KR102375679B1 - Artificial intelligence based fire prediction method and device using pattern analysis - Google Patents

Abstract

The present invention relates to an artificial intelligence based fire prediction method using pattern analysis and a device thereof. The method comprises the steps of: collecting diagnostic data about an operating state of a facility device from a monitoring terminal installed in the facility device and including a plurality of sensors; deriving an operating pattern for each sensor, which is defined as an operating state for a specific time interval, based on the diagnostic data; generating a fire diagnosis model which predicts the risk of fire occurrence for the facility device through learning about the correlation between the operation patterns of each sensor; and diagnosing the fire occurrence in the facility device by applying the real-time diagnostic data collected from the facility device to the fire diagnosis model. Therefore, the present invention can predict the probability of fire occurrence.

Description

Artificial intelligence-based fire prediction method and device using pattern analysis

The present invention relates to an artificial intelligence-based fire prediction technology, and more particularly, an artificial intelligence-based fire prediction method using pattern analysis that predicts the occurrence of a fire by constantly monitoring the operation status of facility equipment and enables rapid response, and It's about the device.

Recently, ultra-high voltage and large-capacity of various power devices have been progressed due to the rapid development of the industry, and much effort is being made to improve the stability and reliability of the power system in order to supply high-quality power, and more advanced for diagnosis of power facilities Advanced scientific equipment is required.

Because a large amount of power is required to upgrade the equipment, the degree of heat generation varies depending on the amount of power, and in some cases, the temperature rises outside the normal operating range.

On the other hand, the temperature rise inside the facility may cause damage to other components, and as a result, may act as a cause to increase the risk of fire in the facility.

Accordingly, there is a need for a technology capable of diagnosing fire occurrence in advance and providing a notification for preventive measures through monitoring of environmental factors such as temperature and humidity inside the facility.

Korean Patent Publication No. 10-2010-0052437 (2010.05.19)

An embodiment of the present invention is to provide an artificial intelligence-based fire prediction method and apparatus using pattern analysis that predicts the occurrence of a fire by constantly monitoring the operation status of facility equipment and enables rapid response.

An embodiment of the present invention is an artificial intelligence-based fire using pattern analysis that can predict the probability of a fire by deriving a driving pattern of a facility device based on data collected from various sensors and analyzing the correlation between the driving patterns. An object of the present invention is to provide a prediction method and apparatus.

An embodiment of the present invention uses pattern analysis that can improve predictability of fire occurrence by deriving fire precursor symptom elements related to abnormalities in facility equipment, and patterning and learning the time change related to fire precursor symptom elements. To provide an artificial intelligence-based fire prediction method and device.

Among the embodiments, the artificial intelligence-based fire prediction method using pattern analysis includes collecting diagnostic data on the operating state of the facility device from a monitoring terminal installed in the facility device and including a plurality of sensors; A fire diagnosis model for predicting the risk of fire in relation to the facility equipment by deriving a driving pattern for each sensor defined as the operating state for a specific time interval based on the basis, and learning about the correlation between the driving patterns for each sensor and diagnosing the occurrence of a fire in the corresponding facility by applying real-time diagnostic data collected from the facility to the fire diagnosis model.

The monitoring terminal measures signals related to temperature, humidity, spark, smoke and gas in the course of operation of the facility device through the plurality of sensors, and collecting the diagnostic data includes electrical It may include the step of converting the signal into a digital signal by sampling (sampling) at a specific period.

The step of deriving the driving pattern is a step of deriving a two-dimensional pattern graph related to the diagnostic data for each sensor, aligning the two-dimensional pattern graph based on a time axis, and then each sensor based on a value axis Determining a pre-fire section that deviates from each operation range and determining the sections before and after the pre-fire section on the basis of the time axis, and operating each sensor for the change in the pattern graph from the previous section to the section after the corresponding section determining as a pattern.

The generating of the fire diagnosis model is a result of determining the representative characteristic value for the driving pattern for each sensor for the specific time interval and calculating the rate of change for the representative characteristic value between consecutive specific time intervals. It may include the step of performing the learning by generating a feature vector having a change rate as a component value as learning data.

The generating of the fire diagnosis model includes obtaining two-dimensional images related to the driving pattern for each sensor for the specific time interval, and an n-dimensional (where n is a natural number) feature vector for each of the two-dimensional images. generating, integrating the feature vectors for each sensor into one to generate a representative feature vector representing the specific time interval; and generating the fire diagnosis model by learning about the representative feature vector. there is.

The generating of the fire diagnosis model includes generating, as the fire diagnosis model, a fire risk function that outputs a predicted probability regarding the possibility of fire based on input data, and diagnosing the fire occurrence includes the steps of: The method may include applying the real-time diagnosis data to a fire risk function and diagnosing the occurrence of a fire in a corresponding facility device according to an output value of the fire diagnosis model obtained as a result of the application.

The step of diagnosing the occurrence of a fire may include providing an alarm regarding the occurrence of a fire when the output value of the fire diagnosis model for the specific time interval exceeds a preset threshold and is continuously repeated more than a preset number of times. can

Among the embodiments, the artificial intelligence-based fire prediction device using pattern analysis includes a diagnostic data collection unit that collects diagnostic data on operating conditions from a plurality of sensors installed in a facility device, and a specific information for each sensor based on the diagnostic data. A driving pattern derivation unit for deriving a driving pattern defined as a driving state for a time interval, and a fire diagnosis model generating a fire diagnosis model that predicts the occurrence of a fire in the facility equipment through learning about the correlation between the driving patterns and a fire occurrence diagnosis unit for diagnosing the occurrence of a fire in the corresponding facility by applying the real-time diagnosis data collected from the unit and the facility device to the fire diagnosis model.

The disclosed technology may have the following effects. However, this does not mean that a specific embodiment should include all of the following effects or only the following effects, so the scope of the disclosed technology should not be understood as being limited thereby.

Artificial intelligence-based fire prediction method and apparatus using pattern analysis according to an embodiment of the present invention derives operation patterns of facility equipment based on data collected from various sensors and analyzes correlations between operation patterns The probability of occurrence can be predicted.

An artificial intelligence-based fire prediction method and apparatus using pattern analysis according to an embodiment of the present invention derives a fire precursor symptom element related to an abnormality in a facility device, and pattern changes over time for the fire precursor symptom element by learning the pattern. It can improve the predictability of fire occurrence.

1 is a view for explaining a fire prediction system according to the present invention.
FIG. 2 is a block diagram illustrating a physical configuration of the fire prediction device of FIG. 1 .
3 is a block diagram illustrating a functional configuration of the fire prediction device of FIG. 1 .
4 is a flowchart illustrating an AI-based fire prediction process using pattern analysis performed by the fire prediction device of FIG. 1 .
FIG. 5 is a view for explaining an embodiment of an AI-based fire prediction process using pattern analysis performed by the fire prediction device of FIG. 1 .
FIG. 6 is a view for explaining a monitoring map for each type used in the fire prediction device of FIG. 1 .
FIG. 7 is a view for explaining a process of determining a region of interest performed by the fire prediction apparatus of FIG. 1 .

Since the description of the present invention is merely an embodiment for structural or functional description, the scope of the present invention should not be construed as being limited by the embodiment described in the text. That is, since the embodiment may have various changes and may have various forms, it should be understood that the scope of the present invention includes equivalents capable of realizing the technical idea. In addition, since the object or effect presented in the present invention does not mean that a specific embodiment should include all of them or only such effects, it should not be understood that the scope of the present invention is limited thereby.

On the other hand, the meaning of the terms described in the present application should be understood as follows.

Terms such as “first” and “second” are for distinguishing one component from another, and the scope of rights should not be limited by these terms. For example, a first component may be termed a second component, and similarly, a second component may also be termed a first component.

When a component is referred to as being “connected to” another component, it may be directly connected to the other component, but it should be understood that other components may exist in between. On the other hand, when it is mentioned that a certain element is "directly connected" to another element, it should be understood that the other element does not exist in the middle. Meanwhile, other expressions describing the relationship between elements, that is, "between" and "between" or "neighboring to" and "directly adjacent to", etc., should be interpreted similarly.

The singular expression is to be understood as including the plural expression unless the context clearly dictates otherwise, and terms such as "comprises" or "have" refer to the embodied feature, number, step, action, component, part or these It is intended to indicate that a combination exists, and it should be understood that it does not preclude the possibility of the existence or addition of one or more other features or numbers, steps, operations, components, parts, or combinations thereof.

In each step, identification numbers (eg, a, b, c, etc.) are used for convenience of description, and identification numbers do not describe the order of each step, and each step clearly indicates a specific order in context. Unless otherwise specified, it may occur in a different order from the specified order. That is, each step may occur in the same order as specified, may be performed substantially simultaneously, or may be performed in the reverse order.

The present invention can be embodied as computer-readable codes on a computer-readable recording medium, and the computer-readable recording medium includes all types of recording devices in which data readable by a computer system is stored. . Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like. In addition, the computer-readable recording medium is distributed in a computer system connected to a network, so that the computer-readable code can be stored and executed in a distributed manner.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs, unless otherwise defined. Terms defined in the dictionary should be interpreted as being consistent with the meaning of the context of the related art, and cannot be interpreted as having an ideal or excessively formal meaning unless explicitly defined in the present application.

Electrical equipment may correspond to machines, appliances, dams, waterways, reservoirs, electric wires, security communication lines and other facilities installed for power generation, transmission, substation, distribution, electricity supply or use of electricity. In addition, the operating facility may correspond to a facility that operates and manages electric facilities. For example, electric equipment includes electric power equipment such as water substation equipment, self-generation equipment, wiring equipment, and lighting equipment, emergency lighting equipment, emergency warning equipment, disaster prevention equipment such as guidance lamp equipment, telephone equipment, broadcasting equipment, and intercom equipment. It may include communication information facilities, such as elevator facilities, escalator facilities, transport facilities such as pipeline facilities.

On the other hand, facility equipment may correspond to a term including electrical equipment and operating equipment, but is not necessarily limited thereto, and may be installed and operated in various places and facilities, and may be used for early diagnosis of fire in order to prevent damage due to fire. It can be understood as devices to which monitoring technology for

1 is a view for explaining a fire prediction system according to the present invention.

Referring to FIG. 1 , the fire prediction system 100 may include a monitoring terminal 110 , a fire prediction device 130 , a database 150 , and a manager terminal 170 .

The monitoring terminal 110 may correspond to a device that is installed in a facility device and can measure information about the operating state of the facility. The monitoring terminal 110 may be implemented to operate in conjunction with an external system, and may include a communication module that provides a network communication function for this purpose. The monitoring terminal 110 may be connected to the fire prediction apparatus 130 through a network, and a plurality of monitoring terminals 110 may be simultaneously connected to the fire prediction apparatus 130 .

In one embodiment, the monitoring terminal 110 may be implemented to include a plurality of sensors for information collection. For example, the monitoring terminal 110 may basically include a temperature sensor, a humidity sensor, a spark detection sensor, a smoke detection sensor, and a gas detection sensor, but is not necessarily limited thereto, and a current sensor, a wattage sensor and Various sensors for detecting fire occurrence, such as a thermal image sensor, may be further included. That is, the monitoring terminal 110 may collect electrical signals related to temperature, humidity, sparks, smoke and gas in the process of operating the facility through a plurality of sensors. The monitoring terminal 110 may transmit the collected information to the fire prediction device 130 through a network in real time or periodically.

In one embodiment, the monitoring terminal 110 is a sensor module composed of a plurality of sensors, a data conversion module for converting a signal measured through a plurality of sensors, a communication module for processing communication with an external system, and the measured signal It may be implemented including a memory module for storing and a control module for controlling related operations. The monitoring terminal 110 may store a variety of information collected through the memory module, and for example, may store information about a sensor type, a sensor identification code, a range of a sensing signal, a sensor output method, and the like.

In an embodiment, the monitoring terminal 110 may be electrically connected to a fire extinguishing device installed in a facility device to collect information on a fire extinguishing configuration for fire extinguishing disaster prevention. For example, the fire extinguishing configuration may correspond to a fire extinguishing tube container containing a fire extinguishing material, and in this case, the monitoring terminal 110 may further measure a signal regarding the internal pressure of the fire extinguishing tube container.

On the other hand, the fire extinguishing device may correspond to a device that provides fire delay and suppression action by automatically ejecting the fire extinguishing material inside when it is installed inside the facility device and is exposed to nearby fire or heat. It may include a tube container for fire extinguishing, a fire monitor for monitoring the condition of the tube container for fire extinguishing, and a container fixing portion that provides a physical coupling between the tube container for fire extinguishing and the fire monitor.

The fire prediction device 130 may be implemented as a server corresponding to a computer or program capable of predicting the occurrence of a fire according to the operation pattern of the facility device using the diagnostic data received from the plurality of monitoring terminals 110 . The fire prediction device 130 may be wirelessly connected to the monitoring terminal 110 through Bluetooth, WiFi, a communication network, etc., and may exchange data with the monitoring terminal 110 through the network.

In an embodiment, the fire prediction device 130 may manage information for diagnosing fire occurrence in conjunction with the database 150 . Meanwhile, unlike FIG. 1 , the fire prediction device 130 may be implemented by including the database 150 therein. In addition, the fire prediction device 130 may be implemented including a processor, a memory, a user input/output unit, and a network input/output unit, which will be described in more detail with reference to FIG. 2 .

The database 150 may correspond to a storage device for storing various types of information required in an AI-based fire prediction process using pattern analysis. The database 150 may store diagnostic data received from each monitoring terminal 110 and may store information on a fire diagnosis model for predicting fire occurrence, but is not necessarily limited thereto, and the fire prediction device 130 may It is possible to store information collected or processed in various forms in the process of monitoring the operation of various facility devices in conjunction with the plurality of monitoring terminals 110, learning the fire diagnosis model, and predicting the occurrence of a fire through this.

The manager terminal 170 may correspond to a computing device that can check monitoring information for each facility device or check a fire diagnosis result, and may be implemented as a smart phone, a laptop computer, or a computer, but is not necessarily limited thereto, and a tablet It can also be implemented in various devices such as a PC. The manager terminal 170 may be connected to the fire prediction apparatus 130 through a network, and a plurality of manager terminals 170 may be simultaneously connected to the fire prediction apparatus 130 .

FIG. 2 is a block diagram illustrating a physical configuration of the fire prediction device of FIG. 1 .

Referring to FIG. 2 , the fire prediction device 130 may include a processor 210 , a memory 230 , a user input/output unit 250 , and a network input/output unit 270 .

The processor 210 may execute a procedure for processing each step in the process in which the fire prediction device 130 operates, and may manage the memory 230 that is read or written throughout the process, and the memory 230 ) can schedule the synchronization time between volatile and non-volatile memory in The processor 210 may control the overall operation of the fire prediction device 130, and is electrically connected to the memory 230, the user input/output unit 250, and the network input/output unit 270 to control the data flow between them. can The processor 210 may be implemented as a central processing unit (CPU) of the fire prediction device 130 .

The memory 230 is implemented as a non-volatile memory, such as a solid state drive (SSD) or a hard disk drive (HDD), and may include an auxiliary storage device used to store overall data required for the fire prediction device 130, It may include a main memory implemented as a volatile memory such as random access memory (RAM).

The user input/output unit 250 may include an environment for receiving a user input and an environment for outputting specific information to the user. For example, the user input/output unit 250 may include an input device including an adapter such as a touch pad, a touch screen, an on-screen keyboard, or a pointing device, and an output device including an adapter such as a monitor or a touch screen. In an embodiment, the user input/output unit 250 may correspond to a computing device connected through a remote connection, and in such a case, the fire prediction device 130 may be implemented as a server.

The network input/output unit 270 includes an environment for connecting with an external device or system through a network, for example, a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), and a VAN (Wide Area Network) (VAN). It may include an adapter for communication such as Value Added Network).

3 is a block diagram illustrating a functional configuration of the fire prediction device of FIG. 1 .

Referring to FIG. 3 , the fire prediction device 130 includes a diagnosis data receiving unit 310 , a driving pattern deriving unit 330 , a fire diagnosis model generating unit 350 , a fire occurrence diagnosis unit 370 , and a control unit 390 . may include

The diagnostic data receiving unit 310 may collect diagnostic data on the operating state of the facility from the monitoring terminal 110 installed in the facility and including a plurality of sensors. That is, the monitoring terminal 110 can collect various information that can determine whether the operation of the facility is normal through various sensors as diagnostic data, and provides the data to the fire prediction device 130 connected by wire or wirelessly. can provide

At this time, the diagnostic data receiving unit 310 of the fire prediction device 130 may take charge of the corresponding operation and process it, and the diagnostic data receiving unit 310 additionally performs a data classification operation or a pre-processing operation for analysis along with the data reception. can do. For example, the diagnostic data receiving unit 310 may perform a validity check on data, and may perform normalization or data conversion.

In an embodiment, the diagnostic data receiver 310 may convert an electrical signal among diagnostic data into a digital signal by sampling at a specific period. That is, the monitoring terminal 110 may measure a signal related to temperature, humidity, spark, smoke and gas in the operation process of the facility device through a plurality of sensors, and at this time, the measured signal is an analog signal as a sensor Depending on the characteristics of , it can be expressed as an unstructured eigenvalue.

Also, the diagnostic data receiving unit 310 may amplify the diagnostic data at a predetermined ratio. For example, the diagnostic data receiving unit 310 may process a conversion operation from the electrical signal generated by the sensor to a digitized signal through sampling at a fixed period, in which case the sampling rate is the fire prediction. It may be preset by the device 130 .

Also, if necessary, the diagnostic data receiver 310 may perform calibration after conversion of the digital signal. To this end, the diagnostic data receiver 310 may use a calibration function stored in the database 150 .

The driving pattern deriving unit 330 may derive a driving pattern for each sensor defined as a driving state for a specific time interval based on the diagnostic data. The driving pattern deriving unit 330 may determine a diagnosis cycle in consideration of the operating characteristics of the facility equipment, and may determine a specific time interval according to the diagnosis cycle. For example, when the diagnosis period is T, the driving pattern derivation unit 330 may determine T/m (in this case, m is a natural number) or m*T as a specific time interval. Also, the driving pattern deriving unit 330 may determine the driving pattern by expressing the driving state derived from the diagnostic data for each sensor in a visualized form.

In one embodiment, the driving pattern deriving unit 330 derives a two-dimensional pattern graph related to diagnostic data for each sensor, aligns the two-dimensional pattern graph based on the time axis, and then based on the value axis Determining the section before and after the fire warning section is determined based on the time axis, and changing the pattern graph from the previous section to the next section as the operating pattern for each sensor. can decide

More specifically, the driving pattern deriving unit 330 may derive a pattern graph as a result of visualizing the diagnostic data collected for each sensor on a two-dimensional coordinate system composed of a time axis and a value axis. That is, the pattern graph may correspond to a graphic interface that visualizes characteristics measured by a sensor in a two-dimensional coordinate system in the operation process of the facility device.

On the other hand, the driving pattern derivation unit 330 may define the signal range measured in a normal state for each sensor as the driving range, rather than simply arranging the collected data, and detect the moment when the collected diagnostic data is out of the driving range. It is possible to determine the defined fire precursor section including the corresponding section. The driving pattern derivation unit 330 may determine each of the front and rear sections having the same time interval as the fire warning section based on the time axis, and converts the change in the pattern graph from the previous section to the next section as the driving pattern for each sensor. can decide

The fire diagnosis model generation unit 350 may predict the risk of a fire with respect to the facility device through learning about the correlation between driving patterns for each sensor. The fire diagnosis model generating unit 350 may diagnose fire occurrence only with a driving pattern for a single sensor, and may diagnose fire by analyzing a correlation between driving patterns for a plurality of sensors. When diagnosing fire occurrence based on the correlation between driving patterns, the fire diagnosis model generating unit 350 extracts a feature value for each driving pattern, defines the correlation between the driving patterns with the corresponding feature values, and utilizes this A learning operation can be performed.

In an embodiment, the fire diagnosis model generation unit 350 determines a representative characteristic value for a driving pattern for each sensor for a specific time interval and calculates a rate of change for the representative characteristic value between consecutive specific time intervals. Learning can be performed by generating a feature vector having the rate of change of each sensor as a component value as learning data.

More specifically, the fire diagnosis model generating unit 350 may set a predetermined specific time interval as a unit section, and may derive a driving pattern for each sensor for each unit section. In addition, the fire diagnosis model generating unit 350 may set a representative value that can represent the driving pattern for each unit section, for example, through the average of the characteristic values related to the driving state during the unit section, the corresponding unit A representative feature value for the interval may be determined.

In addition, the fire diagnosis model generation unit 350 may calculate a change rate between representative feature values derived for each unit section, and may generate an n-dimensional feature vector having a change rate for each sensor as a component as learning data. In this case, n may correspond to the number of sensors installed in the facility device.

In one embodiment, the fire diagnosis model generation unit 350 obtains two-dimensional images about the driving pattern for each sensor for a specific time interval, and an n-dimensional (where n is a natural number) feature vector for each of the two-dimensional images. A fire diagnosis model can be created by generating a representative feature vector representing a specific time interval by integrating the feature vectors for each sensor into one, and learning about the representative feature vector.

More specifically, the fire diagnosis model generation unit 350 may generate a two-dimensional pattern graph as a result of visualizing the driving pattern for each sensor, and divide the time axis of the corresponding two-dimensional pattern graph into specific intervals at each specific interval. It is possible to obtain a two-dimensional image of the pattern graph for In addition, the fire diagnosis model generator 350 may generate an n-dimensional feature vector through image analysis of the two-dimensional image. For example, the fire diagnosis model generator 350 may generate an n-dimensional feature vector for a specific image using a Convolution Neural Network (CNN) algorithm.

That is, the fire diagnosis model generating unit 350 may integrate the n-dimensional feature vectors obtained for each sensor into one and generate a representative feature vector as integrated information about the operating state of the facility at the time point, and for each cycle By learning the representative feature vectors at each time point, a fire diagnosis model for fire diagnosis can be built. As a result, the fire diagnosis model built through this can output the fire risk as a dependent variable by using the correlation between the driving patterns for each sensor as an independent variable.

In an embodiment, the fire diagnosis model generation unit 350 may generate a fire risk function for outputting a prediction probability regarding the possibility of fire as a fire diagnosis model based on the input data. That is, the fire diagnosis model generation unit 350 calculates the fire prediction probability according to various input conditions (temperature, humidity, spark, smoke and gas, etc.) The output fire hazard can also be built as a function. Accordingly, the fire prediction device 130 can effectively diagnose the occurrence of a fire according to the operating state collected from the facility device using the fire diagnosis model.

The fire occurrence diagnosis unit 370 may apply real-time diagnosis data collected from the facility device to the fire diagnosis model to diagnose the occurrence of a fire in the facility device. The fire occurrence diagnosis unit 370 may diagnose and monitor the occurrence of a fire in each facility device using a fire diagnosis model built through sufficient prior learning. Meanwhile, the fire occurrence diagnosis unit 370 may process learning and diagnosis in parallel. That is, the fire diagnosis model built through prior learning can perform real-time diagnosis on fire occurrence and at the same time obtain feedback according to the diagnosis result to learn about real-time diagnosis data. It may be updated periodically or in real time by the fire occurrence diagnosis unit 370 based on the data.

In one embodiment, the fire occurrence diagnosis unit 370 applies real-time diagnostic data to the fire occurrence risk function and diagnoses the occurrence of a fire in the corresponding facility device according to the output value of the fire diagnosis model obtained as a result of the application. . The fire diagnosis model can be built through learning based on the previously collected diagnosis data, and as a result, the fire diagnosis unit 370 uses the fire diagnosis model to determine A fire can be effectively diagnosed.

For example, the fire occurrence diagnosis unit 370 may determine a preset fire generation level based on the output value of the fire diagnosis model and perform an operation according to the level. The fire occurrence class may correspond to a definition of the possibility of fire by class, and may be further classified including caution, warning and ignition classes. The fire occurrence diagnosis unit 370 may determine a current grade according to the occurrence probability output by the fire diagnosis model and process an operation according to an action for each grade.

In one embodiment, when the output value of the fire diagnosis model for a specific time interval exceeds a preset threshold value and is continuously repeated more than a preset number of times, the fire occurrence diagnosis unit 370 may provide an alarm regarding the occurrence of a fire. there is. That is, the fire occurrence diagnosis unit 370 predicts the occurrence of a fire based on the number of times the output value of the fire diagnosis model temporarily exceeds the set threshold value in consideration of the possibility that it is outside the normal range, and at the same time generates an alarm for the manager. It can be transmitted to an external system including the terminal 170 .

The control unit 390 controls the overall operation of the fire prediction device 130 , and operates between the diagnosis data receiving unit 310 , the driving pattern deriving unit 330 , the fire diagnosis model generating unit 350 , and the fire occurrence diagnosis unit 370 . It can manage control flow or data flow.

4 is a flowchart illustrating an embodiment of an artificial intelligence-based fire prediction process performed in the fire prediction apparatus of FIG. 1 .

Referring to FIG. 4 , the fire prediction device 130 is installed in the facility through the diagnostic data receiving unit 310 and collects diagnostic data on the operating state of the facility from the monitoring terminal 110 including a plurality of sensors. It can be done (step S410). The fire prediction device 130 may derive a driving pattern for each sensor defined as a driving state for a specific time interval based on the diagnostic data through the driving pattern deriving unit 330 (step S430).

In addition, the fire prediction device 130 may generate a fire diagnosis model for predicting the risk of a fire with respect to the facility equipment through learning about the correlation between operation patterns for each sensor through the fire diagnosis model generation unit 350 . (Step S450). The fire prediction device 130 may apply the real-time diagnosis data collected from the facility device through the fire occurrence diagnosis unit 370 to the fire diagnosis model to diagnose the occurrence of a fire in the facility device (step S470).

FIG. 5 is a view for explaining an embodiment of an AI-based fire prediction process using pattern analysis performed in the fire prediction device of FIG. 1 .

Referring to FIG. 5 , the fire prediction apparatus 130 may derive a driving pattern for each sensor based on the diagnostic data collected from the monitoring terminal 110 . In this case, the driving pattern for each sensor may be defined as a driving state for a specific time interval. In particular, the fire prediction device 130 may derive a two-dimensional pattern graph related to diagnostic data for each sensor, and a fire warning section including a section deviating from the operating range set for each sensor based on the time axis of the pattern graph ( 510) can be determined.

In FIG. 5 , the pre-fire section may correspond to a section including the pre-fire symptom element, and the pre-fire symptom element may correspond to a case in which diagnostic data collected from a specific sensor is outside the operating range set for the specific sensor. there is. That is, the fire precursor symptom element may correspond to an abnormal situation occurring during the operation of a facility device before a fire occurs, and a case in which the sensing value is outside the normal range for no particular reason may correspond to a representative example.

On the other hand, the fire prediction device 130 may generate a fire occurrence risk function that outputs a predicted probability regarding the possibility of a fire based on the input data as a fire diagnosis model, and provides real-time diagnostic data from the facility equipment to the fire occurrence risk function. can be applied to diagnose the occurrence of fire in the relevant facility equipment according to the output value. At this time, the fire prediction device 130 does not simply diagnose by one sensor, but predicts and provides a fire occurrence probability 530 as a result of analyzing the correlation between information collected from various sensors to diagnose fire occurrence. can increase the accuracy of

In one embodiment, the fire prediction device 130 classifies the diagnostic data collected from the plurality of monitoring terminals 110 by type to build the database 150 and generate a monitoring map for each type of a plurality of facility devices. It may be implemented by further including a diagnostic data visualization unit. Here, the monitoring map is data generated based on the diagnostic data, and may be derived by aligning the operation status of each facility according to the passage of time by type on the same plane. That is, the fire prediction device 130 can effectively perform monitoring of a plurality of facility devices through the monitoring map.

In addition, the diagnostic data visualization unit may determine a region of interest based on the distribution on the monitoring map for each type and detect abnormal signals regarding the operation of facilities included in the region of interest. In addition, the diagnostic data visualization unit may provide a notification for a facility device associated with an abnormal signal, and accordingly, the fire prediction device 130 may perform a precise diagnosis on the facility device.

In an embodiment, the diagnostic data visualization unit may generate a monitoring map for each type as a two-dimensional matrix with a plurality of facility devices and measurement times as each dimension using the monitoring data classified by type. For example, a monitoring map regarding the internal temperature of a facility may be expressed as a two-dimensional matrix in which each of a plurality of facility devices corresponds to one dimension, and a temperature signal received from each facility corresponds to the other dimension. there is. The diagnostic data visualization unit may visualize a type-specific monitoring map in the form of a table or graph on one plane.

In an embodiment, the diagnostic data visualization unit expands the two-dimensional matrix in three dimensions by arranging monitoring maps for each type side by side based on the time dimension, and when the number of overlaps between candidate regions of interest determined in each two-dimensional matrix exceeds a preset threshold An ROI may be determined as a set of corresponding candidate ROIs. More specifically, the diagnostic data visualization unit may arrange and arrange the monitoring maps for each type so that the time dimension coincides, and accordingly, the two-dimensional matrix may be expanded into three dimensions (refer to FIG. 6 ).

In addition, the diagnostic data visualization unit can determine the candidate ROI as a result of detecting the presence or absence of abnormalities in the diagnostic data on the monitoring map for each type, and pay attention to the region in which the overlap of the candidate ROI is repeated a certain number or more on the three-dimensionally extended monitoring map. area can be finally determined. That is, the region of interest determined by the diagnostic data visualization unit may be utilized to determine a facility with a high probability of detecting an abnormality in an operation signal.

FIG. 6 is a view for explaining a monitoring map for each type used in the fire prediction device of FIG. 1 .

Referring to FIG. 6 , the fire prediction device 130 generates a monitoring map for each type as a two-dimensional matrix with a plurality of facility devices and measurement times in each dimension using the diagnostic data classified by type through the diagnostic data visualization unit. can In the case of FIG. 6 , it may correspond to a temperature monitoring map generated based on temperature signals measured from facility devices. That is, each row of the temperature monitoring map may correspond to each facility device (facility 1, facility 2, ..., facility n), and each column corresponds to each measurement time (t1, t2, t3, ... , tn). Accordingly, the fire prediction device 130 can effectively detect an abnormal signal exceeding the normal range in each facility device through the temperature monitoring map.

FIG. 7 is a view for explaining a process of determining a region of interest performed by the fire prediction apparatus of FIG. 1 .

Referring to FIG. 7 , the fire prediction device 130 expands the two-dimensional matrix 710 in three dimensions by arranging the monitoring maps for each type side by side based on the time dimension t through the diagnostic data visualization unit, and in each two-dimensional matrix When the number of overlapping between the determined candidate ROIs 730 exceeds a preset threshold, the ROI 770 may be determined as a set of corresponding candidate ROIs. That is, the ROI 770 may correspond to a region including the largest overlapping region 750 of the candidate ROIs 730 .

In FIG. 7, the fire prediction device 130 can build monitoring maps related to temperature, humidity, sparks, etc. as a two-dimensional matrix 710, respectively, and can be extended in three dimensions by arranging them side by side based on the time dimension t. can Each two-dimensional matrix 710 may be arranged such that diagnostic data obtained at the same time t for the same facility device f correspond to each other, and the fire prediction device 130 is a candidate determined on each monitoring map as a result of the arrangement. It may be determined whether the largest overlapping region 750 in which the overlapping number of the regions of interest 730 is greater than or equal to a specific number (eg, 3) may be determined.

If there is a maximum overlapping region 750 with a number of overlaps equal to or greater than 3, the fire prediction apparatus 130 may finally determine a set of candidate regions of interest 730 forming the overlap as the region of interest 770 . Thereafter, the fire prediction device 130 may perform a precise diagnosis by detecting an abnormal signal targeting the facility devices associated with the region of interest 770 .

Although the above has been described with reference to the preferred embodiment of the present invention, those skilled in the art can variously modify and change the present invention within the scope without departing from the spirit and scope of the present invention as described in the claims below. You will understand that it can be done.

100: fire prediction system
110: monitoring terminal 130: fire prediction device
150: database 170: administrator terminal
210: processor 230: memory
250: user input/output unit 270: network input/output unit
310: diagnostic data receiving unit 330: driving pattern deriving unit
350: fire diagnosis model generation unit 370: fire diagnosis unit
390: control unit
510: fire precursor section 530: fire probability
710: two-dimensional matrix 730: candidate region of interest
750: most overlapping region 770: region of interest

Claims (8)

Collecting diagnostic data on the operating state of the facility from a monitoring terminal installed in the facility and including a plurality of sensors;
deriving a driving pattern for each sensor defined as a driving state for a specific time interval based on the diagnostic data;
generating a fire diagnosis model for predicting the risk of fire occurrence with respect to the facility equipment through learning about the correlation between the driving patterns for each sensor; and
and diagnosing the occurrence of a fire in the facility by applying the real-time diagnosis data collected from the facility to the fire diagnosis model,
The monitoring terminal measures signals related to temperature, humidity, spark, smoke and gas in the operation process of the facility device through the plurality of sensors,
The step of collecting the diagnostic data is
Sampling an electrical signal of the diagnostic data at a specific period and converting it into a digital signal,
The step of deriving the driving pattern is
deriving a two-dimensional pattern graph related to the diagnostic data for each sensor;
arranging the two-dimensional pattern graph based on a time axis and then determining a fire warning section that deviates from the operation range for each sensor based on a value axis; and
Determining each of the sections before and after the fire warning section based on the time axis and determining the change in the pattern graph from the previous section to the subsequent section as the driving pattern for each sensor,
The step of generating the fire diagnosis model is
As a result of determining the representative characteristic value for the driving pattern for each sensor for the specific time interval and calculating the rate of change for the representative characteristic value between successive specific time intervals, a feature vector using the rate of change for each sensor as a component value Including the step of performing the learning by generating as learning data,
The step of generating the fire diagnosis model is
acquiring two-dimensional images related to the driving pattern for each sensor for the specific time interval;
generating an n-dimensional (where n is a natural number) feature vector for each of the two-dimensional images;
generating a representative feature vector representing the specific time interval by integrating the feature vectors for each sensor into one; and
generating the fire diagnosis model by performing learning on the representative feature vector,
The step of generating the fire diagnosis model is
Generating a fire risk function that outputs a predicted probability regarding the possibility of a fire based on input data as the fire diagnosis model,
Diagnosing the fire
applying the real-time diagnosis data to the fire occurrence risk function and diagnosing the occurrence of a fire in the facility device according to the output value of the fire diagnosis model obtained as a result of the application,
Diagnosing the fire
and providing an alarm regarding the occurrence of a fire when the output value of the fire diagnosis model for the specific time interval exceeds a preset threshold value and is continuously repeated for more than a preset number of times. A fire prediction method based on artificial intelligence.
delete delete delete delete delete delete a diagnostic data collection unit that collects diagnostic data on operating conditions from a plurality of sensors installed in the facility;
a driving pattern derivation unit for deriving a driving pattern defined as a driving state for a specific time interval for each sensor based on the diagnostic data;
a fire diagnosis model generation unit for generating a fire diagnosis model for predicting the occurrence of a fire with respect to the facility equipment through learning about the correlation between the driving patterns; and
A fire occurrence diagnosis unit for diagnosing the occurrence of a fire in the relevant facility by applying the real-time diagnosis data collected from the facility device to the fire diagnosis model,
The diagnostic data collection unit
Measures signals related to temperature, humidity, spark, smoke and gas in the operation process of the facility device through the plurality of sensors,
The diagnostic data collection unit
An electrical signal among the diagnostic data is sampled at a specific period and converted into a digital signal,
The driving pattern deriving unit
deriving a two-dimensional pattern graph related to the diagnostic data for each sensor;
arranging the two-dimensional pattern graph based on a time axis and then determining a fire warning section that deviates from the operation range for each sensor based on a value axis; and
Determining each of the sections before and after the fire warning section based on the time axis, and determining the change in the pattern graph from the previous section to the subsequent section as a driving pattern for each sensor,
The fire diagnosis model generation unit
As a result of determining the representative characteristic value for the driving pattern for each sensor for the specific time interval and calculating the rate of change for the representative characteristic value between successive specific time intervals, a feature vector using the rate of change for each sensor as a component value Perform the learning by generating it as learning data,
The fire diagnosis model generation unit
acquiring two-dimensional images related to the driving pattern for each sensor for the specific time interval;
generating an n-dimensional (where n is a natural number) feature vector for each of the two-dimensional images;
generating a representative feature vector representing the specific time interval by integrating the feature vectors for each sensor into one; and
performing the learning on the representative feature vector to generate the fire diagnosis model;
The fire diagnosis model generation unit
Generate a fire risk function that outputs a predicted probability about the possibility of a fire based on the input data as the fire diagnosis model,
The fire diagnosis unit
Applying the real-time diagnosis data to the fire occurrence risk function and diagnosing the occurrence of a fire in the facility device according to the output value of the fire diagnosis model obtained as a result of the application,
The fire diagnosis unit
Artificial intelligence-based using pattern analysis, characterized in that when the output value of the fire diagnosis model for the specific time interval exceeds a preset threshold value and is continuously repeated for more than a preset number of times, an alarm about the occurrence of a fire is provided. fire prediction device.

Images