Abnormality discrimination program, monitoring system equipped with the same, and fire monitoring system

Abstract

To provide an abnormality discrimination program that can increase or decrease variation in a value of a physical quantity at a normal time and can adjust sensitivity with which a processor discriminates the occurrence of an abnormality.SOLUTION: An abnormality discrimination program is capable of causing a processor 12 to execute arithmetic processing including random number calculation processing S3 for calculating a random number value for increasing significant digits of a physical quantity x, and random number addition processing S4 for adding the random number value to the physical quantity x. The random number calculation processing S3 causes the processor 12 to calculate a random number value in accordance with a first condition that sets an average value of the calculated random number values to 0, and a second condition that sets a set value for determining a width from the upper limit to the lower limit of the calculated random number values to α. By increasing or decreasing the set value α, it is possible to increase or decrease variation in values of the physical quantity x, and adjust sensitivity with which the processor 12 discriminates the occurrence of an abnormality.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to an anomaly discrimination program for forming a population of physical quantities measured under normal conditions and discriminating the occurrence of an anomaly that has a low probability of occurring under normal conditions based on this population, as well as a monitoring system and a fire monitoring system equipped with the same. [Background technology]

[0002] In Japanese Patent Laid-Open Publication No. 2023-168656 (Patent Document 1), the present applicant proposes an anomaly discrimination program that enables early discrimination of the occurrence of an anomaly that has a low probability of occurring under normal circumstances. This anomaly discrimination program applies statistical logic. That is, a population is formed for physical quantities measured under normal circumstances, and events that have a low probability of occurring under normal circumstances are discriminated as an anomaly based on this population.

[0003] For example, when determining whether a fire has occurred based on the temperature of a monitored area, the normal temperature is continuously measured by multiple sensors installed on the ceiling of the monitored area. The normal temperature is measured every second for 60 seconds, for example. The temperatures measured by each sensor are normalized according to a standard normal distribution. The normalized temperatures of each sensor are weighted taking into account the conditions of the monitored area, and the sum of the weighted temperatures of each sensor is calculated. The sum of the temperatures of each sensor, repeatedly calculated over 60 seconds, is stored in a database. This forms a population of normal temperatures in the monitored area. Thereafter, it is determined whether the temperature of the monitored area is abnormal based on this population. Whether the temperature of the monitored area is abnormal is determined based on a predetermined probability threshold. In other words, if the probability of the sum of the temperatures measured by each sensor occurring is below a threshold, the sum of the temperatures is determined to be abnormal. The probability threshold is, for example, 1.0 x 10 -4 ~1.0×10 -8 It is set between.

[0004] This type of conventional anomaly detection program can detect a temperature abnormality caused by a fire when the temperature in the monitored area rises from the normal 20°C to approximately 25°C. The time it takes for the temperature abnormality to be detected is approximately 90 seconds after the fire source is ignited, which is faster than the response time of a "special type" fixed-temperature heat detector, which is said to have the fastest response time. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2023-168656 Summary of the Invention [Problem to be solved by the invention]

[0006] In an actual monitored area, physical quantities change extremely slowly under normal conditions. For example, the room temperature in an office building, a tenant building, or a logistics warehouse varies greatly between day and night, but changes slowly and continuously over a long period of time. For this reason, there is almost no variation in the values ​​of physical quantities measured under normal conditions by sensors installed in the monitored area (for example, temperature measured every second over a period of 60 seconds). As a result, conventional anomaly detection programs have the high sensitivity to instantly identify slight changes in physical quantities in the monitored area as an anomaly.

[0007] However, in an actual monitoring area, physical quantities may change in a relatively short time due to normal causes. For example, if many people enter a room that is a monitoring area, the temperature in the room will rise in a relatively short time due to the body temperatures of the many people. In such cases, conventional anomaly detection programs have the problem of instantly determining that a change in physical quantities due to normal causes is an anomaly.

[0008] To solve this problem, we set a probability threshold (e.g., 1.0 × 10 -4 ~1.0×10 -8However, because there is almost no variation in the values ​​of physical quantities under normal conditions, the probability of a change in the physical quantity itself being present is low, and even if the probability threshold is lowered, it is not possible to delay the timing at which an abnormality is determined.

[0009] The present invention has been made in consideration of the above-mentioned problems, and aims to provide an abnormality detection program that can increase or decrease the variation in the values ​​of physical quantities under normal conditions and that enables a processor to adjust the sensitivity for detecting the occurrence of an abnormality, as well as a monitoring system and a fire monitoring system that are equipped with the program. [Means for solving the problem]

[0010] (A) In order to achieve the above object, the anomaly discrimination program of the present invention is an anomaly discrimination program for causing a processor to discriminate the occurrence of an anomaly that has a low probability of occurring under normal circumstances, based on a database of a physical quantity x measured by a sensor under normal circumstances, and is capable of causing the processor to execute arithmetic processes including a random number calculation process for calculating a random number value to increase the significant digits of the physical quantity x, and a random number addition process for adding the random number value to the physical quantity x, wherein the random number calculation process causes the processor to calculate the random number value in accordance with a first condition that sets an average value of the calculated random number values ​​to 0, and a second condition that sets a set value α to determine a range from an upper limit to a lower limit of the calculated random number value, and the set value α is configured to increase or decrease the variation in the value of the physical quantity x, thereby enabling the processor to adjust the sensitivity for discriminating the occurrence of an anomaly.

[0011] (B) Preferably, in the abnormality discrimination program of (A) above, the random number calculation process causes the processor to calculate the random number value according to a normal distribution whose mean value is 0 and whose standard deviation is the set value α.

[0012] (C) Preferably, in the abnormality discrimination program of (A) above, the random number calculation process causes the processor to calculate the random number value according to a uniform distribution in which the average value is 0 and the range from the upper limit to the lower limit of the calculated random number value is the set value α.

[0013] (D) Preferably, in the abnormality discrimination program of (A) above, the physical quantity x is any one of the values ​​of temperature, smoke concentration, or infrared intensity measured by each of the plurality of sensors installed in a specified monitoring area.

[0014] (E) Preferably, in the anomaly discrimination program of (D) above, a first process of collecting a plurality of physical quantities x measured by a plurality of sensors for a predetermined time or a predetermined number of times under normal conditions where no fire has occurred and converting each of the collected physical quantities x into a normalized physical quantity y; and a first process of converting a weight w corresponding to each of the plurality of sensors into a normalized physical quantity y by multiplying the weight w by the weight w. prev a second process of collecting a plurality of physical quantities x measured by a plurality of the sensors after the predetermined time has elapsed or the predetermined number of times has been exceeded, converting each of the collected physical quantities x into a normalized physical quantity y, and calculating a sum S of the physical quantities y multiplied by a weight w; new The third process calculates the sum S new Average value of [Table 1] and standard deviation σ new The fourth process calculates the average value [Table 2] , standard deviation σ new and the sum S prev Average value of [Table 3] and a fifth process of determining whether or not the probability calculated by applying the cumulative distribution function of the above equation is equal to or less than a preset probability threshold, and determining that an abnormality has occurred if the probability is equal to or less than the threshold.

[0015] (F) Preferably, in the anomaly discrimination program of (D) above, a first process of collecting a plurality of physical quantities x measured by a plurality of sensors for a predetermined time or a predetermined number of times under normal conditions where no fire has occurred and converting each of the collected physical quantities x into a normalized physical quantity y; and a first process of converting a weight w corresponding to each of the plurality of sensors into a normalized physical quantity y by multiplying the weight w by the weight w. prev a second process of collecting a plurality of physical quantities x measured by a plurality of the sensors after the predetermined time has elapsed or the predetermined number of times has been exceeded, converting each of the collected physical quantities x into a normalized physical quantity y, and calculating a sum S of the physical quantities y multiplied by a weight w; new The third process calculates the sum S new Average value of [Table 4] and standard deviation σ new A fourth process of calculating Average [Table 5] , standard deviation σ new and the sum S prev Average value of [Table 6] to the cumulative distribution function is equal to or less than a preset probability threshold, and if it is equal to or less than the threshold, starts counting a predetermined length of discrimination delay time, and if probabilities equal to or less than the threshold are continuously calculated before the discrimination delay time is reached, determines that an abnormality has occurred.

[0016] (G) Preferably, in the abnormality discrimination program of (F) above, the discrimination delay time is set within a range of 15 seconds to 150 seconds.

[0017] (H) Preferably, in the abnormality determination program of (F), the processor is configured to execute a learning mode, which is a process for automatically setting the determination delay time, The learning mode includes: The same processes as the first to fourth processes, respectively; Average [Table 7] , standard deviation σ new and the sum S prev Average value of [Table 8] a low probability discrimination process for determining whether the probability calculated by applying the above formula to the cumulative distribution function is equal to or less than a preset probability threshold; a continuous time counting process that starts counting time when the probability calculated in the low probability discrimination process is equal to or less than the threshold, continues counting time until a probability exceeding the threshold is calculated, and, when a probability exceeding the threshold is calculated, stores the time during which a probability equal to or less than the threshold is continuously calculated in a database; and a discrimination delay time setting process for identifying the longest time from among the plurality of times stored in the database, and setting a time exceeding the longest time as the discrimination delay time.

[0018] (I) Preferably, in the abnormality discrimination program of (H) above, a period is set for the processor to execute the learning mode, and the processor executes the discrimination delay time setting process after the set period has elapsed.

[0019] (J) In order to achieve the above object, the monitoring system of the present invention is a monitoring system that operates in accordance with the abnormality discrimination program described in any one of (A) to (I) above, and includes at least one of the sensors, the processor that executes processing in accordance with the abnormality discrimination program based on a signal indicating a physical quantity x output from the sensor, and an alarm unit that operates when the processor determines that an abnormality has occurred.

[0020] (K) In order to achieve the above object, the first fire monitoring system of the present invention is a fire monitoring system that operates according to the abnormality discrimination program described in any one of (D) to (I) above, and includes a plurality of the sensors and a receiver, the receiver includes a receiving unit, the processor, and an alarm unit, and the abnormality discrimination program is installed, the receiving unit receives signals indicating a physical quantity x output from the plurality of sensors, the processor executes processing according to the abnormality discrimination program based on the physical quantity x, and the alarm unit operates when the processor determines that an abnormality has occurred.

[0021] (L) In order to achieve the above object, the second fire monitoring system of the present invention is a fire monitoring system that operates according to the abnormality discrimination program described in any of (D) to (I) above, and includes a plurality of the sensors and a repeater, the repeater includes a receiving unit and the processor, and the abnormality discrimination program is installed, the receiving unit receives signals indicating a physical quantity x output from the plurality of sensors, and the processor executes processing in accordance with the abnormality discrimination program based on the physical quantity x, and outputs an abnormality detection signal when it determines that an abnormality has occurred. [Effects of the Invention]

[0022] In the anomaly detection program, the monitoring system, and the fire monitoring system including the program, a random value for increasing the number of significant digits is added to a physical quantity under normal conditions. This random value is calculated under a first condition that sets the average value of the calculated random value to 0 and a second condition that sets a set value α for determining the range from the upper limit to the lower limit of the calculated random value. Therefore, by increasing or decreasing the set value α depending on the conditions of the monitored area, the range from the upper limit to the lower limit of the random value added to the physical quantity under normal conditions can be varied, thereby increasing or decreasing the variability in the value of the physical quantity under normal conditions. This allows the processor to adjust its sensitivity for determining the occurrence of an abnormality. That is, by increasing the variability in the value of the physical quantity under normal conditions, the processor's sensitivity for determining the occurrence of an abnormality can be reduced. On the other hand, by decreasing the variability in the value of the physical quantity under normal conditions, the processor's sensitivity for determining the occurrence of an abnormality can be increased. [Brief explanation of the drawings]

[0023] [Figure 1] FIG. 1 is a schematic diagram showing a first embodiment of a fire monitoring system equipped with an abnormality determination program of the present invention. [Figure 2] FIG. 2 is a block diagram showing a receiver that constitutes the above-mentioned fire monitoring system. [Figure 3] Figure 3(a) is a graph showing the relationship between the probability density and random values ​​calculated according to a normal distribution with a mean value of 0 and a standard deviation α of 0.5. Figure 3(b) is a graph showing the relationship between the probability density and random values ​​calculated according to a normal distribution with a mean value of 0 and a standard deviation α of 1.0. [Figure 4] Figure 4(a) is a graph showing the relationship between the probability density and random values ​​calculated according to a normal distribution with a mean value of 0 and a standard deviation α of 1.5. Figure 4(b) is a graph showing the relationship between the probability density and random values ​​calculated according to a normal distribution with a mean value of 0 and a standard deviation α of 2.0. [Figure 5]Figure 5(a) is a graph showing the relationship between the probability density and random values ​​calculated according to a uniform distribution with an average value of 0 and a range α from upper to lower limits set to ±0.5. Figure 5(b) is a graph showing the relationship between the probability density and random values ​​calculated according to a uniform distribution with an average value of 0 and a range α from upper to lower limits set to ±1.0. [Figure 6] Figure 6(a) is a graph showing the relationship between the probability density and random values ​​calculated according to a uniform distribution with an average value of 0 and a range α of ±1.5 from the upper limit to the lower limit. Figure 6(b) is a graph showing the relationship between the probability density and random values ​​calculated according to a uniform distribution with an average value of 0 and a range α of ±2.0 from the upper limit to the lower limit. [Figure 7] FIG. 7 is an explanatory diagram for explaining a physical quantity x measured by a plurality of sensors and a normalized physical quantity y. [Figure 8] FIG. 8 is an explanatory diagram for explaining the sum S of the physical quantities y. [Figure 9] FIG. 9 is an explanatory diagram for explaining the principle of abnormality discrimination in the abnormality discrimination program. [Figure 10] Figure 10 shows the conditions of the monitoring area where temperatures were measured for use in simulating the anomaly detection program, with Figure 10(a) being a side view of the ceiling, sensors 1 to 4, and the fire source, and Figure 10(b) being a plan view of the ceiling, sensors 1 to 4, and the fire source. [Figure 11] 11(a) is a graph showing the temperature measurement values ​​of the sensor 1, and FIG. 11(b) is a graph showing the temperature measurement values ​​of the sensor 2. In FIG. [Figure 12] 12(a) is a graph showing the temperature measurement values ​​of the sensor 3, and FIG. 12(b) is a graph showing the temperature measurement values ​​of the sensor 4. In FIG. [Figure 13] Fig. 13(a) is a graph showing the transition of probability when a random number value of a normal distribution with a mean value of 0 and a standard deviation α of 0.5 is added to a physical quantity x. Fig. 13(b) is a graph showing the transition of probability when a random number value of a normal distribution with a mean value of 0 and a standard deviation α of 1.0 is added to a physical quantity x. [Figure 14]Fig. 14(a) is a graph showing the transition of probability when a random value having a normal distribution with a mean value of 0 and a standard deviation α of 1.5 is added to a physical quantity x. Fig. 14(b) is a graph showing the transition of probability when a random value having a normal distribution with a mean value of 0 and a standard deviation α of 2.0 is added to a physical quantity x. [Figure 15] Fig. 15(a) is a graph showing the transition of probability when a uniformly distributed random value with an average value of 0 and a range α of ±0.5 from the upper limit to the lower limit is added to a physical quantity x. Fig. 15(b) is a graph showing the transition of probability when a uniformly distributed random value with an average value of 0 and a range α of ±1.0 from the upper limit to the lower limit is added to a physical quantity x. [Figure 16] Fig. 16(a) is a graph showing the transition of probability when a uniformly distributed random number value with an average value of 0 and a range α of ±1.5 from the upper limit to the lower limit is added to a physical quantity x. Fig. 16(b) is a graph showing the transition of probability when a uniformly distributed random number value with an average value of 0 and a range α of ±2.0 from the upper limit to the lower limit is added to a physical quantity x. [Figure 17] FIG. 17 is a schematic diagram showing a second embodiment of a fire monitoring system equipped with an abnormality determination program of the present invention. [Figure 18] FIG. 18 is a block diagram showing a receiver constituting a fire monitoring system according to the third embodiment of the present invention. [Figure 19] Figure 19 shows the conditions of the monitoring area (warehouse model) used in the first simulation of the anomaly discrimination program executed by the control unit of the receiver. Figure 19(a) is a plan view of the ceiling, sensors A1 to A10, B1 to B10, and the fire source (six-tier crib), and Figure 19(b) is a side view of the ceiling, sensors A1 to A10, B1 to B10, and the fire source. [Figure 20] Figure 20(a) is a graph showing the change in the heat release rate (kW) of the fire source used in the first simulation. The solid line in Figure 20(a) is a graph of the actual measured values ​​of the heat release rate of the fire source, and the dotted line is a graph of the calculated values ​​of the heat release rate obtained by fitting the actual measured values ​​of the heat release rate of the fire source to the αt2 model. Figure 20(b) is a graph showing the transition of the probability resulting from the first simulation. [Figure 21] Figure 21(a) is a side view showing the conditions of the monitoring area (warehouse model) used in a second simulation of the anomaly detection program executed by the control unit of the receiver. In the second simulation, the fire source installed in the monitoring area is changed from the six-tiered crib to two jet heaters. Figure 21(b) is a graph showing the transition of probability as a result of the second simulation. [Figure 22] FIG. 22 is a block diagram showing the flow of processing in the learning mode of the abnormality determination program executed by the control unit of the receiver. DETAILED DESCRIPTION OF THE INVENTION

[0024] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an abnormality discrimination program and a fire monitoring system including the program according to the present invention will be described below with reference to the accompanying drawings.

[0025] The anomaly discrimination program of the present invention discriminates whether an anomaly has occurred based on the probability that a measured physical quantity x occurs. Therefore, the discrimination target of the anomaly discrimination program of the present invention is not limited to anomalies caused by fires. However, for convenience of explanation, an embodiment of the anomaly discrimination program and the fire monitoring system will be described using an anomaly caused by a fire as an example of the discrimination target.

[0026] 1. First embodiment of fire monitoring system FIG. 1 shows a fire monitoring system according to a first embodiment of the present invention. The fire monitoring system according to this embodiment is installed in a large-scale facility, such as an office building, a tenant building, or a logistics warehouse. The fire monitoring system is primarily composed of multiple sensors 1 to 4 and a receiver 10. The sensors 1 to 4 are installed, for example, on the ceiling of a predetermined monitoring area in the large-scale facility. Although not shown, numerous sensors, each with the same or different configuration as the sensors 1 to 4, are installed in monitoring areas throughout the large-scale facility. Signals output from the sensors 1 to 4 and other sensors are input to the receiver 10 via a signal line 5 indicated by a dashed line in FIG. 1. The receiver 10 monitors the entire large-scale facility based on the signals output from the sensors 1 to 4 and other sensors. The receiver 10 can monitor for fires and other problems. Problems other than fires include, for example, gas-related problems, electrical problems, problems with fire prevention equipment installed in the large-scale facility, and abnormalities in communication between the receiver 10 and peripheral devices linked to the receiver 10.

[0027] 1.2 Multiple Sensors The sensors 1 to 4 are, for example, analog fire detectors. The sensors 1 to 4 have the same configuration and include at least one detection element for measuring a physical quantity x, such as temperature, smoke density, or infrared intensity. In a predetermined monitoring area, each of the sensors 1 to 4 measures the physical quantity x that changes due to the occurrence of a fire and generates an analog signal corresponding to the physical quantity x. Furthermore, each of the sensors 1 to 4 converts the analog signal corresponding to the physical quantity x into a digital signal and outputs it. This digital signal includes an address assigned to each of the sensors 1 to 4. The digital signal output from each of the sensors 1 to 4 is input to the receiver 10 via a signal line 5.

[0028] The multiple sensors 1 to 4 are not limited to multiple fire detectors. For example, the multiple sensors 1 to 4 may be differential distributed detectors. A differential distributed detector is typically configured with multiple thermocouples electrically connected at predetermined intervals along a single electric wire. Each of the multiple thermocouples generates an electromotive force corresponding to the amount of heat detected.

[0029] The sensors 1 to 4 may also be temperature measuring devices using optical fibers. This temperature measuring device is primarily configured with a single optical fiber, a laser light source, and a light detection unit. The optical fiber is installed, for example, on the ceiling of a predetermined monitoring area in a large-scale facility. The laser light source emits pulsed laser light into the input end of the optical fiber. A portion of the incident light is reflected by glass particles in the optical fiber, generating reflected light of Rayleigh scattered light, Brillouin scattered light, and Raman scattered light. These reflected light beams return to the input end. Of these reflected light components, the intensity of the Raman scattered light and the frequency of the Brillouin scattered light change depending on the temperature of the optical fiber. Therefore, it is possible to measure the temperature based on the intensity of the Raman scattered light and the frequency of the Brillouin scattered light. Furthermore, it is possible to identify the position on the optical fiber where the reflected light originates based on the time it takes for the reflected light to return after the incident light enters.

[0030] 1.3 Receiver A receiver 10 shown in FIG. 1 is installed, for example, in a disaster prevention center provided in a large-scale facility. The receiver 10 is mainly composed of a receiving unit 11, a control unit 12, and a notification unit 13 shown in FIG. 2. In FIG. 2, the receiving unit 11 is electrically connected to a plurality of sensors 1 to 4 via a signal line 5 in FIG. 1. The receiving unit 11 receives digital signals output from the plurality of sensors 1 to 4. These digital signals include physical quantities x measured by each of the plurality of sensors 1 to 4. The receiving unit 11 transmits the received digital signals to a control unit 12. For convenience of explanation, the digital signals processed by the control unit 12 will be expressed as the physical quantities x measured by the plurality of sensors 1 to 4.

[0031] The control unit 12 includes a processor for executing calculation processing. The processor of the control unit 12 executes the processing of steps S1 to S11, S12 to S15, and S21 to S26 shown in Fig. 2 in accordance with the abnormality discrimination program according to this embodiment. The abnormality discrimination program is capable of discriminating the occurrence of a fire in the same manner as conventional fire monitoring systems, and also of discriminating slight abnormalities in the early stages of a fire. The abnormality discrimination program according to this embodiment will be described in detail later.

[0032] 1, and an LED and a speaker (not shown). The notification unit 13 displays various information on the touch panel based on a signal output from the control unit 12. For example, when the control unit 12 detects the occurrence of a fire or an abnormality, the notification unit 13 notifies the occurrence of the fire or abnormality by an image displayed on the touch panel, the color of the LED light, and a sound output from the speaker.

[0033] 1.4 Abnormality detection program The processing of the control unit 12 according to the abnormality determination program of this embodiment will be described. In Fig. 2, steps S1 to S11 are the processing of the main routine of the abnormality determination program. Furthermore, steps S12 to S15 and S21 to S26 are the processing of subroutines accompanying the main routine.

[0034] 1.4.1 Fire detection processing The control unit 12 executes a fire detection process in step S1 based on the physical quantity x measured by the plurality of sensors 1 to 4. In the fire detection process S1, the control unit 12 compares the physical quantity x with a first threshold value set in advance in step S12. For example, if the physical quantity x is a temperature, a temperature of "70°C" is set as the first threshold value. In this case, the control unit 12 determines whether the physical quantity x is 70°C or higher. If the result of this determination is that the physical quantity x is 70°C or higher (YES), the control unit 12 transmits a signal to the alarm unit 13 to notify the occurrence of a fire. On the other hand, if the temperature is not 70°C or higher (NO), the control unit 12 executes a physical quantity x collection process in step S2.

[0035] In the abnormality determination program of this embodiment, the fire determination process S1 is executed at the earliest stage, so that even if a fire does break out, the occurrence of the fire is quickly notified by the alarm unit 13. As a result, the disaster prevention center where the receiver 10 is installed can quickly extinguish the fire in its initial stage.

[0036] 1.4.2 Physical quantity x collection processing In the physical quantity x collection process of step S2, the control unit 12 continuously collects the physical quantity x measured by the multiple sensors 1 to 4. The control unit 12 collects the physical quantity x at predetermined time intervals for each of the multiple sensors 1 to 4. The time interval for collecting the physical quantity x is determined depending on the number of the multiple sensors 1 to 4, the length of the signal line 5, the processing capacity of the processor of the control unit 12, etc.

[0037] In step S21, the collected physical quantities x are stored in chronological order in databases corresponding to the plurality of sensors 1 to 4. Every time a new physical quantity x is measured, the physical quantity x is accumulated in the database. When the storage area of ​​the database is filled with the physical quantities x, the oldest physical quantity x is deleted and the latest physical quantity x is added. In this way, the physical quantities x stored in the database are constantly updated during operation of the abnormality determination program of this embodiment.

[0038] Here, the significance of the physical quantity x stored in the database in step S21 will be explained. First, the physical quantity x stored in the database is determined to be less than the first threshold in the fire discrimination process S1 described above. That is, all of the physical quantities x stored in the database are measured under conditions in which no fire has been reported. Next, the physical quantities x stored in the database may have been measured under either normal or abnormal conditions. A normal condition is a condition in which no event that could cause a fire has occurred. An abnormal condition is a condition in which no fire has been reported yet, but an event that could cause a fire has occurred. Then, the collection of time-series physical quantities x measured under normal conditions serves as a basis for determining whether or not an abnormality has occurred thereafter.

[0039] 1.4.3 Random number calculation process Many of the general sensors 1 to 4 can measure the physical quantity x to at most one decimal place. Therefore, the physical quantity x measured by the sensors 1 to 4 in a normal state tends to be a substantially constant value. If the physical quantity x in a normal state becomes a substantially constant value, the control unit 12 will determine that a slight change in the physical quantity x is an abnormality in the abnormality determination process in step S11, which will be described later. In other words, the sensitivity of the control unit 12 for determining the occurrence of an abnormality will be too high.

[0040] Therefore, in the random number calculation process of step S3, the control unit 12 calculates a random value to increase the significant digits of the physical quantity x and to provide variation to the value of the physical quantity x. The control unit 12 calculates a random value to the nth decimal place in accordance with a first condition that sets the average value of the calculated random values ​​to 0 and a second condition that sets a set value α to determine the range from the upper limit to the lower limit of the calculated random values. The reason for setting the average value of the random values ​​to 0 is to keep the random values ​​within the measurement error range of the physical quantity x under normal conditions. Increasing the set value α increases the variation in the random values. On the other hand, decreasing the set value α decreases the variation in the random values. In other words, increasing or decreasing the set value α can increase or decrease the variation in the physical quantity x to which the random value is added, which allows the control unit 12 to adjust the sensitivity for determining the occurrence of an abnormality.

[0041] <Calculating random numbers according to a normal distribution> For example, the control unit 12 can calculate random values ​​according to a normal distribution with a mean value of 0 and a standard deviation of a set value α. Figures 3(a), (b) and 4(a), (b) show normal distributions that the control unit 12 uses to calculate random values.

[0042] Figure 3(a) shows a normal distribution with a mean value of 0 and a standard deviation α of 0.5. The horizontal axis of Figure 3(a) represents the random value calculated by the control unit 12. The vertical axis of Figure 3(a) represents the probability density of the random value calculated by the control unit 12. The random value on the horizontal axis is a continuous random variable with consecutive values ​​to the nth decimal place. The probability density on the vertical axis represents the relative likelihood of a value being obtained in a continuous random variable. A random value with a higher probability density is more likely to be calculated, and a random value with a lower probability density is less likely to be calculated. As shown in Figure 3(a), when the mean value is set to 0 and the standard deviation α is set to 0.5, a random value within the range of 0 ± 0.5 is calculated with a probability of approximately 68%. The probability density of the random value 0, which is the most likely to be calculated, is 0.8.

[0043] Figure 3(b) shows a normal distribution with a mean value of 0 and a standard deviation α of 1.0. When the mean value is 0 and the standard deviation α is 1.0, a random value within the range of 0 ± 1.0 is generated with a probability of approximately 68%. The probability density of the most likely random value, 0, is 0.4. Figure 4(a) shows a normal distribution with a mean value of 0 and a standard deviation α of 1.5. When the mean value is 0 and the standard deviation α is 1.5, a random value within the range of 0 ± 1.5 is generated with a probability of approximately 68%. The probability density of the most likely random value, 0, is 0.27. Figure 4(b) shows a normal distribution with a mean value of 0 and a standard deviation α of 2.0. When the mean value is 0 and the standard deviation α is 2.0, a random value within the range of 0 ± 2.0 is generated with a probability of approximately 68%. The probability density of the most likely random value, 0, is 0.2.

[0044] As shown in Figures 3(a) and 3(b) and Figures 4(a) and 4(b), the smaller the value of the standard deviation α, the narrower the range from the upper limit to the lower limit of the random number values ​​calculated by the control unit 12, and the higher the probability density of random number values ​​close to the average value 0. In other words, the smaller the value of the standard deviation α, the easier it is to calculate random number values ​​close to the average value 0, and the variance of the random number values ​​is reduced. On the other hand, the larger the value of the standard deviation α, the wider the range from the upper limit to the lower limit of the random number values ​​calculated by the control unit 12, and the lower the probability density of random number values ​​close to the average value 0. In other words, the larger the value of the standard deviation α, the harder it is to calculate random number values ​​close to the average value 0, and the variance of the random number values ​​is increased.

[0045] <Calculating random values ​​according to a uniform distribution> Furthermore, for example, the control unit 12 can calculate random values ​​according to a uniform distribution with an average value of 0 and a set value α between the upper and lower limits of the random value. Figures 5(a), (b) and 6(a), (b) show the uniform distribution used by the control unit 12 to calculate random values.

[0046] FIG. 5(a) shows a uniform distribution with a mean value of 0 and a range α from the upper limit to the lower limit of the random value set to ±0.5. The horizontal axis of FIG. 5(a) represents the random value calculated by the control unit 12. The vertical axis of FIG. 5(a) represents the probability density of the random value calculated by the control unit 12. The random value on the horizontal axis is a continuous random variable with consecutive values ​​to the nth decimal place. The probability density on the vertical axis represents the relative likelihood of a value being obtained in a continuous random variable. A random value with a higher probability density is easier to calculate, and a random value with a lower probability density is harder to calculate. As shown in FIG. 5(a), when the mean value is set to 0 and the range α is set to ±0.5, all random value values ​​within the range of 0±0.5 are calculated with the same probability density of 1.

[0047] Figure 5(b) shows a uniform distribution with a mean value of 0 and a range α between the upper and lower limits of the random value set to ±1.0. When the mean value is set to 0 and the range α is set to ±1.0, all random values ​​within the range of 0 ±1.0 are calculated with the same probability density of 0.5. Figure 6(a) shows a uniform distribution with a mean value of 0 and a range α between the upper and lower limits of the random value set to ±1.5. When the mean value is set to 0 and the range α is set to ±1.5, all random values ​​within the range of 0 ±1.5 are calculated with the same probability density of 0.33. Figure 6(b) shows a uniform distribution with a mean value of 0 and a range α between the upper and lower limits of the random value set to ±2.0. When the mean value is set to 0 and the range α is set to ±2.0, all random values ​​within the range of 0 ±2.0 are calculated with the same probability density of 0.25.

[0048] As shown in Figures 5(a) and 5(b) and Figures 6(a) and 6(b), when the value of the width α from the upper limit to the lower limit of the random number values ​​is reduced, the width from the upper limit to the lower limit of the random number values ​​calculated by the control unit 12 simply becomes narrower, and the probability density of all the calculated random number values ​​becomes higher. As a result, the variance of the random number values ​​calculated by the control unit 12 is reduced. When the value of the width α from the upper limit to the lower limit of the random number values ​​is increased, the width from the upper limit to the lower limit of the random number values ​​calculated by the control unit 12 simply becomes wider, and the probability density of all the calculated random number values ​​becomes lower. As a result, the variance of the random number values ​​calculated by the control unit 12 is increased.

[0049] <Calculating random numbers according to other distributions> As described above, the process of calculating random values ​​according to a normal distribution and a uniform distribution has been exemplified, but the algorithm used to calculate the random values ​​is not limited to the normal distribution and the uniform distribution. For example, a random value to be added to the physical quantity x may be calculated by appropriately selecting a t distribution, an F distribution, a binomial distribution, a chi-squared distribution, a Poisson distribution, a gamma distribution, or the like.

[0050] 1.4.4 Random number addition process 2, in the random number addition process of step S4, the control unit 12 adds the random number value having the nth decimal place calculated in step S3 to each of the physical quantities x collected in step S2. In step S21, the physical quantities x to which the random numbers have been added are stored in chronological order in databases corresponding to the multiple sensors 1 to 4. Next, the control unit 12 proceeds to the physical quantity y conversion process of step S5.

[0051] 1.4.5 Physical quantity y conversion process In the physical quantity y conversion process in step S5, the control unit 12 converts each of the physical quantities x to which the random number values ​​have been added in step S4 into a normalized physical quantity y according to the following formula (1).

number

[0052] Here, normalization in the conversion process to the physical quantity y will be described with reference to FIG. 7. As shown in FIG. 7, the physical quantity x measured by the multiple sensors 1 to 4 is a continuous variable and follows a normal distribution N(μ,σ) with a mean value μ and a standard deviation σ. However, the multiple sensors 1 to 4 are installed under different conditions in the predetermined monitoring area shown in FIG. 1. For example, the multiple sensors 1 to 4 are installed at different positions on the ceiling and at different distances from the air outlet of the air conditioner 100. Therefore, the mean value μ and standard deviation σ of the physical quantity x measured by the multiple sensors 1 to 4 will be different from one another. Therefore, the physical quantity x measured by the multiple sensors 1 to 4 is converted into a normalized physical quantity y using the above formula (1). As shown in FIG. 7, the normalized physical quantity y follows a standard normal distribution N(0,1) with a mean value 0 and a standard deviation 1. That is, the normal distribution N(μ,σ) of the physical quantity x measured by the multiple sensors 1 to 4 is aligned with the standard normal distribution N(0,1) of the normalized physical quantity y. A collection of the normalized physical quantities y forms a population that follows the normal distribution N(0,1). In step S22 of FIG. 2, the normalized physical quantities y are stored in chronological order in databases corresponding to the multiple sensors 1 to 4, respectively.

[0053] The average value in the above formula (1) [Table 10] The physical quantity x stored in the database in step S21 is used to calculate the average value. [Table 11] The number of data points of the physical quantity x used to calculate is not particularly limited. For example, the average value [Table 12] At the timing of calculating the physical quantity x, all the physical quantities x stored in the database in step S21 may be used. Also, for example, among the physical quantities x stored in the database in step S21, the physical quantities x for the most recent tens of minutes or tens of seconds may be calculated as an average value. [Table 13] It may also be used to calculate

[0054] 1.4.6 Sum S prev Calculation process The sum S in step S6 prev In the calculation process, the control unit 12 acquires the physical quantity y from the database corresponding to each of the plurality of sensors 1 to 4 in step S22, and calculates the sum S of the physical quantities y multiplied by the weight w. prev is calculated using the following formula (2).

number

[0055] In step S13, the weight w is set in advance for each of the multiple sensors 1 to 4 based on the interrelationships between the sensors 1 to 4 (including the interrelationships between the same sensors). For example, Table 1 below shows weights w set based on the mutual distances r between the multiple sensors 1 to 4. [Table 14]

[0056] The weight w shown in Table 1 is set to "1" when the distance r from sensor P to sensor Q is in the relationship r≦0.18H with respect to the height H (m) from the floor to the ceiling, and set to "r-3 / 2 "

[0057] The sum S of the above formula (2) prev is the physical quantity y of sensor Q for each sensor P. (Q) Weight w (Q,P) 8, (a) to (d) show the physical quantity y in the above formula (2) when the sensors P and Q are sensors 1 to 4, respectively. (Q) and weight w (Q,P) The physical quantities y (Q) Each of these has a weight w (Q,P) The sum of all the multiplied values ​​is the sum S of the above formula (2). prev is.

[0058] In addition, the total S prev The formula for calculating may be either formula (3) or (4) below.

number

number

[0059] Sumwa S prev The physical quantity y used to calculate (p) and physical quantity y (Q) The number of data is not particularly limited. For example, the sum S prev At the timing of calculating all the physical quantities y (p) and physical quantity y (Q) Alternatively, for example, the physical quantity y stored in the database in step S22 may be used. (p) and physical quantity y (Q) Among these, the physical quantity y for the most recent few tens of minutes or tens of seconds (p) and physical quantity y (Q) The sum S prev It may also be used to calculate

[0060] The sum S in step S6 prevIn the calculation process, the sum S is repeatedly calculated until it reaches a preset number of times. prev is calculated. The repeated sum S prev The time interval at which the latest physical quantity y is calculated is not particularly limited. For example, as the shortest time interval, (p) and physical quantity y (Q) Each time is stored in the database in step S22, the sum S prev If the time interval is longer than the shortest time interval, the sum S prev The time interval at which is calculated can be set arbitrarily.

[0061] The sum S in step S6 prev The sum S calculated repeatedly by the calculation process prev is stored in the database in step S23 of FIG. 2. The sum S prev Based on the sum S prev Average value of [Table 15] is calculated. [Table 16] is used to determine whether or not there is an abnormality in the physical quantity x newly measured by the plurality of sensors 1 to 4 in step S11, which will be described later.

[0062] 1.4.7 Sum S new Calculation process The control unit 12 calculates the sum S prev After the calculation process is completed, the sum S new The calculation process proceeds to step S7. new In the calculation process, similarly to steps S2 to S6 described above, the control unit 12 collects a plurality of physical quantities x measured by the plurality of sensors 1 to 4, converts each of the collected physical quantities x into a normalized physical quantity y, and calculates the sum S of the physical quantities y multiplied by a weight w. newThe above formula (1) is used to convert the measured physical quantity x into the normalized physical quantity y. new In the calculation of S7, any one of the above formulas (2) to (4) is used. new In the calculation process, the sum S is repeatedly calculated until it reaches a preset number of times. new is calculated. The sum of the repeated calculations S new is stored in the database in step S24.

[0063] Sumwa S new The physical quantity y used to calculate (p) and physical quantity y (Q) The number of data is not particularly limited, but the sum S prev In the calculation process, the sum S prev The physical quantity y used to calculate (p) and physical quantity y (Q) It is preferable that the number of data is the same as or close to the number of data.

[0064] Also, the sum S new The number of times and the time interval at which the calculation of the sum S in step S6 is repeated are not particularly limited. prev In the calculation process, the sum S prev It is preferable that the number of times and the time intervals at which the calculation of is repeated are the same as or approximate to the number of times and the time intervals at which the calculation of is repeated.

[0065] 1.4.8 Average [Table 17] Calculation process Average value of step S8 [Table 18] In the calculation process, the control unit 12 calculates the sum S new Average value of [Table 19] Calculate the average value of step S8. [Table 20] The sum S calculated in the calculation process new Average value of [Table 21] is stored in the database in step S25.

[0066] 1.4.9 Standard deviation σ new Calculation process Standard deviation σ of step S9 new In the calculation process, the control unit 12 calculates the sum S new Standard deviation σ new Calculate the standard deviation σ in step S9. new Standard deviation σ calculated in the calculation process new is stored in the database in step S26.

[0067] 1.4.10 Probability calculation process In the probability calculation process of step S10, the control unit 12 calculates the average value [Table 22] , standard deviation σ new and the sum S prev Average value of [Table 23] is applied to the cumulative distribution function. For example, the following formula (5) is used as a variable for the cumulative distribution function. The control unit 12 calculates the probability value of the t-distribution based on the value of t calculated by the following formula (5).

number

[0068] 1.4.11 Abnormality detection process In the abnormality determination process in step S11, the control unit 12 compares the probability value calculated in the probability calculation process in step S10 with a second threshold value set in advance in step S14. The second threshold value is, for example, 1.0×10 -4 ~1.0×10 -8 A low probability value within the range is set.

[0069] The control unit 12 determines whether the calculated probability value is equal to or less than the second threshold value. If it is determined that the calculated probability value is equal to or less than the second threshold value (YES), the control unit 12 transmits a signal to the alarm unit 13 to notify the occurrence of an abnormality. Thereafter, the control unit 12 acquires the updated physical quantity y from the database in step S22 and repeats the processing of steps S6 to S11. On the other hand, if it is determined that the calculated probability value is not equal to or less than the second threshold value (NO), the control unit 12 acquires the updated physical quantity y from the database in step S22 without transmitting a signal to the alarm unit 13 and repeats the processing of steps S6 to S11.

[0070] 1.4.12 Principle of abnormality detection The principle of anomaly discrimination in the above-mentioned anomaly discrimination program will be described with reference to FIG. 9. The mean value of a sample population sampled from a population of normal distribution N(μ,σ) with mean value μ and standard deviation σ is [Table 25] , standard deviation is σ samp Then, t calculated by the following equation (6) follows the t distribution shown in FIG.

number

[0071] Therefore, in the abnormality discrimination program of this embodiment, in steps S1 to S6 of FIG. 2, the sum S of the physical quantities y that form the population is calculated based on the physical quantity x in the normal state. prev Then, in step S7 of FIG. 2, the newly measured physical quantity x is converted into a normalized physical quantity y, and the sum S of the physical quantities y is calculated. new Calculate the sum S new is the sum S prev Although it is not a sample population sampled from prev 2, t is calculated, and the probability value of the t distribution is derived from the value of t. In step S11, if the probability value is equal to or less than the second threshold, it is determined that an event with a low probability of occurring under normal circumstances, i.e., an abnormality, has occurred.

[0072] 1.4.13 Changing the weight w setting The value of the weight w preset in step S13 of FIG. 2 may be changed based on the influence of a disturbance on at least one of the plurality of sensors 1 to 4.

[0073] For example, if at least one of the multiple sensors 1 to 4 is affected by the temperature caused by the air conditioner 100 shown in FIG. 1, the value of the weight w applied to at least one of the multiple sensors 1 to 4 is changed based on information related to the operation of the air conditioner 100. Information related to the operation of the air conditioner 100 includes, for example, information on whether the air conditioner 100 is ON / OFF, information on the set temperature of the air conditioner 100, etc. This information is provided to the control unit 12, which then executes the setting change process of step S15. For example, if the sensor 3 shown in FIG. 1 is affected by the temperature caused by the air conditioner 100, the control unit 12 changes the value of the weight w set corresponding to the sensor 3 to "0" as shown in Table 2 below. As a result, the sum S using the above equations (2) to (4) is prev and the sum S new The measurement result of sensor 3 is excluded from the calculation. [Table 27]

[0074] For example, if at least one of the multiple sensors 1 to 4 is affected by the temperature due to sunlight, the value of the weight w applied to at least one of the multiple sensors may be set or changed based on at least one of the time of sunlight, the amount of sunlight, and the temperature of the ceiling.

[0075] 1.5 Effects According to the anomaly detection program of this embodiment and the fire monitoring system equipped with the same, by combining the physical quantities x measured by the multiple sensors 1 to 4 and applying a weight w that takes into account the interrelationships between the multiple sensors 1 to 4, it becomes possible to quickly detect the occurrence of an anomaly that is unlikely to occur under normal circumstances.

[0076] In addition, the sum S is used to determine whether an abnormality has occurred. previs calculated based on the physical quantity x measured by the multiple sensors 1 to 4, there is no need to perform deep learning on the neural network. Furthermore, the calculation processing executed in steps S3 to S11 in Fig. 2 makes it possible to determine whether an abnormality has occurred, and the calculation load for the determination is extremely small.

[0077] Furthermore, in the anomaly discrimination program and the fire monitoring system including the same according to this embodiment, a random number value with n decimal places is added to the physical quantity x in the normal state (step S4 in FIG. 2). This random number value is calculated in accordance with a first condition that the average value of the calculated random number values ​​is 0 and a second condition that a set value α is used to determine the range from the upper limit to the lower limit of the calculated random number value (step S3 in FIG. 2). Therefore, by increasing or decreasing the set value α according to the conditions of the monitoring area, the range from the upper limit to the lower limit of the random number value added to the physical quantity in the normal state can be varied, thereby increasing or decreasing the variation in the value of the physical quantity x in the normal state. This enables the control unit 12 to adjust the sensitivity with which the control unit 12 determines the occurrence of an abnormality. That is, by increasing the variation in the value of the physical quantity x in the normal state, the control unit 12 can lower the sensitivity with which the control unit 12 determines the occurrence of an abnormality. On the other hand, by decreasing the variation in the value of the physical quantity x in the normal state, the control unit 12 can increase the sensitivity with which the control unit 12 determines the occurrence of an abnormality.

[0078] 2. Simulation of anomaly detection program A simulation of the anomaly discrimination program of the present invention will be described below with reference to Figures 10 to 16. The simulation described below will clarify the influence of a random number value added to the physical quantity x on the sensitivity for discriminating the occurrence of an anomaly.

[0079] 2.1 Conditions for the monitoring area First, we set up a monitoring area with the conditions shown in Figure 10(a) and (b). The monitoring area had a ceiling and a floor, with no walls between them. Four sensors 1 to 4 were installed on the ceiling, and one fire source was installed on the floor.

[0080] The ceiling is 14.5m x 14.5m = 210.25m 2 The fire source was a square and was suspended at a height of H = 8 m from the floor. A 1.8 m x 1.8 m six-tiered crib made of evenly spaced wooden pieces was used as the fire source. The fire source was installed directly below the center of the ceiling on the floor. The four sensors 1 to 4 were K-type thermocouples with a glass coating and an outer diameter of φ0.6 mm. The four sensors 1 to 4 protruded 50 mm below the ceiling surface and were positioned at equal intervals, 4 m horizontally from the center of the fire source (which was also the center of the ceiling).

[0081] 2.2 Temperature measurement using four sensors Temperature measurements using the four sensors 1 to 4 began 60 seconds before the fire source ignited. The four sensors 1 to 4 measured the temperature near the ceiling at 1-second intervals. The fire source ignited 60 seconds after temperature measurements began. The time when the fire source ignited was set to 0 seconds, and the temperature was measured continuously until 600 seconds had passed. The temperature measurements from each of the four sensors 1 to 4 are shown in Figures 11(a) and (b) and 12(a) and (b).

[0082] 2.3 Simulation of anomaly detection program Using the temperature measurements of the four sensors 1 to 4 shown in FIGS. 11(a), 11(b), and 12(a), 12(b), as the physical quantity x, eight simulations were performed for steps S2 to S11, S12 to S14, and S21 to S26 of the anomaly discrimination program shown in FIG. 2. In each of the eight simulations, the calculation conditions for the random number values ​​in the random number calculation process in step S3 were all different. The calculation conditions for the random number values ​​were four normal distributions shown in FIGS. 4(a), 4(b), and 5(a), 5(b), and four uniform distributions shown in FIGS. 6(a), 6(b), and 7(a), 7(b). Random number values ​​were calculated according to each of these normal distributions and uniform distributions, and the effects of the random number values ​​added to the physical quantity x on the probability calculation process in step S10 and the anomaly discrimination process in step S11 were confirmed. The results of the eight simulations are shown in FIGS. 13(a), 13(b) to 16(a), 16(b).

[0083] In the eight simulations shown in Figures 13(a), (b) to 16(a), (b), the sum S was calculated based on the temperature measurements (physical quantity x) measured by the four sensors 1 to 4 for 60 seconds before the fire source ignited. prev Average value of [Table 28] The sum S prev Average value of [Table 29] After calculating the sum S, which is the target for determining whether an abnormality has occurred. new Average value of [Table 30] The time lag until the calculation of the probability is set to 180 seconds. The second threshold value for determining whether the probability calculated in the probability calculation process in step S10 is abnormal in the abnormality determination process in step S11 is 1.0×10 -6 was set to.

[0084] 2.3.1 Normal distribution random numbers (mean 0, standard deviation α=0.5) Fig. 13(a) is a graph showing the transition of the probability calculated in the probability calculation process of step S10 when a random number value of a normal distribution with a mean value of 0 and a standard deviation α of 0.5 is added to the physical quantity x. In the simulation shown in Fig. 13(a), the calculated probability at 126 seconds after the fire source was ignited was 1.0 × 10 -6 The average temperature rise when the occurrence of an abnormality was detected was 1.1°C (see the circled area in Figure 13(a)).

[0085] 2.3.2 Normal distribution random numbers (mean 0, standard deviation α=1.0) 13(b) is a graph showing the transition of the probability calculated in the probability calculation process of step S10 when a random number value of a normal distribution with a mean value of 0 and a standard deviation α of 1.0 is added to the physical quantity x. As shown in FIG. 13(b), the calculated probability at 137 seconds after the fire source ignited was 1.0×10 -6 The average temperature rise when the occurrence of an abnormality was detected was 1.6°C (see the circled area in Figure 13(b)).

[0086] 2.3.3 Normal distribution random numbers (mean 0, standard deviation α=1.5) 14(a) is a graph showing the transition of the probability calculated in the probability calculation process of step S10 when a random number value of a normal distribution with a mean value of 0 and a standard deviation α of 1.5 is added to the physical quantity x. As shown in FIG. 14(a), the calculated probability at 236 seconds after the fire source ignited was 1.0×10 -6 The average temperature rise when the occurrence of an abnormality was detected was 7.0°C (see the circled area in Figure 14(a)).

[0087] 2.3.4 Normal distribution random numbers (mean 0, standard deviation α=2.0) Fig. 14(b) is a graph showing the transition of the probability calculated in the probability calculation process of step S10 when a random value of a normal distribution with a mean value of 0 and a standard deviation α of 2.0 is added to the physical quantity x. As shown in Fig. 14(b), the calculated probability at 389 seconds after the fire source ignited was 1.0 × 10 -6 The average temperature rise when the occurrence of an abnormality was detected was 24.8°C (see the circled area in Figure 14(b)).

[0088] 2.3.5 Summary of simulation results using normally distributed random values The simulation results shown in FIGS. 13(a), (b) and 14(a), (b) are summarized in Table 3 below. [Table 31]

[0089] In the simulations shown in Figures 13(a) and 13(b) and Figures 14(a) and 14(b), random values ​​were calculated according to a normal distribution with a mean value of 0 and a standard deviation of α. These random values ​​were added to the temperature measurements (physical quantity x) measured during the 60 seconds before the fire source was ignited to form a population of temperature measurements taken under normal conditions. The standard deviation α determines the range from the upper limit to the lower limit of the random value. Therefore, the smaller the standard deviation α, the smaller the variance in the temperature measurements forming the population under normal conditions, resulting in a higher sensitivity for detecting abnormalities. On the other hand, the larger the standard deviation α, the greater the variance in the temperature measurements forming the population under normal conditions, resulting in a lower sensitivity for detecting abnormalities. As shown in Table 3 above, the smaller the standard deviation α, the shorter the time and the lower the average temperature rise required to detect an abnormality. On the other hand, the larger the standard deviation α, the longer the time and the higher the average temperature rise required to detect an abnormality. In other words, by arbitrarily setting the standard deviation α according to the conditions of the monitored area, it is possible to adjust the sensitivity for detecting abnormalities. For example, when monitoring a room at night when no one is coming or going, the value of the standard deviation α is reduced to increase the sensitivity for determining whether an abnormality has occurred. On the other hand, when monitoring a room during the day when there is a lot of people coming and going, the value of the standard deviation α is increased to decrease the sensitivity for determining whether an abnormality has occurred. This allows the sensitivity to be adjusted so that disturbances such as a temperature rise due to a person's body temperature or a temperature drop due to the inflow of cold air from outside are not determined to be an abnormality.

[0090] 2.3.6 Uniformly distributed random values ​​(mean value 0, range α = ±0.5) 15(a) is a graph showing the transition of the probability calculated in the probability calculation process of step S10 when a random number value from a uniform distribution with an average value of 0 and a range α from the upper limit to the lower limit set to ±0.5 is added to the physical quantity x. In the simulation shown in FIG. 15(a), the calculated probability at 125 seconds after the fire source was ignited was 1.0×10 -6The average temperature rise when the occurrence of an abnormality was detected was 1.0°C (see the circled area in Figure 15(a)).

[0091] 2.3.7 Uniformly distributed random values ​​(mean value 0, range α = ±1.0) Fig. 15(b) is a graph showing the transition of the probability calculated in the probability calculation process of step S10 when a random number value from a uniform distribution, with an average value of 0 and a range α from the upper limit to the lower limit set to ±1.0, is added to the physical quantity x. In the simulation shown in Fig. 15(b), the calculated probability at 177 seconds after the fire source was ignited was 1.0 × 10 -6 The average temperature rise when the occurrence of an abnormality was detected was 3.7°C (see the circled area in Figure 15(b)).

[0092] 2.3.8 Uniformly distributed random values ​​(mean value 0, range α = ±1.5) Fig. 16(a) is a graph showing the transition of the probability calculated in the probability calculation process of step S10 when a random value of a uniform distribution with an average value of 0 and a range α from the upper limit to the lower limit of ±1.5 is added to the physical quantity x. In the simulation shown in Fig. 16(a), the calculated probability at 299 seconds after the fire source was ignited was 1.0 × 10 -6 The average temperature rise when the occurrence of an abnormality was detected was 13.0°C (see the circled area in Figure 16(a)).

[0093] 2.3.9 Uniformly distributed random values ​​(mean value 0, range α = ±2.0) Fig. 16(b) is a graph showing the transition of the probability calculated in the probability calculation process of step S10 when a random number value from a uniform distribution with an average value of 0 and a range α from the upper limit to the lower limit set to ±2.0 is added to the physical quantity x. In the simulation shown in Fig. 16(b), the calculated probability at 301 seconds after the fire source was ignited was 1.0 × 10 -6 The average temperature rise when the occurrence of an abnormality was detected was 13.3°C (see the circled area in Figure 16(b)).

[0094] 2.3.10 Summary of simulation results using uniformly distributed random values The simulation results shown in FIGS. 15(a), (b) and 16(a), (b) are summarized in Table 4 below. [Table 32]

[0095] In the simulations shown in Figures 15(a) and 15(b) and Figures 16(a) and 16(b), random values ​​were calculated according to a uniform distribution with a mean value of 0 and a range α between the upper and lower limits of the calculated random values. These random values ​​were added to the temperature measurements (physical quantity x) measured during the 60 seconds before the fire source ignited to form a population of temperature measurements measured under normal conditions. The smaller the range α, the smaller the variance in the temperature measurements forming the population under normal conditions, resulting in a higher sensitivity for detecting abnormalities. On the other hand, the larger the range α, the greater the variance in the temperature measurements forming the population under normal conditions, resulting in a lower sensitivity for detecting abnormalities. As shown in Table 4 above, the smaller the range α, the shorter the time and the lower the average temperature rise required to detect an abnormality. On the other hand, the larger the range α, the longer the time and the higher the average temperature rise required to detect an abnormality. In other words, by arbitrarily setting the range α according to the conditions of the monitored area, it is possible to adjust the sensitivity for detecting abnormalities.

[0096] 3. Second embodiment of fire monitoring system Fig. 17 shows a fire monitoring system according to a second embodiment of the present invention. The fire monitoring system according to this embodiment is mainly composed of a plurality of sensors 1 to 4 installed in a plurality of monitoring areas, at least one repeater 20, and a receiver 10. The repeater 20 is electrically connected between the sensors 1 to 4 installed in any one of the monitoring areas and the receiver 10 via a signal line 5. The repeater 20 includes a receiving unit 11 and a control unit 12 similar to those in the first embodiment, and has installed therein an abnormality detection program similar to that in the first embodiment. On the other hand, the receiver 10 according to this embodiment is an existing product that is applied as is, and no abnormality detection program is installed in the receiver 10.

[0097] The receiver 11 of the repeater 20 is electrically connected to the plurality of sensors 1 to 4 via a signal line 5. The receiver 11 receives digital signals output from the plurality of sensors 1 to 4. These digital signals include physical quantities x measured by each of the plurality of sensors 1 to 4. The receiver 11 transmits the received digital signals to the controller 12. The controller 12 includes a processor for executing arithmetic processing.

[0098] First, the processor of the control unit 12 transmits digital signals including the physical quantity x received from the multiple sensors 1 to 4 to the receiver 10 and / or other repeaters, other receivers, fire extinguishing equipment, etc. For example, the receiver 10 executes a fire determination process similar to step S1 shown in Fig. 2 based on the digital signals including the physical quantity x to determine whether or not a fire has occurred. If the receiver 10 determines that a fire has occurred, it causes the alarm unit 13 to notify the occurrence of a fire.

[0099] Second, the processor of the control unit 12 executes the processes of steps S2 to S11, S12 to S15, and S21 to S26 shown in Fig. 2 according to the same abnormality determination program as in the first embodiment. If the control unit 12 determines in step S11 that an abnormality has occurred, it transmits an abnormality detection signal 5a to the receiver 10 and / or another repeater, another receiver, or fire extinguishing equipment. For example, the receiver 10 causes the notification unit 13 to notify the occurrence of an abnormality based on the abnormality detection signal 5a.

[0100] In the fire monitoring system of the present embodiment described above, an abnormality detection program is installed in the repeater 20. Therefore, by adding the repeater 20 to an existing fire monitoring system installed in, for example, an office building, a tenant building, a logistics warehouse, etc., it becomes possible to detect the occurrence of an abnormality in a specified monitoring area. In other words, the fire monitoring system of the present invention can be constructed using the existing sensors 1 to 4, signal line 5, and receiver 10.

[0101] The repeater 20 can be selectively applied to multiple monitoring areas, and can detect the occurrence of an abnormality in any monitoring area selected by the user. Furthermore, the repeater 20 can be easily installed by connecting signal lines 5 to the input and output terminals. In addition, by installing an abnormality detection program in the repeater 20, the calculation load on the receiver 10 can be reduced, and the amount of information stored in memory can be reduced.

[0102] 4. Third embodiment of fire monitoring system A third embodiment of a fire monitoring system including an abnormality determination program of the present invention will be described with reference to FIGS.

[0103] FIG. 18 shows a receiver 10 constituting a fire monitoring system of a third embodiment. As with the first embodiment shown in FIG. 2, the receiver 10 includes a receiving unit 11, a control unit 12, and an alarm unit 13. In FIG. 18, steps S1 to S11 are the processing of a main routine of an abnormality discrimination program. Furthermore, steps S12 to S16 and S21 to S26 are the processing of subroutines accompanying the main routine. Such an abnormality discrimination program is executed by the processor of the control unit 12. The fire monitoring system of the third embodiment is characterized in that the discrimination delay time of step S16 is used in the abnormality discrimination processing of step S11.

[0104] In step S16, a predetermined determination delay time for determining whether or not an abnormality has occurred is preset in the abnormality determination program. The length of the determination delay time is set to a value of 1 second or more depending on the temperature change in the monitored area under normal conditions. For example, the processor of the control unit 12 can automatically set an optimal determination delay time by learning the temperature change in a specific monitored area under normal conditions. A method in which the processor of the control unit 12 automatically sets the determination delay time by learning the temperature change under normal conditions will be described in detail later. The length of the determination delay time is not particularly limited, but is preferably selected from the range of 15 to 150 seconds, for example.

[0105] In the abnormality determination process in step S11, the processor of the control unit 12 compares the probability value calculated in the probability calculation process in step S10 with a second threshold value set in advance in step S14. The probability value calculated in step S10 indicates the probability of occurrence of a physical quantity x newly measured by a plurality of sensors. The second threshold value is, for example, 1.0×10 -4 ~1.0×10 -8 A low probability value within the range is set.

[0106] In the abnormality determination process of step S11, the processor of the control unit 12 determines whether the probability value calculated in the probability calculation process of step S10 is equal to or less than a second threshold. If the calculated probability value is equal to or less than the second threshold (YES), the processor of the control unit 12 starts counting a predetermined time (15 to 150 seconds) in step S16. Thereafter, the processor of the control unit 12 repeatedly determines whether the probability value calculated in the probability calculation process of step S10 is equal to or less than the second threshold. Then, if probability values ​​equal to or less than the second threshold are calculated consecutively before the determination delay time is reached (YES), the processor of the control unit 12 determines that an abnormality has occurred. On the other hand, if a probability value exceeding the second threshold is calculated before the determination delay time is reached (NO), the processor of the control unit 12 resets the count of the determination time (15 to 150 seconds) in step S16.

[0107] The fire monitoring system of the third embodiment allows the processor of the control unit 12 to adjust the sensitivity with which it determines whether an abnormality has occurred. Specifically, in the fire monitoring system of the first embodiment described above, the processor of the control unit 12 determines whether an abnormality has occurred when the condition that the calculated probability value is equal to or less than the second threshold is satisfied. On the other hand, in the fire monitoring system of the third embodiment, the processor of the control unit 12 determines whether an abnormality has occurred when probability values ​​equal to or less than the second threshold are consecutively calculated within the determination delay time. Specifically, in the fire monitoring system of the third embodiment, an abnormality is determined when a physical quantity x with a low probability of occurrence continues until the determination delay time is reached. Conversely, an abnormality is not determined when the physical quantity x does not continue until the determination delay time is reached. Therefore, the fire monitoring system of the third embodiment allows the processor of the control unit 12 to adjust the sensitivity with which it determines whether an abnormality has occurred by changing the length of the determination delay time in step S16. Furthermore, the fire monitoring system of the third embodiment prevents erroneous determination of an abnormality even when a probability value equal to or less than the second threshold is calculated within a very short time under normal conditions.

[0108] 4.1 First simulation of anomaly detection program using a warehouse model Next, the first simulation of the abnormality determination program shown in FIG. 18 will be described with reference to FIGS.

[0109] With the development of logistics networks in recent years, warehouses are becoming larger in size. In a 2022 survey, 82% of logistics companies in Japan responded that they plan to expand the total floor space of their warehouses in the future. It is easy to imagine that if a fire breaks out in a large warehouse and spreads, it could cause extensive damage. In large warehouses, early detection of fires is important to minimize damage. The first simulation described below verifies the operation of the anomaly detection program shown in Figure 18 when a fire breaks out in a large warehouse.

[0110] 4.1.1 Conditions for the monitoring area 19(a) and (b) show the conditions of the monitoring area used in the first simulation. The monitoring area in the first simulation is a floor area of ​​326 m 2 This is a warehouse model. As shown in Figure 19(a), the ceiling of the warehouse model is composed of a deck plate 30 with a width W = 11.0 m and a length = 29.6 m. Beneath the ceiling surface of the deck plate 30, beams are provided, consisting of large and small H-shaped steel beams 41 and 42 combined vertically and horizontally. The large and small H-shaped steel beams 41 and 42 divide the ceiling surface of the deck plate 30 into 20 beam spaces arranged in 2 columns and 10 rows. Each beam space is rectangular in plan view. A sensor is located in the center of each of the 20 beam spaces arranged in 2 columns and 10 rows. The 10 sensors arranged in the first beam space are referred to as "sensors A1 to A10," and the 10 sensors arranged in the second beam space are referred to as "sensors B1 to B10." Sensors A1 to A10 and B1 to B10 are analog fire detectors equipped with detection elements for measuring temperature (physical quantity x).

[0111] As shown in Figure 19(b), the deck plate 30 is made of a metal plate with a height H1 = 0.15 m that is bent to create a repeated uneven pattern. Such a deck plate 30 is placed at a height H2 = 6.3 m from the floor. Meanwhile, 20 sensors A1 to A10 and B1 to B10 are all suspended at a height H3 = 0.25 m from the ceiling surface of the deck plate 30. A 1.8 m x 1.8 m six-tier crib C made of evenly spaced wooden pieces was used as the fire source. As shown in Figure 19(a), the six-tier crib C was placed on the floor so that the linear distances from sensors A2, A3, B2, and B3 were equal.

[0112] 4.1.2 Changes in heat release rate of fire sources FIG. 20(a) is a graph showing the change in the heat release rate (kW) of the 6-stage crib C. The data on the change in the heat release rate (kW) of the 6-stage crib C is obtained by calculating the actual measured values ​​of the thermogravimetric analysis of the 6-stage crib C by αt 2 It is obtained by fitting to the model. 2 The model is Q=αt 2It is calculated by the formula below, where Q is the heat release rate (kW), α is the fire growth rate, and t is time (seconds).

[0113] 4.1.3 Temperature output value of each sensor The spatial distribution and temporal changes in temperature within the warehouse model in the event of a fire in six-tiered crib C are analyzed using the Fire Dynamics Simulator (FDS). The FDS is a fire dynamics simulator developed by NIST (National Institute of Standards and Technology). By inputting the above-mentioned warehouse model conditions and data on the change in heat release rate (kW) of six-tiered crib C into the FDS, the temperature outputs of 20 sensors A1-A10 and B1-B10 are obtained. The FDS divides the space within the warehouse model into a three-dimensional mesh and analyzes changes in temperature and airflow for each mesh. Areas with drastic changes in temperature and airflow are finely divided into smaller meshes. On the other hand, areas with gentle changes in temperature and airflow are roughly divided into larger meshes. Each mesh is a cube with sides measuring 5 cm to 20 cm. A total of 3.88 million meshes are placed within the warehouse model, with the areas near six-tiered crib C and the ceiling being particularly finely divided into smaller meshes. Using FDS, the temperature outputs of the 20 sensors A1 to A10 and B1 to B10 were obtained for 40 seconds before the sixth-stage crib C was ignited and for 300 seconds after the sixth-stage crib C was ignited.

[0114] 4.1.4 Operation of the abnormality detection program Based on the temperature outputs of the 20 sensors A1 to A10 and B1 to B10 obtained by the FDS analysis for a total of 340 seconds, the abnormality determination program shown in Fig. 18 was run. The processor of the control unit 12 executes steps S1 to S6 in Fig. 18 based on the temperature (physical quantity x) for 40 seconds before the sixth-stage crib C ignites, and calculates the sum S of the physical quantities y multiplied by the weight w. prev is calculated by the following formula (7).

number

[0115] The sum S in step S6 prev The sum S calculated repeatedly by the calculation process prev is stored in the database in step S23 in Fig. 18. The processor of the control unit 12 calculates the sum S prev Based on the sum S prev Average value of [Table 33] Calculate the average value [Table 34] is used in step S11 in FIG. 18 to determine whether there is an abnormality in the temperature (physical quantity x) after the sixth-stage crib C is ignited.

[0116] Next, the processor of the control unit 12 executes step S7 in FIG. 18 based on the temperature (physical quantity x) after the sixth-stage crib C is ignited, and calculates the sum S of the physical quantities y multiplied by the weight w. new is calculated by the above formula (7). new The sum S calculated repeatedly by the calculation process new is stored in the database in step S24 in Fig. 18. The processor of the control unit 12 executes step S8 in Fig. 18 and stores the sum S new Based on the sum S new Average value of [Table 35] Calculate the average value [Table 36] is stored in the database in step S25.

[0117] Next, the processor of the control unit 12 executes step S9 in FIG. 18 to obtain the sum S new Standard deviation σ new Calculate the standard deviation σ in step S9. new Standard deviation σ calculated in the calculation process new is stored in the database in step S26.

[0118] Next, the processor of the control unit 12 executes step S10 in FIG. 18 to calculate the average value [Table 37] , standard deviation σ new and the sum S prev Average value of [Table 38] is applied to the cumulative distribution function. For example, the following formula (8) is used as a variable for the cumulative distribution function. The processor of the control unit 12 calculates the probability value of the t-distribution based on the value of t calculated by the following formula (8).

number

[0119] Next, the processor of the control unit 12 executes step S11 in Fig. 18. In the abnormality determination process of step S11, the processor of the control unit 12 compares the probability value calculated in the probability calculation process of step S10 with a second threshold value set in advance in step S14. In the first simulation, the second threshold value is set to 1.0 x 10 -6 is set.

[0120] If the calculated probability value is equal to or less than the second threshold value (YES), the processor of the control unit 12 starts counting the discrimination delay time in step S16. In the first simulation, the discrimination delay time is set to, for example, 60 seconds. Thereafter, the processor of the control unit 12 repeatedly determines whether the probability value calculated in the probability calculation process of step S10 is equal to or less than the second threshold value. Then, if probability values ​​equal to or less than the second threshold value are calculated consecutively before the discrimination delay time reaches 60 seconds (YES), the processor of the control unit 12 determines that an abnormality has occurred.

[0121] 20(b) is a graph showing the transition of the probability value calculated in the probability calculation process of step S10. The dashed line in the figure indicates the second threshold value of 1.0×10 -6 20(b), the calculated probability value becomes equal to or less than the second threshold value approximately 50 seconds after the 6-stage crib C ignites. At this point, the processor of the control unit 12 starts counting the 60-second discrimination delay time. Thereafter, probability values ​​equal to or less than the second threshold value are continuously calculated until the discrimination delay time reaches 60 seconds. The processor of the control unit 12 determines that an abnormality has occurred when the discrimination delay time reaches 60 seconds. In other words, according to the abnormality discrimination program shown in FIG. 18, it is possible to notify the occurrence of an abnormality approximately 110 seconds after the start of a fire.

[0122] 4.2 Second simulation of the anomaly detection program using the warehouse model Next, the second simulation of the abnormality determination program shown in FIG. 18 will be described with reference to FIG.

[0123] Large warehouses are susceptible to the effects of outside temperatures and can get very cold in the winter. Furthermore, the vastness of the space makes it difficult to adjust the room temperature using an air conditioner. For this reason, large warehouses are typically heated in the winter using large commercial heaters. Large commercial heaters have extremely high heating capacity and begin to produce a heating effect within a short time after being turned on. When the temperature (physical quantity x) inside a large warehouse rises in a relatively short time due to a large commercial heater, conventional anomaly detection programs have the problem of instantly determining that an anomaly has occurred. The second simulation described below verifies the operation of the anomaly detection program shown in Figure 18 when a large commercial heater is used in a large warehouse.

[0124] 4.2.1 Conditions for the monitoring area Figure 21(a) shows the conditions of the monitoring area used in the second simulation. The monitoring area in the second simulation uses the same warehouse model as in the first simulation shown in Figures 19(a) and (b). However, the fire source installed in the warehouse model is changed from the six-tiered crib C to two jet heaters J.

[0125] Jet Heater J is a large commercial heater that burns fuel such as kerosene and blows out heated air with a fan. The two Jet Heaters J are manufactured by Orion Machinery Co., Ltd. under the product name "Jet Heater HPE370". The specifications of the two Jet Heaters J are as follows: The combustion method is a high-pressure spray type. The heat output can be switched between "strong" and "weak". The heat output at "strong" is 43kW. The heat output at "weak" is 32kW. The hot air blowout volume is 18m 3 / min. The duct diameter is φ278mm (=0.6m 2 ) The fuel used was kerosene (JIS No. 1 kerosene). As shown in Figure 19(a), each of the two jet heaters J was installed 5 m away from the left and right walls of the warehouse model. Assuming a severe cold season, the initial temperature of each of the two jet heaters J was set to 5°C.

[0126] 4.2.2 Changes in heat release rate of fire sources Unlike the first simulation described above, the two jet heaters J are used with a thermal output of either 43 kW or 32 kW. Therefore, the heat generation rate (kW) data for the two jet heaters J was set to a constant value of 43 kW.

[0127] 4.2.3 Temperature output value of each sensor As with the first simulation described above, the spatial distribution and temporal changes in temperature within the warehouse model when each of the two jet heaters J is activated are analyzed using FDS. By inputting the warehouse model conditions described above and data on changes in the heat generation rate (kW) of the two jet heaters J into FDS, the temperature outputs of each of the 20 sensors A1-A10 and B1-B10 are obtained. A total of 3.99 million meshes are placed within the warehouse model, and these are finely divided by small meshes, particularly around the path of the hot air blown out from the jet heater J and the ceiling surface. Using FDS, the temperature outputs of each of the 20 sensors A1-A10 and B1-B10 are obtained for 40 seconds before the two jet heaters J are ignited and for 300 seconds after the two jet heaters J are ignited.

[0128] 4.2.4 Operation of the abnormality detection program As in the first simulation described above, the abnormality determination program shown in Fig. 18 was run based on the temperature outputs of the 20 sensors A1 to A10 and B1 to B10 obtained by FDS analysis for a total of 340 seconds. The processor of the control unit 12 executes steps S1 to S10 in Fig. 18 and calculates the probability value of the t-distribution based on the value of t calculated by the above equation (8).

[0129] 21(b) is a graph showing the transition of the probability value calculated in the probability calculation process of step S10. The dashed line in the figure indicates the second threshold value of 1.0×10 -621(b), the calculated probability value becomes equal to or less than the second threshold value approximately 55 seconds after the two jet heaters J are ignited. At this point, the processor of the control unit 12 starts counting the 60-second discrimination delay time. Subsequently, a probability value exceeding the second threshold value is calculated before the discrimination delay time reaches 60 seconds. The processor of the control unit 12 resets the discrimination delay time count at the point at which the probability value exceeding the second threshold value is calculated. That is, the processor of the control unit 12 did not determine that the temperature change in the warehouse model due to the ignition of the two jet heaters J was abnormal. In other words, the processor of the control unit 12 determined that the temperature change in the warehouse model due to the ignition of the two jet heaters J was normal. Thus, according to the anomaly discrimination program shown in FIG. 18, even if a probability value equal to or less than the second threshold value is calculated within a very short period of time under normal circumstances, it is not erroneously determined that an anomaly has occurred. In other words, it is possible to prevent the anomaly discrimination program from misjudging or reporting an error due to the operation of an air conditioner or heater.

[0130] 4.3 Learning mode for setting the discrimination delay time Next, the learning mode of the abnormality discrimination program executed in the fire monitoring system of the third embodiment will be described with reference to Fig. 22. The learning mode of the abnormality discrimination program is a process for automatically setting the discrimination delay time in step S16 by having the processor of the control unit 12 learn temperature changes under normal conditions in a specific monitoring area.

[0131] Fig. 22 shows the flow of processing in the learning mode of the anomaly discrimination program. Steps S1 to S10, S12 to S15, and S21 to S26 in Fig. 22 are the same as the processing for anomaly discrimination shown in Fig. 18. In the learning mode, the processor of the control unit 12 executes steps S1 to S10 in Fig. 22 and calculates the probability value of the t-distribution based on the value of t calculated by the above equation (8).

[0132] 4.3.1 Low-probability discrimination processing In step S31, the processor of the control unit 12 compares the probability value calculated in the probability calculation process in step S10 with a second threshold value set in advance in step S14. For example, the second threshold value is set to 1.0×10 -6 If the calculated probability value is equal to or less than the second threshold value (YES), the processor of the control unit 12 executes a continuous time count process in step S32.

[0133] 4.3.2 Continuous Time Counting In step S32, the processor of the control unit 12 starts counting time. The counting of time in step S32 continues as long as probability values ​​equal to or less than the second threshold are continuously calculated in step S31. On the other hand, if a probability value exceeding the second threshold is calculated in step S31 (NO), the processor of the control unit 12 stores the time (continuous time CT) during which probability values ​​equal to or less than the second threshold are continuously calculated in the database in step S34. Thereafter, the processor of the control unit 12 resets the counting of time in step S32.

[0134] Steps S1 to S10 and S31 to S33 in the learning mode are repeatedly executed for a predetermined period of time. As a result, data of multiple continuous time CTs with different values ​​is accumulated in the database of step S34. The period during which the learning mode is executed is determined based on the magnitude of temperature change in the monitored area under normal conditions. For example, if the temperature change in the monitored area under normal conditions is small, the period during which the learning mode is executed can be several hours. On the other hand, if the temperature change in the monitored area under normal conditions is large, the learning mode needs to be executed for a period of several days or about one month.

[0135] 4.3.3 Determination delay time setting process If the learning mode has been executed for a predetermined period of time, the processor of the control unit 12 executes a discrimination delay time setting process in step S33. That is, the processor of the control unit 12 refers to the multiple values ​​of continuous time CT stored in the database in step S34 and identifies the longest value of continuous time CT. Next, the processor of the control unit 12 sets a time exceeding the identified value of continuous time CT as the discrimination delay time in step S16. For example, the processor of the control unit 12 sets a value obtained by adding several tens of seconds to the longest continuous time CT as the discrimination delay time in step S16. [Explanation of symbols]

[0136] 1~4 sensors 5 Signal line 10 Receivers 11 Receiving unit 12 Control unit (processor) 13. Information Department 20 Repeater 22 Control unit (processor) 5a Abnormality detection signal

Claims

1. An abnormality determination program for causing a processor to determine the occurrence of an abnormality that has a low probability of occurring under normal conditions based on a database of physical quantities x measured by a sensor under normal conditions, The processor can be caused to execute a calculation process including a random number calculation process for calculating a random number value for increasing the significant digits of a physical quantity x, and a random number addition process for adding the random number value to the physical quantity x, The random number calculation process causes the processor to calculate the random number in accordance with a first condition that the average value of the calculated random number is 0, and a second condition that a set value for determining the range from the upper limit to the lower limit of the calculated random number is α; an anomaly detection program that can increase or decrease the set value α to increase or decrease the variation in the value of the physical quantity x, thereby adjusting the sensitivity with which the processor detects the occurrence of an anomaly.

2. 2. The abnormality determination program according to claim 1, wherein the random number calculation process causes the processor to calculate the random number value according to a normal distribution whose mean value is 0 and whose standard deviation is the set value α.

3. 2. The abnormality detection program according to claim 1, wherein the random number calculation process causes the processor to calculate the random number value according to a uniform distribution in which the average value is 0 and the range from the upper limit to the lower limit of the calculated random number value is the set value α.

4. 2. The abnormality detection program according to claim 1, wherein the physical quantity x is any one of a temperature, a smoke density, and an infrared intensity measured by each of the plurality of sensors installed in a predetermined monitoring area.

5. a first process of collecting a plurality of physical quantities x measured by a plurality of sensors for a predetermined time or a predetermined number of times under normal conditions where no fire has occurred, and converting each of the collected physical quantities x into a normalized physical quantity y; A weight w corresponding to each of the plurality of sensors is set in advance, and the sum S of the physical quantities y multiplied by the weight w is prev A second process of calculating After the predetermined time has elapsed or the predetermined number of times has been exceeded, a plurality of physical quantities x measured by a plurality of the sensors are collected, each of the collected physical quantities x is converted into a normalized physical quantity y, and the sum S of the physical quantities y multiplied by a weight w is calculated. new A third process of calculating Sumwa S new Average value of Table 40 and standard deviation σ new A fourth process of calculating Average Table 41 , standard deviation σ new and the sum S prev Average value of Table 42 a fifth process of determining whether the probability calculated by applying the above formula to the cumulative distribution function is equal to or less than a preset probability threshold, and determining that an abnormality has occurred if the probability is equal to or less than the threshold; 5. The abnormality determination program according to claim 4, which causes the processor to execute the following:

6. a first process of collecting a plurality of physical quantities x measured by a plurality of sensors for a predetermined time or a predetermined number of times under normal conditions where no fire has occurred, and converting each of the collected physical quantities x into a normalized physical quantity y; A weight w corresponding to each of the plurality of sensors is set in advance, and the sum S of the physical quantities y multiplied by the weight w is prev A second process of calculating After the predetermined time has elapsed or the predetermined number of times has been exceeded, a plurality of physical quantities x measured by a plurality of the sensors are collected, each of the collected physical quantities x is converted into a normalized physical quantity y, and the sum S of the physical quantities y multiplied by a weight w is calculated. new A third process of calculating Sumwa S new Average value of Table 43 and standard deviation σ new A fourth process of calculating Average Table 44 , standard deviation σ new and the sum S prev Average value of Table 45 a fifth process of determining whether a probability calculated by applying the cumulative distribution function of the formula (I) to the cumulative distribution function is equal to or less than a predetermined probability threshold, and if the probability is equal to or less than the threshold, starting counting a predetermined length of discrimination delay time, and determining that an abnormality has occurred if probabilities equal to or less than the threshold are continuously calculated before the discrimination delay time is reached.

7. 7. The abnormality determination program according to claim 6, wherein the determination delay time is set within a range of 15 seconds to 150 seconds.

8. The processor is configured to execute a learning mode, which is a process for automatically setting the discrimination delay time, The learning mode includes: The same processes as the first to fourth processes, respectively; Average Table 46 , standard deviation σ new and the sum S prev Average value of Table 47 a low probability discrimination process for determining whether the probability calculated by applying the above formula to the cumulative distribution function is equal to or less than a preset probability threshold; a continuous time counting process that starts counting time when the probability calculated in the low probability discrimination process is equal to or less than the threshold, continues counting time until a probability exceeding the threshold is calculated, and, when a probability exceeding the threshold is calculated, stores the time during which a probability equal to or less than the threshold is continuously calculated in a database; a discrimination delay time setting process for identifying the longest time from among the plurality of times stored in the database, and setting a time exceeding the longest time as the discrimination delay time.

9. 9. The abnormality determination program according to claim 8, wherein a period for the processor to execute the learning mode is set, and the processor executes the determination delay time setting process after the set period has elapsed.

10. A monitoring system that operates in accordance with the abnormality determination program according to any one of claims 1 to 9, at least one of said sensors; the processor executing processing according to the abnormality determination program based on a signal indicating a physical quantity x output from the sensor; a notification unit that operates when the processor determines that an abnormality has occurred.

11. A fire monitoring system that operates according to the abnormality determination program according to any one of claims 4 to 9, a plurality of the sensors and receivers; the receiver includes a receiving unit, the processor, and a notification unit, and the abnormality determination program is installed therein; the receiving unit receives signals indicating physical quantities x output from the plurality of sensors; the processor executes processing in accordance with the abnormality determination program based on the physical quantity x; A fire monitoring system in which the alarm unit operates when the processor determines that an abnormality has occurred.

12. A fire monitoring system that operates according to the abnormality determination program according to any one of claims 4 to 9, a plurality of the sensors and repeaters; the repeater includes a receiving unit and the processor, and the abnormality determination program is installed therein; the receiving unit receives signals indicating physical quantities x output from the plurality of sensors; The processor executes processing in accordance with the abnormality determination program based on the physical quantity x, and outputs an abnormality detection signal when it determines that an abnormality has occurred.

Images