KR102685581B1 - Anticipation and damage prevention system for hazardous substance accident risk areas in high-risk industrial sites using artificial intelligence - Google Patents

Abstract

The present invention relates to a system for predicting hazardous material accident risk areas and preventing damage in high-risk industrial sites using artificial intelligence. The present invention relates to an IoT sensor device installed at an industrial site that collects sensor data and the IoT sensor that transmits sensor data to itself. Groups devices, receives sensor data from IoT sensor devices belonging to the group, and receives various sensor data from the gateway and transmits it to the risk analysis server, determines the starting point of risk occurrence according to the type of sensor and the value of the sensor data, It is characterized by including a risk analysis server that predicts accident risk areas by inputting weather, terrain, types of buildings and facilities, and each sensor data type and sensor data value into an artificial intelligence risk analysis engine.
The present invention determines the starting point of risk based on the type of sensor and the value of sensor data received from the gateway, and analyzes weather, terrain, types of buildings and facilities, and the type and value of each sensor data using an artificial intelligence risk analysis engine. When entered, it is possible to accurately predict the area where accident risks occur by comprehensively reflecting not only the type and value of sensor data but also weather, terrain, and building conditions.

Description

A system for predicting and preventing hazardous material accidents at high-risk industrial sites using artificial intelligence.

The present invention relates to a system for predicting areas at risk of hazardous material accidents and preventing damage in high-risk industrial sites using artificial intelligence. More specifically, it relates to accidents such as fires, explosions, and gas leaks using various risk detection sensors installed at industrial sites. This is about a system for predicting hazardous substance accident risk areas and preventing damage in high-risk industrial sites using artificial intelligence, which analyzes environmental information such as occurrence type, weather, terrain, and buildings using artificial intelligence to predict and display accident risk areas.

Currently, there are about 200,000 types of chemicals in circulation worldwide, and about 3,000 types of new chemicals are being developed every year. In Korea, as of 2018, it is understood that over 29,000 types of chemicals are being handled at over 39,000 workplaces.

From 2016 to 2020, a total of 351 chemical accidents occurred, with an annual average of about 70. By region, Gyeonggi, Gyeongbuk, Ulsan, Chungnam, and Jeollanam-do were the most common. In Korea, in the process of forming modern cities, industrial complexes and residential areas became adjacent to each other while buffer green areas were not secured due to rapid urbanization, industrialization, and reckless development. The damage from the hydrogen fluoride leak accident at Huve Global Co., Ltd. in Gumi, Gyeongbuk clearly revealed the seriousness of this problem, and it has been repeatedly seen in the subsequent leak accidents at SK Materials in Yeongju, Gyeongbuk and Ire Chemical in Seo-gu, Incheon. In fact, it is a structural problem with no exceptions anywhere in the country.

On the other hand, in Europe, following the Seveso chemical accident that occurred in Italy in 1976, the 'Seveso Guidelines IⅠⅡⅢ' led to the preparation of process safety reports, regulation of land use in surrounding areas, establishment of emergency response plans, activation of risk communication, accident investigation, and sharing of lessons learned. We are continuously establishing a response system for chemical accidents to prevent similar accidents from recurring.

In Korea, the use of domestic toxic substances and chemicals continues to increase due to the development of national new growth engine industries and technologies such as semiconductors, displays, and solar energy. Therefore, standard action procedures according to the types of hazardous substances and types of leaks, fires, and explosions are efficiently implemented. There is a need for an integrated system that can be applied in the field and quickly delivered to relevant organizations and field personnel.

In addition, in the event of a disaster involving a leak of hazardous chemicals within an industrial complex, the primary casualty of direct human and material damage in the area at the time of the leak and the secondary damage to nearby areas that spread over time will be covered. We also need a system that takes everyone into account.

As a prior art to solve these conventional problems, there is Patent No. 10-2244634, ‘IPV-based hazardous chemical substance leak monitoring system’.

Registered Patent No. 10-2244634 is arranged in a plurality of matrix arrays in areas or inside factories at risk of hazardous chemical leaks, and measures the concentration of at least one hazardous chemical in the air through at least one hazardous chemical leak detection sensor. At least one IoT (Internet of Things) device that transmits unique identification information stored in advance along with each measured hazardous chemical concentration data, and unique identification information along with each hazardous chemical concentration data transmitted from each IoT device. Based on this, metadata is assigned through multiple channels encoded at different bit rates depending on the type and size of each hazardous chemical concentration data, and each hazardous chemical concentration data for which metadata is given is corresponded to a cloud-compatible protocol. A data processing device that converts and transmits the concentration status information data of each hazardous chemical substance, and the concentration status information data of each hazardous chemical substance converted and transmitted from the data processing device is provided, and based on this, the manager determines the concentration status information data of each hazardous chemical substance. It is about visual monitoring of leakage status.

Registered Patent No. 10-2244634 has structural limitations in that it cannot quickly and accurately determine the initial location where an explosion, fire, gas leak, etc. occurred and the path along which fire, smoke, and gas move.

In order to solve the limitations of Registration Patent No. 10-2244634, the applicant (Zubix Co., Ltd.) applied for a 'risk monitoring system that quickly detects the location and direction of risk in high-risk industrial sites' (Application No. 10-2022- 0160469) and registered (Registered Patent No. 10-2552726).

Registered Patent No. 10-2552726 sets up several virtual groups by grouping IoT sensor devices in industrial sites where risks occur and belonging to one gateway with IoT sensor devices in industrial sites that belong to another gateway, and sets up several virtual groups that belong to the virtual groups. By applying a method to determine the initial location of risk and the path of risk by checking sensor data collected from multiple IoT sensor devices, the system flexibly expands beyond the physical group managed by each gateway to a virtual group. When the risk of an IoT sensor device belonging to a virtual group is lifted, the virtual group is reset by switching to another IoT sensor device, and the risk movement path is identified through the virtual group to prevent fire, explosion, or explosion. It is characterized by accurately and quickly identifying the global movement path of gases, etc.

Registered Patent No. 10-2552726 determines the movement path of fire, explosion, and gas hazards based only on sensor data values received from IoT sensor devices, and has the advantage of being able to identify the global movement path of accident risk, but , the terrain, buildings, and facilities are not comprehensively considered, and it is not possible to predict accident risk areas and respond in advance.

Therefore, in the event of an accident such as a fire, explosion, or gas leak, the starting point of risk is determined, and the weather, terrain, types of buildings and facilities, and the type and value of each sensor data are input into the artificial intelligence risk analysis engine to determine the accident risk. We plan to develop a system that predicts the occurrence area and can spread the word in advance to the area where the risk of an accident is expected to occur.

Republic of Korea Patent Registration No. 10-2244634 (IoT-based hazardous chemical substance leak monitoring system) Republic of Korea Patent Registration No. 10-1926368 (Real-time damage prediction monitoring system in case of hazardous chemical leakage) Republic of Korea Patent Registration No. 10-2552726 (Risk monitoring system that quickly detects the location and direction of risk in high-risk industrial sites)

The present invention is intended to solve the limitations of the prior art. In areas where IoT sensor devices are intensively installed, several IoT sensor devices collect sensor data and transmit it to the gateway, and the gateway transmits the collected sensor data to the risk analysis server. When transmitted, the risk analysis server receives various sensor data and determines the starting point of risk according to the type of sensor and the value of the sensor data. Predicting and preventing damage from hazardous substances at high-risk industrial sites using artificial intelligence, which predicts areas where accident risks may occur by inputting values into an artificial intelligence risk analysis engine and prevents damage by spreading the information in advance to areas where accident risks are expected to occur. This is to provide a system.

In order to achieve the above object, the present invention groups IoT sensor devices that are installed in industrial sites and collect sensor data and the IoT sensor devices that transmit sensor data to themselves, and collects sensor data from IoT sensor devices belonging to the group. Receives various sensor data from the gateway and transmits it to the risk analysis server, determines the starting point of risk according to the type of sensor and the value of the sensor data, and determines the starting point of risk, weather, terrain, types of buildings and facilities, and each sensor data. It includes a risk analysis server that predicts accident risk areas by inputting the type and sensor data values into an artificial intelligence risk analysis engine.

The risk analysis server is composed as follows.

When sensor data is received from the gateway, there is a sensor database that accumulates and stores them in chronological order, a weather database that stores weather information by date, an environment database that stores types of terrain, buildings, and facilities by location, and the types of sensors and numbers of sensor data. It learns with the risk start point judgment unit that determines the risk start point by checking the risk start point, learning sensor data, weather information from the weather database, and terrain, building and facility information from the environmental database, and various sensor data stored in the risk start point and sensor database. An artificial intelligence risk analysis engine that predicts accident risk areas using weather information stored in the weather database and terrain, building and facility information stored in the environment database, and the accident risk risk area predicted by the artificial intelligence risk analysis engine is displayed on the screen. It is composed of an accident risk area display unit.

In addition, the risk analysis server cancels the risk occurrence if the probability value of the point predicted by the artificial intelligence risk analysis engine as the accident risk occurrence area does not reach the preset level, or updates the risk occurrence start point so that the artificial intelligence risk analysis engine determines the accident risk occurrence area. It can be configured to further include a risk starting point update unit to re-anticipate.

The present invention provides the expected value of sensor data stored in the database of the risk analysis server up to the area where the risk analysis server predicted the accident risk and the date when the risk of the accident was expected to occur when the risk analysis server's prediction of the accident risk occurrence area is wrong. It further includes a risk prediction failure cause inspection server that analyzes the difference between the actual value and the change trend of the expected value to inspect the cause of the expected error in the accident risk occurrence area, and the risk analysis server inspects the risk expectation failure cause inspection server. The results are reflected and learned in the artificial intelligence risk analysis engine.

The risk prediction failure cause inspection server is composed as follows.

A sensor data trend analysis unit that identifies one or more outliers by analyzing the difference between the expected and actual values of sensor data and the change trend in the expected values, as well as the terrain, types of buildings and facilities, and outliers around the IoT sensor device where the outlier occurred. It consists of a singularity occurrence output unit that analyzes the weather at the time of occurrence and outputs the cause of the singularity occurrence, and a singularity reflection portion that reflects the singularity occurrence cause output by the singularity occurrence output unit into the artificial intelligence risk analysis engine.

In addition, the risk prediction failure cause inspection server analyzes the cause of singular point occurrence for each singular point output by the singularity cause output unit, and if it determines that the cause is problematic, it excludes it from the cause of singular point occurrence and outputs only the rest by filtering them as causes of singular point occurrence. It further includes a singularity verification unit, and the singularity reflection unit can reflect the cause of the singularity filtered and output by the singularity verification unit into the artificial intelligence risk analysis engine.

The present invention determines the starting point of risk based on the type of sensor and the value of sensor data received from the gateway, and analyzes weather, terrain, types of buildings and facilities, and the type and value of each sensor data using an artificial intelligence risk analysis engine. When entered, it is possible to accurately predict the area where accident risks occur by comprehensively reflecting not only the type and value of sensor data but also weather, terrain, and building conditions.

In addition, the present invention provides the expected value of the sensor data stored in the database of the risk analysis server up to the area where the risk analysis server predicted the accident risk and the date when the risk of the accident was expected to occur if the risk analysis server's prediction of the accident risk occurrence area is wrong. By analyzing the difference between the actual value and the change trend in the expected value, we examine the cause of the prediction of the accident risk occurrence area being wrong, and the inspection results are reflected in the artificial intelligence risk analysis engine for learning, further improving the ability to predict the accident risk occurrence area. You can do it.

Figure 1 is a diagram showing the configuration of a system for predicting hazardous material accident risk areas and preventing damage in high-risk industrial sites using artificial intelligence of the present invention.
Figure 2 shows the IoT sensors, gateways, and the arrangement of various terrains, buildings, and facilities at industrial sites, and fire, explosion, and gas, which constitute a system for predicting and preventing damage from hazardous substances in high-risk industrial sites using artificial intelligence of the present invention. These are various examples of leak accidents occurring.
Figure 3 is a photo of an IoT sensor device installed in the field used in the present invention.
Figure 4 is a functional block diagram of the gateway and risk analysis server that constitute the system for predicting hazardous material accident risk areas and preventing damage in high-risk industrial sites using artificial intelligence of the present invention.
Figure 5 is a functional block diagram of a risk prediction failure cause inspection server that constitutes a system for predicting hazardous material accident risk areas and preventing damage in high-risk industrial sites using artificial intelligence of the present invention.

The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings.

However, the present invention is not limited to the embodiments disclosed below and will be implemented in various different forms.

The examples herein are provided to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention.

And the present invention is only defined by the scope of the claims.

Accordingly, in some embodiments, well-known components, well-known operations and well-known techniques are not specifically described in order to avoid ambiguous interpretation of the present invention.

In addition, the same reference numerals refer to the same elements throughout the specification, and the terms used (mentioned) in the specification are for explaining embodiments and are not intended to limit the present invention.

In this specification, the singular also includes the plural unless specifically stated in the phrase, and elements and operations referred to as 'including (or, including)' do not exclude the presence or addition of one or more other elements and operations. .

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used with meanings that can be commonly understood by those skilled in the art to which the present invention pertains.

Additionally, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless they are defined.

Hereinafter, preferred embodiments of the present invention will be described with reference to the attached drawings.

Referring to Figures 1 to 5, the system for predicting hazardous material accident risk areas and preventing damage at high-risk industrial sites using artificial intelligence includes IoT sensor devices (101-110), a gateway (200), a risk analysis server (300), and It is configured to include a risk prediction failure cause inspection server (400).

IoT sensor devices 101 to 110 are installed at industrial sites and collect sensor data. IoT sensor devices (101-110) include various sensor devices such as harmful gas sensors, fire sensors, precipitation sensors, and wind direction/speed sensors. IoT sensor devices (101 to 110) collect harmful gas measurement data, fire smoke measurement data, precipitation data, and wind direction/speed data as sensor data.

The gateway 200 groups IoT sensor devices 101 to 110 that transmit sensor data to itself, receives sensor data from IoT sensor devices 101 to 110 belonging to the group, and transmits it to the risk analysis server 300. .

In FIG. 1, four IoT sensor devices 101 to 104 are shown transmitting sensor data to the gateway 210, but this is not necessarily limited, and three IoT sensor devices transmitting sensor data to the gateway 220 ( 105 to 107) and three IoT sensor devices 108 to 110 that transmit sensor data to the gateway 230 are also provided as examples, and are not necessarily limited thereto.

Referring to FIGS. 1 and 2, each gateway 210, 220, and 230 groups IoT sensor devices 101 to 110 that transmit sensor data to each gateway, which are referred to as physical groups.

The gateway 200 includes a sensor group management unit 201.

The sensor group management unit 201 registers the type and location information of IoT sensor devices that transmit sensor data to itself and manages them as a group.

Gateway 210 registers the type and location information of IoT sensor devices (101 to 104) that transmit sensor data to itself and manages them as a group. Gateway 220 registers the type and location information of IoT sensor devices (105 to 107) that transmit sensor data to itself and manages them as a group. Gateway 230 registers the type and location information of IoT sensor devices (108 to 110) that transmit sensor data to itself and manages them as a group. A group refers to a physical group that actually transmits sensor data to the gateway.

The risk analysis server 300 receives various sensor data from the gateway 200 and determines the starting point of risk according to the type of sensor and the value of the sensor data, weather, terrain, types of buildings and facilities, and the type of each sensor data. and sensor data values are input into the artificial intelligence risk analysis engine to predict accident risk areas.

Depending on the type of sensor, such as a harmful gas sensor, fire sensor, precipitation sensor, or wind direction/speed sensor, the range of sensor data values is different, and the risk occurrence value indicating the level of risk is also different. The risk analysis server 300 determines the IoT sensor device whose sensor data values of the various IoT sensor devices 101 to 110 received from the gateway 200 are at a dangerous level as the starting point of risk occurrence.

Referring to Figure 2, industrial sites include mountains in the middle, factories, apartments, sewage treatment facilities, and factory dormitories. In addition, various accidents occur at industrial sites, such as fires, explosions, gas leaks, etc. These accidents can occur simultaneously or sequentially, and the weather at the time of the accident (rain, wind, wind speed, etc.) Depending on the condition and type of accident (fire, explosion, gas leak, etc.), the location of the mountain, factory, apartment, sewage treatment facility, factory dormitory, and the time of accident occurrence also vary.

For example, since a fire occurring near a mountain is more dangerous than a gas leak occurring near a mountain, the extent and direction of spread of the fire are determined depending on the weather conditions at the time of a fire occurring near the mountain, and the surrounding area is determined by the weather conditions. Predict accident risk areas depending on the type of facility.

If a gas leak occurs near a mountain, it can be judged not to be serious, but the situation is different if a gas leak occurs near an apartment or sewage treatment facility. A gas leak accident occurs near an apartment or sewage treatment facility, and the degree of spread and direction of spread are determined depending on weather conditions such as wind and wind speed at the time of occurrence. There are apartments or dormitories in the expected direction, and the time of the accident occurs. Since damage will be significant after 8 p.m., apartments and dormitories are expected to be areas at risk of accidents.

However, gas leak accidents occur near apartments or sewage treatment facilities, and the degree of spread and direction of spread are determined by weather conditions such as wind and wind speed at the time of occurrence, there is a factory in the direction where progress is expected, and the time of the accident is 8 p.m. Since there will be no casualties at the factory after midnight, the factory will not be expected to be an accident risk area.

An explosion occurs near a mountain or factory, and the extent and direction of spread are determined depending on weather conditions such as wind and wind speed at the time of occurrence. Factories, mountains, and apartments are located in the expected direction, and the accident occurs in the evening. Since the damage will be significant after 8 o'clock, factories, mountains, and apartments are expected to be areas at risk of accidents. Since an explosion accident causes a fire, the damage spreads in the direction in which the fire is expected to progress, and rain may occur at a certain distance or the wind direction may change. In this case, the area expected to be at risk of an accident may occur. Accident damage may not occur.

The risk analysis server 300 includes a sensor database 301, a weather database 302, an environmental database 303, a risk starting point determination unit 304, an artificial intelligence risk analysis engine 305, and an accident risk area display unit 306. It consists of:

When the sensor database 301 receives various sensor data from the gateway 200, it accumulates and stores each sensor data in chronological order.

The weather database 302 stores weather information (rain, wind, wind speed, clouds, etc.) by date and time and region obtained through communication using API with the Korea Meteorological Administration server.

The environmental database 303 stores the types of terrain, buildings, and facilities for each location. As shown in Figure 2, information on mountains, fields, apartments, factories, sewage treatment facilities, dormitories, etc. are stored for each coordinate on the map, and information on the occupied area is also stored.

The risk start point determination unit 304 determines the risk occurrence start point by checking the type of sensor and the value of the sensor data. Looking at Figure 2, there are several different types of IoT sensor devices 101 to 110 at each location. When a fire, explosion, or gas leak occurs, the IoT sensor devices (101 to 110) detect it, and when it exceeds the value of the detected sensor data and corresponds to a dangerous level, the point where the corresponding IoT sensor device is located is judged as the starting point of the risk. The starting point of risk occurrence may be observed in multiple places at the same time.

The artificial intelligence risk analysis engine 305 learns with learning sensor data, weather information from the weather database 302, and terrain, building and facility information from the environmental database 303. Sensor data for learning is a value that is randomly generated and used for learning within the allowable range of sensor data values for each type of sensor. Weather information is learned using recent weather data, and terrain, building, and facility information is learned by realistically reflecting information from industrial sites.

And the artificial intelligence risk analysis engine 305 determines the risk occurrence point determined by the risk start point determination unit 304, various sensor data stored in real time in the sensor database 301, and the Korea Meteorological Administration's latest weather stored in the weather database 302. Accident risk areas are predicted using terrain, building, and facility information stored in the information and environment database 303.

The accident risk area display unit 306 displays the accident risk area predicted by the artificial intelligence risk analysis engine 305 on the screen and informs the manager monitoring the accident site, enabling immediate response by the manager.

The risk analysis server 300 is configured to include a risk start point update unit 307.

The risk start point update unit 307 cancels the risk occurrence when the probability value of the point predicted by the artificial intelligence risk analysis engine 305 as the accident risk occurrence area is below the preset level, or updates the risk occurrence start point to perform artificial intelligence risk analysis. The engine 305 re-anticipates the accident risk area.

Let us assume that there are three points predicted by the artificial intelligence risk analysis engine (305) to be accident risk areas. If the predicted probability values of the three accident risk areas are 35%, 28%, and 42%, respectively, the risk occurrence is canceled because it falls below the preset level of 50%.

When the risk occurrence is canceled, the risk start point determination unit 304 checks the type of sensor and the value of the sensor data to re-determine the risk occurrence start point and newly determines the point where another IoT sensor device is located as the risk occurrence start point. Of course, if the sensor data of the IoT sensor device does not correspond to a dangerous level, the risk release status is maintained.

The artificial intelligence risk analysis engine 305 calculates the risk occurrence start point re-determined by the risk start point determination unit 304, various sensor data stored in real time in the sensor database 301, and the latest weather from the Korea Meteorological Administration stored in the weather database 302. The accident risk occurrence area is predicted again using terrain, building, and facility information stored in the information and environment database 303. Let us assume that there are two points that the artificial intelligence risk analysis engine 305 predicts as accident risk areas. If the predicted probability values of the two accident risk areas are 55% and 60%, respectively, the location is expected to be an accident risk area, and the accident risk area display unit 306 displays the location on the screen.

In the above example, if the accident risk does not occur in a location different from the expected location, but the probability value of the expected accident risk area is below the preset level, the risk occurrence is canceled or the risk occurrence start point is updated to analyze the artificial intelligence risk. It was explained that the engine 305 re-anticipates the accident risk occurrence area.

Below is an explanation of a case where an accident risk does not occur in the accident risk area predicted by the artificial intelligence risk analysis engine 305, but an accident risk occurs in another area.

If the prediction of the accident risk occurrence area of the risk analysis server 300 is incorrect, the inspection server 400, which is the cause of the risk prediction failure, is The difference between the expected value and the actual value of the sensor data stored in the database of the risk analysis server 300 and the change trend in the expected value are analyzed to examine the cause of the unexpected error in the accident risk area.

In FIG. 2, let's say that the risk analysis server 300 predicted that a fire accident risk would occur at location A between 9:00 and 10:00 p.m. on November 28, 2023. However, if the risk of an accident did not occur at location A even after 10 p.m. on November 28, but a risk of an accident occurred at location B, which is a different area, the risk analysis server 300's prediction of location A as the area where the risk of an accident occurred was wrong. will be.

As of 5 p.m. on November 28, the risk analysis server 300 reports the risk occurrence start point, various sensor data stored in the sensor database 301, weather information stored in the weather database 302, and the environmental database 303. Using the stored terrain, building and facility information, it was predicted that an accident risk would occur in area A between 9 PM and 10 PM on November 28, but the prediction of the accident risk area was wrong.

Cause of risk prediction failure The inspection server 400 conducts risk analysis on the area (A) where the risk analysis server 300 predicted an accident risk would occur and from 5 PM to 10 PM on November 28, when the risk of an accident was expected to occur. The difference between the expected value and the actual value of the sensor data stored in the database of the server 300 and the change trend in the expected value are analyzed to examine the cause of the unexpected error in the accident risk area. At this time, the cause of the accident risk actually occurring in area B is also examined and analyzed.

From 5 p.m. to 8 p.m. on November 28, the difference between the predicted and actual values of various sensor data was not large and progress was made as expected, but after 8 p.m., changes occurred in the wind direction and wind speed, leading to area B rather than A. The risk of accidents has arisen in the area. In addition, the inspection reflects the type of facility located in area B and the distance information to area B.

The risk analysis server 300 reflects the test results of the risk prediction failure cause test server 400 into the artificial intelligence risk analysis engine and learns it. Data reflected in learning test results includes not only weather information, but also types of buildings and facilities, time zone information, and distance information to area B.

The risk prediction failure cause inspection server 400 includes a sensor data change trend analysis unit 401, a singularity cause output unit 402, and a singularity reflection unit 403.

The sensor data change trend analysis unit 401 analyzes the difference between the expected value and actual value of the sensor data and the change trend of the expected value to identify one or more outliers. The difference between the expected value and the actual value is small, but from 8 PM on November 28, the difference increases beyond a certain value, and if a large change occurs in the expected value, it is identified as an outlier.

The singularity occurrence cause output unit 402 analyzes the terrain around the IoT sensor devices (101 to 110) where the singularity occurred, the types of buildings and facilities, and the weather at the time the singularity occurred, and outputs the singularity occurrence cause. In addition, the distance from the location of the IoT sensor device (101 to 110) where the singularity occurred to Area B where the actual accident occurred, information on buildings, facilities, and terrain located in between are also output along with the cause of the singularity. In other words, in addition to changes in sensor data values, not only weather information but also the distance to area B and topographic information such as buildings, facilities, mountains, and fields located in between are important.

The singularity reflection unit 403 reflects and learns the cause of singularity occurrence output from the singularity occurrence output unit 402 in the artificial intelligence risk analysis engine 305, thereby accurately predicting the accident risk occurrence area when a similar situation occurs in the future. .

The risk prediction failure cause inspection server 400 includes a singularity verification unit 404.

The singular point verification unit 404 analyzes the cause of singular point occurrence for each singular point output by the singular point occurrence output unit 402, and if it determines that the cause is problematic, it excludes it from the singular point occurrence cause and filters only the rest as the singular point occurrence cause. and print it out.

For example, multiple outliers may be observed as the cause of a fire in area B. Let's say that one IoT sensor device is designated as singular point 1, and another IoT sensor device is designated as singular point 2.

Considering the distance, direction, weather conditions from the IoT sensor device corresponding to Singularity 1 to Area B where the fire occurred, and the buildings, facilities, and terrain located in between, Singularity 1 determined that there was no problem with the cause of the Singularity and determined the cause of Singularity. Select and print.

Considering the distance, direction, weather conditions from the IoT sensor device corresponding to singular point 2 to area B where the fire occurred, and the buildings, facilities, and terrain located in between, the distance between the IoT sensor device corresponding to singular point 2 and area B where the fire occurred If there is a stream, Singularity 2 is a case where there is a problem in the cause of the occurrence. In other words, Singularity 2 must be excluded from the causes of Singularity. Opening is an example, and considering the distance from the IoT sensor device corresponding to Singularity 2 to area B where the fire occurred and the time for the fire to spread to that distance, if the possibility is judged to be low, the cause of the outbreak is problematic. This is the case. In this case, Singularity 2 is excluded from the causes of Singularity.

The singularity reflection unit 403 reflects the cause of singularity occurrence filtered and output by the singularity verification unit 404 in the artificial intelligence risk analysis engine 305. In the above example, by excluding Singularity 2 and learning the cause of Singularity 1 by reflecting it in the artificial intelligence risk analysis engine 305, the accident risk occurrence area can be accurately predicted when a similar situation occurs in the future.

The present invention is not limited to the specific preferred embodiments described above, and various modifications can be made by anyone skilled in the art without departing from the gist of the invention as claimed in the claims. Of course, it is obvious that such changes fall within the scope of the claims.

101~110: IoT sensor device 200, 210, 220, 230: Gateway
201: Sensor group management department 300: Risk analysis server
301: sensor database 302: weather database
303: Environmental database 304: Risk starting point determination unit
305: Artificial intelligence risk analysis engine 306: Accident risk area display unit
307: Risk start point update unit 400: Risk prediction failure cause inspection server
401: Sensor data change trend analysis unit 402: Singularity cause output unit
403: Singularity reflection unit 404: Singularity verification unit

Claims (6)

IoT sensor devices installed at industrial sites and collecting sensor data;
A gateway that groups the IoT sensor devices that transmit sensor data to itself, receives sensor data from IoT sensor devices belonging to the group, and transmits it to a risk analysis server; and
Receives various sensor data from the gateway, determines the starting point of risk according to the type of sensor and the value of the sensor data, and analyzes the weather, terrain, types of buildings and facilities, and the type and value of each sensor data through artificial intelligence risk analysis. Includes a risk analysis server that inputs input into the engine and predicts accident risk areas,
If the risk analysis server's prediction of the accident risk area is wrong, the expected and actual values of the sensor data stored in the risk analysis server's database up to the area where the risk analysis server predicted the accident risk will occur and the date when the accident risk is expected to occur. It further includes a risk prediction failure cause inspection server that analyzes the difference in and the change trend in the expected value to check the cause of the expected error in the accident risk occurrence area,
The risk analysis server is a system for predicting hazardous substance accident risk areas and preventing damage in high-risk industrial sites using artificial intelligence, which is characterized by reflecting and learning the test results of the inspection server for risk prediction failure causes in the artificial intelligence risk analysis engine.
In claim 1,
The risk analysis server,
A sensor database that accumulates and stores sensor data in chronological order when it receives sensor data from the gateway;
A weather database that stores weather information by date and time;
Environmental database that stores types of terrain, buildings, and facilities by location;
A risk start point determination unit that determines the start point of risk by checking the type of sensor and the value of sensor data;
Learning is conducted using sensor data for learning, weather information from the weather database, terrain, buildings and facility information from the environmental database, and the starting point of risk occurrence, various sensor data stored in the sensor database, weather information stored in the weather database, and information stored in the environmental database. Artificial intelligence risk analysis engine that predicts accident risk areas using terrain, building and facility information; and
An accident risk area display unit that displays on the screen the accident risk area predicted by the artificial intelligence risk analysis engine. A system for predicting and preventing damage from hazardous substances at high-risk industrial sites using artificial intelligence.
In claim 2,
The risk analysis server,
If the probability value of the point predicted by the artificial intelligence risk analysis engine as the accident risk occurrence area falls below the preset level, the risk occurrence is canceled or the risk occurrence start point is updated to allow the artificial intelligence risk analysis engine to re-expect the accident risk occurrence area. A system for predicting hazardous material accident risk areas and preventing damage in high-risk industrial sites using artificial intelligence, further comprising a risk start point update unit.
delete In claim 1,
The risk expected failure cause inspection server is,
A sensor data change trend analysis unit that analyzes the difference between the expected value and actual value of the sensor data and the change trend of the expected value to identify one or more outliers;
A singularity source output unit that analyzes the terrain, types of buildings and facilities around the IoT sensor device where the singularity occurred, and the weather at the time the singularity occurred and outputs the cause of the singularity; and
A system for predicting hazardous material accident risk areas and preventing damage in high-risk industrial sites using artificial intelligence, which includes a singularity reflection unit that reflects the cause of singularity output output from the output unit to an artificial intelligence risk analysis engine.
In claim 5,
The risk expected failure cause inspection server is,
It further includes a singularity verification unit that analyzes the cause of singularity for each singularity output by the singularity cause output unit, excludes it from the singularity cause if it determines that the cause is problematic, and outputs only the remaining points by filtering them as causes of singularity. do,
The singularity reflection unit is a system for predicting hazardous substance accident risk areas and preventing damage in high-risk industrial sites using artificial intelligence, characterized in that the singularity occurrence cause filtered and output by the singularity verification unit is reflected in the artificial intelligence risk analysis engine.