 may, for example, set the threshold to 70% of the abnormal value, but it is not limited to 70% and may be any value between 1% and 99%. The measurement data acquisition unit 13 acquires measurement data from each of several types of sensors that exceeds the threshold reset by the threshold reset unit 19 for each of the multiple indicators. The threshold resetting unit 19 may set multiple thresholds for a single type of measurement data. It resets the lower and upper limits of the measurement data to be transmitted to the fire prediction detection system 10. If only measurement data for one bandwidth is transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10, two thresholds are reset. For example, if the lower and upper thresholds are reset, measurement data with values ​​between the reset lower and upper thresholds is transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10. If measurement data for two bandwidths is transmitted from the environmental measurement terminal 20 to the fire prediction detection system 10, a total of four thresholds are reset for the measurement data of one indicator. According to the threshold resetting unit 19, the threshold used by the measurement data acquisition unit 13 can be reset based on the abnormal values ​​extracted by the abnormal value extraction unit 17.

[0111] (Regarding the experimental results of the fire prediction detection system 10) Next, the experimental results of the fire prediction detection system 10 will be explained with reference to Figure 10. Figure 10 is a diagram illustrating an example of the experimental results of the fire prediction detection system 10.

[0112] Figure 10 shows the temperature distribution measurement results of the 2D thermopile radiation temperature sensor 22, divided into a total of 64 sections arranged in an 8x8 grid on the 2D plane of the area to be monitored. Each square in Figure 10 corresponds to one section of the 2D plane, and the number in each square indicates the temperature (°C) of that section. As shown in Figure 10, the temperature distribution measurement results from the two-dimensional thermopile radiation temperature sensor 22 indicate the area with the highest temperature, suggesting that the area may contain a potential source of fire. Furthermore, by changing the color of the cells showing the temperature distribution measurement results according to the temperature of the area, the temperature distribution becomes easier to recognize visually.

[0113] (Regarding fire precursor detection methods and fire precursor detection programs) Next, with reference to Figure 11, a fire precursor detection program according to one embodiment of the present invention will be described along with a fire precursor detection method. Figure 11 is an example of a flowchart of the fire precursor detection program according to this embodiment.

[0114] The fire precursor detection method is executed by the CPU 10e of the fire precursor detection system 10 based on the fire precursor detection program. As shown in Figure 11, the fire precursor detection program includes a reception step S12, a measurement data acquisition step S13, a precursor presence / absence determination step S14, a risk level determination step S15, and a notification step S16.

[0115] The fire precursor detection program implements various functions for the CPU 10e of the fire precursor detection system 10, including reception, measurement data acquisition, prediction of the presence or absence of a precursor, risk level determination, and notification. These functions are executed in the order shown in the flowchart of Figure 11, but the order can be changed as appropriate. Since each function overlaps with the description of the various functions of the fire precursor detection system 10 mentioned above, a detailed explanation is omitted.

[0116] The reception function accepts the type of area to be monitored (S12: reception step). The reception function accepts the type of area to be monitored, entered by the user of the fire prediction detection system 10, or the owner or supervisor of the area to be monitored.

[0117] The measurement data acquisition function acquires measurement data for each of multiple indicators of the environment of the monitored area, which are acquired by multiple types of sensors (S13: Measurement data acquisition step). The measurement data acquisition function acquires measurement data from multiple types of sensors from the environmental measurement terminal 20.

[0118] The fire warning determination function determines whether or not there are signs of a fire in the monitored area based on a combination of measurement data for predetermined indicators selected from the measurement data acquired by the measurement data acquisition function (S14: Fire Warning Determination Step).

[0119] The risk assessment function determines the risk level of the fire precursor when the indicator presence / absence assessment function determines that there is a fire precursor in the monitored area, based on a combination of measurement data of multiple predetermined environmental indicators for that area that formed the basis of the determination in the indicator presence / absence assessment function (S15: Risk Assessment Step).

[0120] The notification function, when the fire warning detection function determines that there are signs of a fire in the area being monitored, notifies the monitor of that area of ​​the presence of signs of a fire (S16: notification step).

[0121] (Regarding fire precursor detection methods and fire precursor detection programs according to other embodiments) Referring to Figure 12, a fire precursor detection program according to another embodiment will be described along with a fire precursor detection method. Figure 12 is an example of a flowchart for a fire precursor detection program according to another embodiment. The flowchart of the fire precursor detection program according to another embodiment shown in Figure 12 differs from the flowchart of the fire precursor detection program shown in Figure 11 in that it includes an anomaly value extraction step S17, an anomaly value prediction step S18, and a threshold reset step S19. The fire precursor detection method according to the other embodiment is executed by the CPU 10e of the fire precursor detection system 10 based on the fire precursor detection program according to the other embodiment shown in Figure 12. The fire precursor detection program according to another embodiment shown in Figure 12 includes a reception step S12, a measurement data acquisition step S13, a precursor presence determination step S14, a risk level determination step S15, a notification step S16, an abnormal value extraction step S17, an abnormal value precursor determination step S18, and a threshold reset step S19.

[0122] The fire precursor detection program according to another embodiment shown in Figure 12 provides the CPU 10e of the fire precursor detection system 10 with functions such as reception, measurement data acquisition, prediction of precursors, risk level determination, notification, abnormal value extraction, abnormal value prediction, and threshold resetting. These functions are executed in the order shown in the flowchart of Figure 12, but the order can be changed as appropriate. Below, we will describe the fire precursor detection method and fire precursor detection program according to another embodiment shown in Figure 12, focusing only on the differences between them and the fire precursor detection method and fire precursor detection program shown in Figure 11. Furthermore, since each function overlaps with the descriptions of the various functional parts of the fire prediction detection system 10 mentioned above, detailed explanations will be omitted.

[0123] The outlier extraction function extracts outliers from the measurement data of multiple environmental indicators within the measurement data area by performing unsupervised clustering on the measurement data acquired by the measurement data acquisition function, treating each of the measurement data of the multiple indicators as anomalies (S17: Outlier Extraction Step).

[0124] The abnormal value prediction function determines whether or not there are signs of a fire in the area based on the abnormal values ​​(S18: abnormal value prediction step).

[0125] The threshold reset function resets the thresholds for each measurement data of multiple indicators based on outliers (S19: Threshold Reset Step).

[0126] According to the fire precursor detection system 10 of the above-described embodiment, multiple sensors, including an odor sensor 21, a two-dimensional thermopile radiation temperature sensor 22, a pressure sensor 23, a temperature and humidity sensor 24, an air pollution sensor 25, a smoke sensor 26, a carbon monoxide sensor 27, a carbon dioxide sensor 28, an illuminance sensor 29, a noise sensor 30, and a camera with night vision capabilities 31, are used to monitor the area to be monitored using multiple environmental indicators, thereby enabling early detection of fire precursors.

[0127] Furthermore, the fire prediction detection system 10 determines the presence or absence of fire signs in a monitored area using a machine learning-developed prediction learning model based on a combination of measurement data of multiple environmental indicators in that area. Compared to a system that sets a threshold for the measurement data of environmental indicators and determines the presence or absence of fire signs based on whether the measurement data exceeds that threshold, the system determines the presence or absence of fire signs by considering the correlation of multiple environmental indicators without fixing a threshold for the measurement data, thus enabling earlier and more realistic determinations.

[0128] Furthermore, according to the fire precursor detection system 10, a combination of measurement data for environmental indicators of the area to be monitored is determined according to the type of area being monitored. Therefore, it is possible to determine whether or not there are signs of fire in accordance with the characteristics of the area, and thus the presence or absence of signs of fire can be determined more quickly.

[0129] Furthermore, the fire prediction detection system 10 uses a machine learning-based risk assessment model to determine the risk level of an area where signs of fire are detected, thus enabling a more accurate determination of the risk level.

[0130] Furthermore, the fire precursor detection system 10 can determine whether or not there are signs of a fire in the area being monitored by taking into account the oxygen concentration in that area.

[0131] Furthermore, the fire prediction detection system 10 can determine the degree of fire risk by taking into account the oxygen concentration in the area being monitored.

[0132] Furthermore, according to the fire prediction detection system 10, since the environmental measurement terminal 20 is mounted on the unmanned aerial vehicle 35 or robot 36, the user can move the environmental measurement terminal 20 to any area of ​​their choice and determine whether or not there are any signs of fire in that area.

[0133] Furthermore, according to the fire prediction detection system 10, by using clustering, a type of unsupervised learning, it is possible to extract abnormal values ​​from the vast amount of measurement data acquired by the measurement data acquisition unit 13 without having to go through the time-consuming process of creating supervised learning data.

[0134] Furthermore, according to the fire precursor detection system 10, the abnormal value precursor determination unit 18 and the precursor presence / absence determination unit 14 can determine the presence or absence of fire precursors in the monitored area based on two different criteria (one being an abnormal value, and the other being a combination of measurement data).

[0135] Furthermore, according to the fire prediction detection system 10, the threshold resetting unit 19 can reset the threshold used by the measurement data acquisition unit 13 based on the abnormal values ​​extracted by the abnormal value extraction unit 17.

[0136] It should be noted that the present invention is not limited to the fire precursor detection system 10, fire precursor detection method, and fire precursor detection program according to the above-described embodiment, and can be implemented by various other modifications or applications without departing from the gist of the present invention as described in the claims. [Explanation of symbols]

[0137] 10. Fire Prediction Detection System 10a communication interface 10b ROM 10c RAM 10d storage section 10e CPU 10f Input / Output Interface 10g input device 10h output device 11 Internet 12 Reception Department 13 Measurement data acquisition unit 14. Precursor detection unit 15. Risk Assessment Unit 16 Hochi Department 17. Anomaly detection unit 18. Anomaly Prediction Unit 19. Threshold reset section 20 Environmental measurement terminals 20a communication interface 20b ROM 20c RAM 20d storage section 20e CPU 20f Input / Output Interface 21 Odor Sensor 22 2D Thermopile Radiation Temperature Sensor 23 Barometric pressure sensor 24 Temperature and Humidity Sensors 25 Air Pollution Sensor 26 Smoke Sensor 27 Carbon monoxide sensor 28. Carbon dioxide sensor 29 Illuminance sensor 30 Noise Sensor 31. Camera with night vision function 32 Oxygen Sensor 33 Ceiling 34 Stands 35 Unmanned aircraft 36. Four-legged robot 37 Helmets 41. Graph line of carbon monoxide concentration 42 Temperature graph line 43. Graph of carbon dioxide concentration

Claims

1. A fire prediction detection system that is installed in an area to be monitored, is capable of communicating with multiple types of sensors that measure the environment of that area, and detects signs of fire in that area, A measurement data acquisition unit that acquires measurement data for each of the multiple indicators of the environment of the region acquired by the multiple types of sensors, A fire precursor determination unit determines whether or not there are signs of fire in the area based on a combination of measurement data of a predetermined indicator selected from the measurement data acquired by the measurement data acquisition unit, If the aforementioned fire warning detection unit determines that there are signs of fire in the area, the notification unit notifies the monitor of the area that there are signs of fire. A fire precursor detection system characterized by being equipped with the following features.

2. The fire precursor detection system according to claim 1, characterized in that the precursor determination unit uses a precursor learning model that has previously learned the correspondence between a combination of measurement data of a predetermined indicator and the presence or absence of fire precursors in the area, and inputs a combination of measurement data of a predetermined indicator selected from the measurement data acquired by the measurement data acquisition unit into the precursor learning model to determine whether or not there are fire precursors in the area.

3. The system further includes a reception unit that accepts the type of area to be monitored, The aforementioned warning sign determination unit is: The fire prediction detection system according to claim 1, characterized in that the receiving unit determines the combination of measurement data for the indicator based on the type of area to be monitored, by referring to a table that defines the correspondence between the type of area to be monitored and the combination of measurement data for the indicator.

4. If the aforementioned fire warning determination unit determines that there are signs of fire in the area, the system further comprises a risk determination unit that determines the degree of risk of the fire warning based on a combination of measurement data of predetermined indicators that formed the basis for the determination by the aforementioned fire warning determination unit. The fire precursor detection system according to claim 1, characterized in that the notification unit notifies the monitor of the area that there is a fire precursor and the degree of danger of the precursor as determined by the danger determination unit when the presence or absence of precursor determination unit determines that there is a fire precursor in the area.

5. The fire precursor detection system according to claim 4, characterized in that the risk determination unit uses a risk learning model that has been pre-learned the correspondence between a predetermined combination of measurement data of the indicators and the risk level of fire precursors in the area, and inputs the predetermined combination of measurement data of the indicators that formed the basis for the determination by the precursor presence / absence determination unit that there are fire precursors in the area to the risk learning model, thereby determining the risk level of fire precursors in the area.

6. The aforementioned measurement data acquisition unit is One or more thresholds are set for the measurement data of each of the multiple indicators, and the measurement data that exceeds the threshold for each of the multiple types of sensors is acquired from each of the multiple types of sensors. A fire precursor detection system as described in item 1.

7. The fire prediction detection system according to claim 1, characterized in that the plurality of types of sensors include an odor sensor for measuring the odor in the area.

8. The fire prediction detection system according to claim 1, characterized in that the plurality of types of sensors include a two-dimensional thermopile radiation temperature sensor that measures a two-dimensional temperature distribution that divides the two-dimensional plane of the region into predetermined sections and indicates the temperature of each section as a numerical value.

9. The fire precursor detection system according to claim 1, characterized in that the plurality of types of sensors include a pressure sensor for measuring the atmospheric pressure in the region.

10. The fire precursor detection system according to claim 1, characterized in that the plurality of types of sensors include a temperature and humidity sensor that measures the temperature and humidity in a steady state of the region.

11. The fire prediction detection system according to claim 1, characterized in that the plurality of types of sensors include an air pollution sensor for measuring the air pollution of the area.

12. The fire prediction detection system according to claim 1, characterized in that the plurality of types of sensors include a smoke sensor for detecting the concentration of smoke in the area.

13. The fire precursor detection system according to claim 1, characterized in that the plurality of types of sensors include a carbon monoxide sensor for detecting the concentration of carbon monoxide in the region.

14. The fire prediction detection system according to claim 1, characterized in that the plurality of types of sensors include a carbon dioxide sensor for detecting the concentration of carbon dioxide in the region.

15. The fire prediction detection system according to claim 1, characterized in that the plurality of types of sensors include an illuminance sensor for measuring the illuminance of the area.

16. The fire precursor detection system according to claim 1, characterized in that the plurality of types of sensors include a noise sensor for measuring the magnitude of noise in the area.

17. The fire prediction detection system according to claim 1, characterized in that the plurality of types of sensors include an oxygen sensor for measuring the oxygen concentration in the region.

18. The aforementioned multiple types of sensors are mounted on an unmanned aerial vehicle or robot. The fire precursor detection system according to claim 1, characterized in that the unmanned aerial vehicle or robot flies or travels in the area by remote control or autonomous control.

19. An outlier extraction unit extracts outliers from the measurement data of each of the multiple environmental indicators of the region included in the measurement data by performing unsupervised clustering on the measurement data acquired by the measurement data acquisition unit, as abnormal values ​​for each of the multiple environmental indicators. An abnormal value prediction unit that determines whether or not there are signs of fire in the area based on the abnormal value, The fire precursor detection system according to claim 1, further comprising the following:

20. An outlier extraction unit extracts outliers from the measurement data of each of the multiple environmental indicators of the region included in the measurement data by performing unsupervised clustering on the measurement data acquired by the measurement data acquisition unit, as abnormal values ​​for each of the multiple environmental indicators. The system further includes a threshold resetting unit that resets the threshold values ​​for each of the measurement data of the multiple indicators based on the aforementioned abnormal values, The aforementioned measurement data acquisition unit is Measurement data exceeding the threshold reset by the threshold resetting unit for each of the multiple indicators is acquired from each of the multiple types of sensors. The fire precursor detection system according to claim 6.

21. A fire precursor detection method that is installed in an area to be monitored, is capable of communicating with multiple types of sensors that measure the environment of the area, and detects signs of fire in the area, Computers A measurement data acquisition step involves acquiring measurement data for each of the multiple indicators of the environment of the region obtained by the multiple types of sensors, A premonitory presence / absence determination step, which determines whether or not there are signs of fire in the area based on a combination of measurement data of a predetermined indicator selected from the measurement data acquired in the measurement data acquisition step, If, in the aforementioned indicator presence determination step, it is determined that there are signs of fire in the area, a notification step is made to notify the monitor of the area that there are signs of fire; A fire precursor detection method characterized by performing the following actions.

22. A fire precursor detection program that is installed in an area to be monitored, is capable of communicating with multiple types of sensors that measure the environment of that area, and detects signs of fire in that area, On the computer, A measurement data acquisition function that acquires measurement data for each of the multiple indicators of the environment of the region acquired by the multiple types of sensors, A premonitory sign determination function that determines whether or not there are signs of fire in the area based on a combination of measurement data of a predetermined indicator selected from the measurement data acquired in the measurement data acquisition function, If the aforementioned fire warning detection function determines that there are signs of fire in the area, the notification function notifies the monitor of the area that there are signs of fire. A fire precursor detection program characterized by its ability to perform the following actions.

Images