KR102480449B1 - Disaster prediction method using multiple sensors - Google Patents

Abstract

The present invention relates to a disaster situation prediction and diagnosis method based on heterogeneous multiple sensors and, more specifically, to a method in which an information system analyzes, by deep learning, data collected from heterogeneous multiple sensors which are of various types and which consist of multiple sensors, predicts signs of a disaster which may occur within a facility, and diagnoses whether the disaster has occurred, the type of disaster which has occurred, the rate of spread, and the situation to be developed in stages.

Description

Disaster prediction and diagnosis method based on heterogeneous multiple sensors {Disaster prediction method using multiple sensors}

The present invention relates to a method for predicting and diagnosing a disaster situation based on heterogeneous multi-sensors. More specifically, the information system analyzes the data collected from heterogeneous multi-sensors, which are of various types and each type consists of a plurality of sensors, by deep learning to predict signs of disasters that may occur within the facility, It is about the method of diagnosing whether a disaster has occurred, the type of disaster that has occurred, the rate of spread, and the situation that will develop in stages.

Disaster safety industry, which uses ICT technology to prevent disasters and minimize damage, is a future convergence industry and one of the rapidly growing high-growth industries. If ICT technology is used in the field of disaster safety, disaster damage can be minimized by detecting and responding to and suppressing disasters at an early stage. It is expected that various solutions and equipment to implement the system will be developed and provided. In particular, various solutions and services converged with digital twin technology are expected to emerge. According to the Information and Communications Technology Planning and Evaluation Institute, the ICT market size in the public safety and disaster safety fields is expected to be approximately $400 billion and $60 billion, respectively, in 2022, with an average annual growth rate of 41.75% and 10.95%.

Disaster management refers to a multifaceted response process to minimize social or physical impact in the event of a large-scale disaster event or accident, such as a large-scale fire, explosion, flood, earthquake, or collapse. Based on the timing, it is composed of four stages: prevention - preparedness - response - recovery. In addition, disaster management is managed by dividing into the stages of interest - caution - alert - serious.

The government as well as the private sector are actively participating in disaster prevention activities and research. Therefore, government agencies, businesses, universities, research institutes, etc. are making efforts to develop technology in the field of disaster prevention and safety management for scientific and systematic disaster prevention. In particular, efforts have been made to develop, apply, and operate disaster response technologies using ICT technology. We are building a smart disaster safety management system.

On the other hand, in cities with large populations, especially megacities such as special cities and metropolitan cities, not only are large buildings such as residential or office buildings and multi-use facilities concentrated, but also underground utility tunnels, tunnels, bridges, substations, charging facilities, etc. Likewise, essential infrastructure facilities of the city are mixed. Most of these urban infrastructure facilities are managed unmanned, and are equipped with monitoring facilities such as various sensors or cctv in preparation for disasters or accidents. However, disasters can occur due to various causes, and once a disaster occurs, the extent of damage and the speed of spread increase uncontrollably because it is located in a densely populated area.

Although these facilities are equipped with various state-of-the-art monitoring facilities such as various sensors and cctv equipment, there is a possibility of disaster. It occurs because the early stage of a disaster or the signs before a disaster is not noticed because it is being analyzed.

Therefore, by analyzing various types of sensor data generated in a specific space or facility and supplementing them by running simulations, it is possible to find signs before a disaster occurs, and to prevent spread and minimize damage at an early stage in the event of a disaster. In addition, it is required to prepare a flexible disaster response plan, and to prepare a plan to prevent disasters and minimize damage at various national levels through quick response and modeling for disasters that may occur in these facilities.

Korean Patent Publication No. 10-2016-0107512 (2016.9.19) Korean Patent Registration No. 10-2208152 (2021.1.28) Korean Patent Registration No. 10-2440097 (2022.9.6) Korean Patent Publication 10-2016-0085033 (2016.7.15)

Lee Mi-sook, Jung Woo-seok, Kim Eun-sol, “A Study on Disaster Safety Management Methods for Digital Twin-Based Underground Utilities,” Journal of Information Science, 39(2), pp. 16-24(2021).

In order to prevent the occurrence of large-scale disasters that may occur in various infrastructure facilities in urban space, such as underground utility tunnels, tunnels, power facilities, etc. The purpose of providing a method to prevent the occurrence of a disaster in advance and prevent its spread when it occurs by using an information system, such as notifying the symptoms of a disaster and capturing and notifying it at an early stage if it is determined that a disaster has occurred do.

The present invention also allows the information system to predict signs of disasters in advance, such as predicting signs of occurrence of disasters, types of disasters with a risk of occurrence, and predicted places of occurrence by using various data collected through heterogeneous multi-sensors. In addition, the purpose is to fundamentally prevent disasters from occurring in facilities by strengthening surveillance activities when detecting signs of disasters.

The present invention also enables the information system to analyze various data collected through heterogeneous multiple sensors to quickly and accurately determine whether a disaster has occurred, the type and place of the disaster, so that even if a disaster occurs, it can be quickly and initially The purpose of this study is to provide a method for implementing an information system capable of minimizing damage caused by a disaster by being able to know and respond accordingly.

Another object of the present invention is to provide a method by which, when it is determined that a disaster has occurred, an information system can analyze data collected from each of heterogeneous multiple sensors to predict the spread speed and direction of the disaster.

Another object of the present invention is to enable an information system to infer the expected spread range of the disaster and the situation to be developed in stages when it is determined that a disaster has occurred, and to predict and inform the risk level of facilities accordingly.

The present invention also builds an information system that can provide timely judgment data that enables disaster managers to make decisions according to the situation, such as capturing signs of a disaster, whether a disaster has occurred, the type of disaster, the extent of spread, and the degree of risk. It aims to make it possible.

The technical problems to be achieved by the present invention are not limited to the above-mentioned technical problems, and other technical problems not mentioned can be clearly understood by those skilled in the art from the description below. There will be.

의 데이터 중에서 재난 발생의 전조증상과 관련되는 센서 종류에서 발생하는 데이터를 선별 및 분석하여 재난의 발생 징후, 발생 위험이 있는 재난의 종류 및 발생 예상 장소를 예측하고 학습하는 재난징후 예측단계(S1); 재난유형 판단모델을 이용하여, 상기 이기종 다중센서 각각에서 수집되는 상기 탐지구간

의 데이터 중에서 상호 간의 변화 연관성을 수평적으로 클러스터링하여 재난의 발생 여부, 발생한 재난의 종류 및 발생 장소를 판단하고 학습하되, - 상기 재난징후 예측단계에서 탐지구간

에 대한 상기 발생 징후를 예측한 경우에는 상기 재난유형 판단모델 중 제1모델을 이용하며, - 상기 발생 징후를 예측하지 아니한 경우에는 상기 재난유형 판단모델 중 제2모델을 이용하는 재난발생 감지단계(S2); 상기 재난발생 감지단계(S2)에서 재난이 발생한 것으로 판단한 경우 재난확산 예측모델을 이용하여, 상기 이기종 다중센서 각각에서 수집되는 데이터 각각에 대한 시계열 변화에 대한 수직적 분석과 상기 변화 연관성에 대한 수평적 분석을 통하여 탐지구간

에서의 재난의 확산속도 및 확산방향을 예측하고 학습하는 재난확산 예측단계(S3); 상기 재난발생 감지단계(S2)에서 재난이 발생한 것으로 판단한 경우 위험도 예측모델을 이용하여, 상기 발생한 재난의 종류, 상기 발생 장소, 상기 확산속도 및 상기 확산방향에 따른 재난의 확산 예상범위 및 단계별로 전개될 상황을 추론하고 이에 따른 시설물 위험도를 예측하고 학습하는 위험도 예측단계(S4); 및 상기 재난의 발생 징후, 상기 발생 위험이 있는 재난의 종류, 상기 발생 예상 장소, 상기 재난의 발생 여부, 상기 발생한 재난의 종류, 상기 발생 장소, 상기 확산속도, 상기 확산방향, 상기 확산 예상범위, 상기 전개될 상황 및/또는 상기 시설물 위험도를 포함하는 재난 이벤트를 관리시스템에 제공하는 재난 이벤트 전파단계(S5);로 이루어진 것을 특징으로 하는, 이기종 다중센서 기반의 재난 상황 예측 및 진단 방법으로 하는 것이 바람직하다.In order to achieve the above object, the present invention analyzes big data collected from heterogeneous multi-sensors, which are of various types and each type consists of a plurality of sensors, by deep learning in an information system to prevent disasters that may occur in facilities. As a method of predicting the occurrence of a disaster and diagnosing whether a disaster has occurred, the type of disaster that has occurred, the rate of spread, and the situation to be developed in stages, using a disaster symptom prediction model, detection intervals collected from each of the heterogeneous multi-sensors

Disaster symptom prediction step (S1) of predicting and learning the signs of a disaster, the type of disaster with a risk of occurrence, and the expected place of occurrence by selecting and analyzing the data generated by the types of sensors related to the precursor symptoms of a disaster among the data. ; The detection section collected from each of the heterogeneous multi-sensors using a disaster type determination model

Among the data of, horizontally clustering the correlation between changes to determine whether a disaster has occurred, the type and location of the disaster that has occurred, and learn, - Detection section in the disaster symptom prediction step

In the case of predicting the occurrence symptom for , the first model among the disaster type determination models is used, and - in the case of not predicting the occurrence symptom, the disaster occurrence detection step using the second model among the disaster type determination models (S2 ); When it is determined that a disaster has occurred in the disaster occurrence detection step (S2), a vertical analysis of time series changes for each of the data collected from each of the heterogeneous multi-sensors and a horizontal analysis of the correlation of the changes are performed using a disaster diffusion prediction model. Through the detection zone

Disaster spread prediction step (S3) of predicting and learning the spread speed and spread direction of disasters in; When it is determined that a disaster has occurred in the disaster occurrence detection step (S2), a risk prediction model is used to predict the spread of the disaster according to the type of the disaster, the place of occurrence, the spread rate, and the spread direction, and the spread of the disaster in stages. A risk prediction step (S4) of inferring the situation to be and predicting and learning the risk of the facility accordingly; And the occurrence symptom of the disaster, the type of disaster with a risk of occurrence, the expected location, whether or not the disaster occurred, the type of the disaster, the place of occurrence, the spread rate, the spread direction, the expected spread range, A disaster event propagation step (S5) of providing a disaster event including the situation to be developed and/or the risk of the facility to the management system; desirable.

In addition to the above-described features, the present invention also uses an ontology-based disaster knowledge database to determine the type of disaster with the risk of occurrence, the type of disaster that has occurred, the place of occurrence, the rate of spread, the direction of spread, and the expected spread. A decision-making information providing step (S6) of providing information for decision-making to the management system for each disaster situation according to the range, the situation to be developed or the risk level of the facility; A digital twin generation step (S7) in which the management system implements the disaster event as a digital twin model as a disaster situation prediction and diagnosis method or in addition thereto; And it is preferable to use it as a diagnostic method.

에 대한 상기 재난발생 감지단계(S2)를 상기 제1모델을 이용하여 수행하는 것을 특징으로 하는, 이기종 다중센서 기반의 재난 상황 예측 및 진단 방법으로 하는 것이 바람직하다.The present invention, in addition to the above-described features, the disaster event propagation step (S5) and the decision-making information providing step (S5) include the disaster symptom prediction step (S1), the disaster occurrence detection step (S2), the A method for predicting and diagnosing a disaster situation based on heterogeneous multi-sensors, characterized in that each is performed when the disaster event is predicted or determined in the disaster spread prediction step (S3) or the risk prediction step (S4), or in addition to the above In the case of predicting the occurrence symptom in the disaster symptom prediction step (S1), - In the disaster event propagation step (S5), an alarm is generated and at the same time, the location information or CCTV screen for a place with a risk of disaster occurrence is managed. It is transmitted to the system to induce intensive monitoring, - the detection section

It is preferable to use a method for predicting and diagnosing a disaster situation based on heterogeneous multi-sensors, characterized in that the disaster occurrence detection step (S2) is performed using the first model.

In addition to the above-described features, the present invention is also a correlation of the change frequency of the data value generated in each of the heterogeneous multiple sensors and the correlation of the change amount or trend according to the rise or fall of the data value, and the disaster The type learning model is a method for predicting and diagnosing disaster situations based on heterogeneous multi-sensors, characterized in that the type of disaster is discriminated and learned by clustering data generated from each of the heterogeneous multi-sensors according to the change correlation, or In addition, the disaster diffusion learning model calculates the change in slope for data obtained as a result of clustering by the disaster type learning model to determine the spread rate of the disaster and the situation to be developed in each step, characterized in that, heterogeneous It is desirable to use a multi-sensor-based disaster situation prediction and diagnosis method.

The present invention is also a heterogeneous multi-sensor-based disaster situation prediction and diagnosis method, characterized in that, in addition to the above-described features, the first model has a lower criterion for the change correlation than the second model, or in addition to this A data normalization step (S0) of normalizing the data collected from each of the heterogeneous multi-sensors through pre-processing before the disaster symptom prediction step (S1); determining the type of each of the multi-sensors and standardizing the structure and terms of data collected from each of the heterogeneous multi-sensors; and - a unit standardization process of defining a number type having the largest number of types as a standard after classifying heterogeneous integers or real numbers for each type of the heterogeneous multi-sensor through dimension reduction. It is also desirable to use a disaster situation prediction and diagnosis method based on the

As described above, in the present invention, the information system uses a disaster symptom prediction model, which is a deep learning learning model, to select and detect data generated from sensor types related to precursor symptoms of disaster occurrence among various data collected from heterogeneous multi-sensors. Since it includes a disaster symptom prediction step (S1) of analyzing and predicting and learning the signs of a disaster, the type of disaster with a risk of occurrence, and the expected place of occurrence, the probability of occurrence of a disaster among the combinations of data collected by sensors that detect facilities If a combination of data with this is captured as a precursor symptom, it can be notified as a sign of a disaster, which has the effect of fundamentally preventing the occurrence of a disaster.

In the present invention, the information system horizontally clusters the correlation of changes among data collected from each of heterogeneous multiple sensors using a disaster type determination model, which is a deep learning learning model, to determine whether a disaster has occurred, the type of disaster that has occurred, and Since it includes a disaster occurrence detection step (S2) that determines and learns the place of occurrence, it is possible to accurately and quickly determine whether a disaster has occurred, so that in the event of a disaster, it is notified promptly in the early stage and responding accordingly. It has the effect of minimizing the damage caused by it.

The present invention also uses two models as disaster type judgment models used in the disaster occurrence detection step (S2). In the case where signs of occurrence are not predicted, the second model is used, and different models are used to determine whether there are signs of occurrence, such as to increase the accuracy of the judgment of disaster occurrence. When determining the type of disaster, collected data can be sensitively judged, and accordingly, there is an effect of increasing the accuracy and sensitivity of determining the occurrence of a disaster.

In addition, when it is determined that a disaster has occurred in the disaster occurrence detection step (S2), the present invention uses a disaster diffusion prediction model for vertical analysis of time series changes for each of the data collected from each of heterogeneous multi-sensors and horizontal analysis for change correlation Since it includes a disaster spread prediction step (S3) of predicting and learning the spread speed and direction of the disaster through enemy analysis, when it is determined that a disaster has occurred, the information system informs the manager of the spread speed of the disaster, so that the It has the effect of enabling quick and accurate decision-making in determining the size and scope of dispatch.

In the present invention, when it is determined that a disaster has occurred in the disaster occurrence detection step (S2), the expected range of the disaster spread according to the type, place of occurrence, spread speed and direction of the disaster using a risk prediction model and Since it includes a risk prediction step (S4) of inferring the situation and predicting and learning the risk of the facility accordingly, when it is determined that a disaster has occurred, the information system infers the expected spread of the disaster and the situation that will develop in stages, It has the effect of predicting and informing the risk level, and accordingly, the disaster manager, etc. can respond quickly in consideration of the expected scope of the spread of the disaster.

The present invention also uses an ontology-based disaster knowledge database to classify disaster situations according to the types of disasters with risk of occurrence, types of disasters that have occurred, place of occurrence, spread rate, spread direction, expected range of spread, situation to be developed or risk of facilities, etc. Since the decision-making information provision step (S6) is included to provide information for decision-making to the management system, managers, etc. can calmly respond to manuals in the event of a disaster, thereby minimizing damage caused by a disaster. There are effects that can be done.

1 shows a conventional disaster monitoring system using a multi-sensor network.
2 is a system configuration diagram for implementing a method for predicting and diagnosing a disaster situation based on heterogeneous multi-sensors according to the present invention.
3 is a flowchart illustrating a method for predicting and diagnosing a disaster situation based on heterogeneous multi-sensors according to the present invention.
4 and 5 are examples of sensor combinations used for predicting symptoms for each type of disaster in the present invention.
6 illustrates a learning process of predicting and diagnosing a disaster situation based on heterogeneous multi-sensors according to the present invention.
7 is a system configuration diagram for implementing another embodiment of the present invention.
8 is a procedure diagram in which another embodiment of the present invention is performed.
9 and 10 are conceptual diagrams of the ontology-based disaster knowledge database included in the present invention.

Hereinafter, the present invention will be described in detail so that the above-described objects and characteristics become clear, and accordingly, those skilled in the art to which the present invention belongs will be able to easily implement the technical spirit of the present invention. In addition, in describing the present invention, among the known technologies related to the present invention, the detailed description is given when it is determined that the detailed description of the known technology may unnecessarily obscure the subject matter of the present invention. to omit

In addition, the terms used in the present invention have been selected from general terms that are currently widely used as much as possible, but in certain cases, there are terms arbitrarily selected by the applicant. It is intended to clarify that the present invention should be understood as the meaning of the term, not the name of. Terms used in the description of the embodiments are only used to describe specific embodiments, and are not intended to limit the embodiments. Singular expressions include plural expressions unless the context clearly dictates otherwise.

Embodiments may be changed in various forms and may have various additional embodiments. Here, specific embodiments are shown in the drawings and related detailed descriptions are described. However, this is not intended to limit the embodiments to a specific form, and it should be understood to include all changes, equivalents, or substitutes included in the spirit and technical scope of the embodiments.

Expressions such as “first”, “second”, “first” or “second” in the description of various embodiments may modify various components of the embodiments, but do not limit the corresponding components. For example, the above expressions do not limit the order and/or importance of corresponding components. The above expressions may be used to distinguish one component from another.

Hereinafter, the present invention will be described with reference to the accompanying drawings. A large population lives in a densely populated city, and large structures such as bridges and tunnels, as well as various infrastructure facilities to provide power, energy supply, and communication to them, as well as large residential or office buildings, are scattered in a complicated manner, resulting in fires and other hazards. The risk of a major disaster still exists. In addition, when disasters occur due to population growth, urban density, and consequent complexity and diversification of facilities, not only are disasters becoming more complex and larger, resulting in greater damage, but also disasters such as sinkholes, building collapses, or landslides in downtown areas. The types of disasters are also diversifying, such as new types of disasters that have not appeared before.

On the other hand, various infrastructures such as roads, bridges, tunnels, as well as various buildings, structures, and infrastructures in the downtown area are equipped with various infrastructures that can manage them and monitor their status for smooth operation or to ensure safety such as crime prevention or disaster occurrence. Many monitoring devices such as sensors and video equipment are installed. Therefore, in recent years, a hyper-connected network environment through linkage of data collected from these existing monitoring devices, Internet of Things (IoT), big data, and artificial intelligence-based predictive analysis environments are configured as a platform structure to respond to disasters. Strategies are presented.

In addition, a procedure that supports the establishment of a much more effective and flexible contingency plan by analyzing various types of data and running simulations to complement them is also being studied. For example, in an effort to configure the national disaster safety management system as a digital twin, digital twin platform construction and research for disaster safety management of underground utility conduits are also being conducted. This is because it can be linked with various national facilities and disasters through quick response and modeling of national disasters that occur in a specific space. The digital twin disaster safety management platform applies digital twin disaster prediction data to visualize the digital twin model of underground utility tunnels by applying a procedure that expresses and integrates information collected in the field, spatial information, modeling and simulation information, and agency manual information by hierarchy. generate Then, it manages time and space with the created digital twin model and applies shape management for disaster safety management to the platform.

As described above, in general, as shown in FIG. 1, in urban space, numerous monitoring means such as sensors and video equipment are connected through a communication network, and the data collected by the numerous monitoring means is integrated through a communication network where experts reside. It is provided to the control center to be monitored, and disaster occurrence is automatically monitored through big data analysis. And when it is judged that a disaster has occurred, it contacts the department related to disaster suppression such as the fire department so that suppression personnel can be dispatched, and if necessary, intelligent robots are dispatched. However, these disaster response systems have limitations in the use of data collected from related sensors or video equipment, so they are highly dependent on surveillance personnel, and in the case of big data analysis, they analyze specific types of sensor data or video data. It is true that there are limits.

Therefore, in the present invention, all possible types of sensors or monitoring installed for the purpose of management or monitoring or data acquisition of various facilities such as underground utility tunnels, tunnels, bridges, roads, power facilities, energy supply facilities, water supply and sewage facilities, etc. After collecting the data collected through the equipment and analyzing it by deep learning, predict the signs of a disaster that may occur within the facility and diagnose whether or not a disaster has occurred, the type of disaster that has occurred, the speed of spread, and how to diagnose the situation that will develop in stages. to provide.

2 shows a configuration diagram of an embodiment of the present invention. As shown in FIG. 2, in the present invention, from 'heterogeneous multi-sensors 100' composed of "various types and a plurality of sensors or measuring devices for each type" that collects data within each facility 10. It is preferable to have the information system 300 analyze the collected big data by deep learning. The 'heterogeneous multi-sensor 100' included in the present invention includes, for example, various sensors such as a temperature sensor, a humidity sensor, an air pressure sensor, a vibration sensor, a motion sensor, and a noise sensor, as well as an ammeter, a voltmeter, a harmonic meter, and an electromagnetic wave sensor. It is preferable to include various measuring instruments such as a measuring instrument, a pressure gauge, a flowmeter, a current meter, and an earth leakage measuring instrument, and to configure a plurality of each type of sensor or measuring instrument.

For example, in the case of various power supply facilities such as power facilities of unmanned substations or power cables, connectors, and cutoff devices installed in underground utility tunnels, the basic power supply status such as voltage, current amount, amount of power, power factor, and leakage current, and abnormalities in power lines or devices It not only includes a plurality of various instruments that can monitor whether or not this occurs, but also includes various sensors to measure basic conditions such as temperature, humidity, and vibration in underground cavities or substations, and also includes fire detection, gas leak, etc. Video surveillance equipment such as sensors and cctv in preparation for safety accidents are also installed. In addition, sensors for detecting visitors, such as motion sensors, noise sensors, and sensors monitoring door opening and closing, are also installed. In the present invention, it is preferable to collect all the data measured by all these measuring devices and sensors installed in the facility and use them to determine signs and occurrences of disasters. Hereinafter, in the description and claims of the present invention, the heterogeneous multi-sensor 100 is used as a concept including all of such sensors, measuring devices, and monitoring devices.

In general, in order to collect and utilize the information collected from the heterogeneous multi-sensor 100, it is preferable to configure each sensor network 101 for each type of sensor or measuring device. And, in the present invention, it is preferable to further include a multi-sensor gateway 200 capable of collecting information collected from all of the respective sensor networks 101 into one. That is, in the present invention, in the heterogeneous multi-sensor 100, the data collected for each type of sensor or measuring device (100a, 100b, 100c) is collected through the respective sensor networks 101a, 101b, and 101c, and the multi-sensor gateway In step 200, it is preferable to collect data collected and sent by each of the sensor networks 101a, 101b, and 101c and send them to the information system 300 for big data analysis.

In addition, the information system 300 preferably performs data pre-processing for stabilization of data standard and quality with respect to the collected data. The information system 300 first determines the heterogeneous multi-sensor 100 according to the type of sensor or measuring device for data pre-processing, and standardizes the structure and terminology of data transmitted for each type of the heterogeneous multi-sensor 100. It is desirable to proceed with normalization based on the criterion. And, as for the unit of normalization for data preprocessing, it is preferable to classify heterogeneous integers or real numbers through dimensionality reduction according to the type of the heterogeneous multi-sensor 100 and to define the number type having the most types as a standard. In addition, it is desirable to standardize terms used in data structures by similarity or function and apply them to unification of data and normalization of numeric data. Accordingly, normalization is performed to ensure quality based on converted numeric data of the most number types, 3D location information, and time.

In addition, the information system 300 performs preprocessing on the data collected from the heterogeneous multi-sensors 100, and then determines whether there is a sign of a disaster based on the standardized data, the type of disaster to occur, and the actual disaster It quickly predicts the stage of disaster, such as predicting whether it has occurred and the degree of progress. That is, the information system 300 pre-processes the data collected from the heterogeneous multi-sensors 100 and then deep learning analyzes the pre-processed data to predict signs of occurrence of disasters that may occur within the facility 10, and It is desirable to diagnose whether the occurrence of the disaster, the type of disaster that occurred, the rate of spread, and the situation to be developed in stages are transmitted to the management system 400.

3 is a flowchart of a process in which a method for predicting and diagnosing a disaster situation based on heterogeneous multi-sensors according to the present invention is performed. As described above, in the present invention, the information system 300 analyzes the big data collected from the heterogeneous multi-sensors 100, which are of various types and each type consists of a plurality of sensors, by deep learning, It is about a method for predicting signs of occurrence of disasters that may occur in the future and diagnosing whether or not a disaster has occurred, the type of disaster that has occurred, the rate of spread, and the situation that will develop in stages.

To this end, first, as shown in FIG. 3, when the heterogeneous multi-sensors 100 transmit various data they each sense or measure to the multi-sensor gateway 200 through the sensor network 101 (s101), Preferably, the multi-sensor gateway 200 collects and sends them to the information system 300 (s102). Then, the information system 300 preferably performs a data normalization step (S0) of normalizing the data transmitted from the multi-sensor gateway 200 through preprocessing (s103). In the data normalization step (S0), the type of each of the heterogeneous multi-sensors 100 is determined, and a process of standardizing the structure and terms of data collected from each of the heterogeneous multi-sensors 100 is performed, and then the heterogeneous multi-sensors 100 are standardized. It is more preferable to perform a unit standardization process of defining a number type having the most types as a standard after distinguishing heterogeneous integers or real numbers through dimension reduction for each type of the multi-sensor 100 .

(i는 정수)의 데이터 중에서 재난 발생의 전조증상과 관련되는 센서 종류에서 발생하는 데이터를 선별 및 분석하여 재난의 발생 징후, 발생 위험이 있는 재난의 종류 및 발생 예상 장소를 예측하고 학습하는 재난징후 예측단계(S1)를 수행하도록 하는 것이 바람직하다(s110). 여기서 상기 탐지구간

는 분석할 데이터를 재난발생 전조증상과의 연관성을 판단하기 위한 시계열 상의 탐지구간이다. 따라서 상기 재난징후 예측모델은 상기 탐지구간

에서 상기 이기종 다중센서(100) 각각에서 수집된 데이터를 탐지구간

또는 탐지구간

(n은 정수)의 데이터들과 비교하는 것도 가능하다. 그리고 상기 탐지구간

와 상기 탐지구간

는 서로 겹치는 시간대가 있도록 하는 것도 가능하다.After performing the data normalization step (S0), the information system 300 uses a disaster symptom prediction model (s115) to detect detection intervals collected from each of the heterogeneous multi-sensors 100.

Disaster sign that predicts and learns signs of occurrence of disasters, types of disasters with risk of occurrence, and predicted locations by selecting and analyzing data generated from sensor types related to precursor symptoms of disasters among data of (i is an integer) It is preferable to perform the prediction step (S1) (s110). Here, the detection interval

is the detection interval on the time series to determine the correlation of the data to be analyzed with the precursor symptoms of a disaster. Therefore, the disaster symptom prediction model is the detection interval

In the detection period, the data collected from each of the heterogeneous multi-sensors 100

or detection interval

It is also possible to compare data of (n is an integer). And the detection period

and the detection interval

It is also possible to have time zones that overlap each other.

As described above, the disaster symptom prediction model does not use all of the data collected from each of the heterogeneous multiple sensors 100, but rather selects and analyzes data generated from sensor types related to the precursor symptoms of disaster occurrence to prevent disasters. It is desirable to predict symptoms. 4 and 5 show examples of sensors used for predicting signs and detected phenomena for each type of disaster.

For example, in the case shown in FIG. 4, collisions or collisions of cars in a tunnel often lead to disasters such as large-scale fires. Therefore, when a car collision or collision accident occurs in a tunnel, it is desirable to predict it as one sign of a fire disaster because there is a high possibility of a fire disaster occurring in the tunnel. Therefore, in the disaster symptom prediction step (S1) using the disaster symptom prediction model, the information system 300 has no 119 report or received information that an accident has occurred in the tunnel, but noise among the heterogeneous multi-sensors 100 If the detection sensor detects an abnormal noise corresponding to an impact or frictional sound, and the vehicle speed at the exit outside the tunnel is normal among the data detected by the speed sensor, but the vehicle speed in the tunnel is rapidly decreasing, an impact accident has occurred in the tunnel. Since there is a possibility of a fire due to a vehicle collision, this can be predicted as a sign of a disaster.

In addition, when analyzing the cctv screen attached to the tunnel entrance, it can be determined that the signs of a fire disaster have increased if a vehicle loaded with dangerous goods has entered the tunnel. In addition, if a movement of a person other than a vehicle is detected, since people got off the vehicle due to an accident, this can also be used as data to be judged as a sign of a disaster in the tunnel. Therefore, in order to capture the signs of disaster in the tunnel, among the heterogeneous multi-sensors 100 installed in the tunnel, a noise detection sensor in the tunnel, a vehicle speed measurer installed at the entrance and exit of the tunnel, a cctv camera installed at the tunnel entrance, and a tunnel By analyzing the data collected from my motion detection sensor with the disaster sign prediction model, it will be possible to predict the signs of disaster and the risk of fire accidents in the tunnel.

As another example, as shown in FIG. 5, there is a possibility of electrical fire due to power facilities in an underground utility or an unmanned substation. In particular, there is a possibility of a fire due to an arc discharge in power facilities such as power cables installed in underground utility tunnels or unmanned substations, and if a fire breaks out in an underground utility tunnel, there is a high possibility that it will develop into a fairly wide range of fires, and there is a high possibility that it will develop into an unmanned substation or residential power facility. When a fire breaks out in the back, there is a possibility that it will develop into a large-scale disaster such as a forest fire. That is, when a contact defect or disconnection occurs in a power facility, an arc discharge occurs at that part, and the arc discharge is accompanied by sparks, flashes, and heat. However, in most cases, even if an arc discharge occurs, it is difficult to detect the arc discharge at an early stage because it does not immediately lead to a fire. However, if the arc discharge is left unattended, it leads to deterioration of the insulation part, heat generation caused by it, and thermal runaway phenomenon, which causes a fire to spread to the surroundings and eventually develop into a large-scale fire.

However, when the disaster symptom prediction step (S1) of analyzing the data of the heterogeneous multi-sensor 100 is performed using the disaster symptom prediction model according to the present invention, the possibility of a fire occurring even if a fire does not occur is reduced. You can find out the signs of an arc discharge you have and take action at an early stage. That is, when arc discharge occurs, harmonics increase and electromagnetic waves increase at the same time, and since arc discharge can accompany heat and flash, among the data of the heterogeneous multi-sensor 100, the harmonics measurer, the electromagnetic wave measurer, the temperature sensor, the illuminance sensor, etc. It is possible to predict the sign of a fire caused by a power facility by using the change value of the generated data. That is, if the harmonics increase and the electromagnetic waves increase at the same time, it is determined that an arc discharge occurs, and if the surface temperature of the enclosure or cable rises or even a change in illumination is detected, this can be predicted as the immediately preceding stage of a fire.

As such, the present invention selects and combines the data of the heterogeneous multi-sensor 100 that may occur as a precursor to a disaster, as illustrated in FIGS. 4 and 5, and analyzes them using the disaster symptom prediction model. It is possible to predict the symptoms of a disaster, the type of disaster with a risk of occurrence, and the expected place of occurrence. In addition to this, the prediction accuracy can be further increased by the disaster symptom prediction model learning using the prediction result. And if it is determined whether a disaster has occurred in the subsequent disaster detection step (S2), the disaster symptom prediction model can learn whether or not an actual disaster has occurred. And it is possible to more elaborately predict the expected place of occurrence.

In addition to this, there are various factors that can be signs of occurrence of fire disasters that may occur in facilities such as underground common pits. That is, arson or misfire due to intrusion by outsiders, overload of power facilities, static electricity, electrical discharge, brain (lightning), etc. can also be causes, and smoking or bringing in inflammable materials by workers or outsiders, overheating of electric heaters or lights, It may be caused by a variety of causes, such as static electricity, welding, and oil vapor generation due to maintenance work. In the present invention, by including and learning various possibilities of such occurrence symptoms in the disaster symptom prediction model, it will be possible to predict disaster symptoms quickly and precisely.

의 데이터 중에서 상호 간의 변화 연관성을 수평적으로 클러스터링하여 재난의 발생 여부, 발생한 재난의 종류 및 발생 장소를 판단하고 학습하는 재난발생 감지단계(S2)를 수행하도록 하는 것이 바람직하다(s130, s131). 상기 재난징후 예측단계(S1)에서 적용되는 딥러닝 모델인 상기 재난징후 예측모델에서는 상기 이기종 다중센서(100) 각각에 수집되는 데이터 중에서 재난 발생의 전조증상과 관련되는 센서 종류에서 발생하는 데이터만 선별하여 분석했으나, 상기 재난발생 감지단계(S2)에 적용되는 딥러닝 모델인 상기 재난유형 판단모델에서는 상기 이기종 다중센서(100) 각각에 수집되는 데이터 모두를 사용하여 데이터들 상호 간의 변화 연관성을 수평적으로 클러스터링하는 것이 바람직하다.On the other hand, as shown in FIG. 3, after the information system 300 performs the disaster symptom prediction step (S1), the detection collected from each of the heterogeneous multi-sensors 100 is performed using a disaster type determination model. section

It is preferable to perform a disaster detection step (S2) of determining whether a disaster has occurred, the type and location of the disaster, and learning by horizontally clustering the correlation of changes among the data of (s130, s131). In the disaster symptom prediction model, which is a deep learning model applied in the disaster symptom prediction step (S1), only data generated from sensor types related to precursor symptoms of disaster occurrence are selected from among the data collected by each of the heterogeneous multi-sensors 100. However, in the disaster type determination model, which is a deep learning model applied to the disaster occurrence detection step (S2), all of the data collected by each of the heterogeneous multiple sensors 100 is used to horizontally correlate changes between data It is preferable to cluster with .

Here, the change correlation is the correlation between the frequency of change of the data value collected from each of the heterogeneous multiple sensors 100 and the correlation between the change amount or trend according to the rise or fall of the data value, and the disaster type learning model It is preferable to discriminate and learn the type of disaster by clustering data collected from each of the heterogeneous multi-sensors 100 according to change correlation.

As an example of the change correlation, when the data values of the temperature sensor and the humidity sensor among the heterogeneous multi-sensors 100 in the underground utility tunnel are analyzed for the amount of change or the correlation of the trend according to the rise or fall, from before a certain point in time from the present point in time. If the temperature rises and the humidity decreases, the temperature rises due to the fire and the humidity decreases due to the heat, so it can be determined that the fire has occurred in the underground common port. In addition, if the illuminance is lowered, it may be that the illuminance is lowered due to the smoke caused by the fire, so it can be a more reliable basis for determining the occurrence of a fire. In addition, gas detection also serves as judgment data for fire occurrence. In addition, fire detection by automatic fire detection equipment will be important data for determining fire occurrence. In addition, if data from temperature and humidity sensors are added, it will be possible to quickly and accurately determine the occurrence of fire in the underground common tunnel.

In addition, taking a fire in a tunnel during a disaster, for example, if the temperature value rises in the temperature sensor among the heterogeneous multiple sensors 100 and the gas concentration rises in the gas sensor, it will be able to determine that a fire has occurred. If the sensor detects an abnormal noise corresponding to an impact or frictional sound, and the vehicle speed at the exit outside the tunnel is normal among the data detected by the speed sensor, but the vehicle speed in the tunnel is rapidly decreasing, it is due to a vehicle collision in the tunnel. It can be judged that a fire has occurred. In addition, if the illuminance is lowered by the illuminance sensor and the movement of a person other than the vehicle is detected by the motion sensor, this can also be used as data to determine that a fire has occurred in the tunnel because people got off the vehicle due to a fire accident.

As such, in the present invention, using the disaster type determination model, among the data collected from each of the heterogeneous multi-sensors 100, horizontal clustering of correlations between changes is performed to determine whether a disaster has occurred, the type and location of the disaster, and In this case, the data that can be collected from the heterogeneous multi-sensor 100 will be very diverse. For example, vibration, temperature, wind speed, illuminance, humidity, speed, acceleration, precipitation, rainfall, snowfall, water leakage, elasticity, density, air pressure, frictional force, slippage, roughness, tension, force, distance, color, color temperature, Contact count, resistance, turbidity, concentration, viscosity, adhesiveness, water level, flow rate, flow rate, air volume, wind direction, current, voltage, power, wattage, number of revolutions, dryness, tensile force, tension, pressure, bonding force, strength, elasticity, The data that can be measured, such as vacuum, slope, centrifugal force, centripetal force, angle, refractive index, reflectance, and salinity, are quite diverse, and in the present invention, the disaster type determination model horizontally clusters the correlation between these various data It determines whether a disaster has occurred, the type of disaster that has occurred, and the place where it occurred and learns.

에 대하여 재난의 발생 징후를 예측한 경우에는 상기 재난유형 판단모델 중 제1모델을 이용하도록 하고(s135), 상기 재난징후 예측단계에서 재난의 발생 징후를 예측하지 아니한 경우에는 상기 재난유형 판단모델 중 제2모델을 이용하도록 하는 것이 바람직하다(s136). 여기서 상기 제1모델은 상기 제2모델보다 상기 변화 연관성에 대한 판단기준을 낮게 적용하는 것이 더욱 바람직하다. 즉, 제1모델은 상기 재난징후 예측단계에서 재난의 징후를 예측한 이후에 재난의 발생 여부를 판단하는 것이어서 재난의 발생 가능성이 높은 상태이므로 판단의 민감도를 더 높이기 위하여 그 판단에 있어서 다소 변화 연관성이 낮더라도 재난 발생으로 판단하게 하는 것이다. On the other hand, as shown in FIG. 3, the disaster type judgment model applied in the disaster occurrence detection step (S2) uses a disaster type judgment model of a different type according to the prediction result of the previously performed disaster symptom prediction step (S1). It is desirable to apply That is, the detection interval in the disaster symptom prediction step

In the case of predicting a symptom of a disaster, the first model among the disaster type determination models is used (s135), and if the symptom of a disaster is not predicted in the disaster symptom prediction step, among the disaster type determination models It is preferable to use the second model (s136). Here, it is more preferable that the first model applies a lower criterion for the change correlation than the second model. That is, since the first model determines whether a disaster will occur after predicting the signs of a disaster in the disaster symptom prediction step, the probability of occurrence of a disaster is high. Therefore, in order to further increase the sensitivity of the judgment, the correlation of some changes in the judgment Even if it is low, it is judged as a disaster.

뿐만이 아니라 상기 탐지구간

에 대한 상기 재난발생 감지단계(S2)도 상기 제1모델을 이용하여 수행하도록 하는 것이 바람직한데, 이는 상기 재난징후 예측단계(S1)에서 상기 발생 징후를 예측하였기 때문에 재난의 발생 가능성이 높을 것이므로 당분간 민감도를 높여서 감시하도록 하는 것이 바람직하기 때문이다.And, when the occurrence symptom is predicted in the disaster symptom prediction step (S1), the detection interval

In addition, the detection interval

It is preferable to perform the disaster occurrence detection step (S2) for the disaster occurrence using the first model as well. Since the occurrence symptom is predicted in the disaster symptom prediction step (S1), the possibility of occurrence of the disaster is high, so for the time being This is because it is desirable to monitor with increased sensitivity.

에서의 재난의 확산속도 및 확산 방향을 예측하고 학습하는 재난확산 예측단계(S3)를 수행하도록 하는 것이 바람직하다(s140).On the other hand, as shown in FIG. 3, when it is determined that a disaster has occurred in the disaster occurrence detection step (S2) (s140), using a disaster diffusion prediction model, each of the data collected from each of the heterogeneous multi-sensors 100 Detection interval through vertical analysis of time series change for and horizontal analysis of correlation of change

It is preferable to perform the disaster spread prediction step (S3) of predicting and learning the spread speed and spread direction of the disaster in (s140).

It is preferable that the disaster diffusion learning model calculates a change in slope for data acquired as a result of clustering by the disaster type learning model to determine the spread rate of the disaster and the situation to be developed in each step. In other words, in the case of the fire in the underground utility tunnel mentioned above, the vertical analysis of the time-series change for each data is the time-series change of data values such as how fast the temperature rises, and through this, how quickly the severity of the fire You can predict what's going on. In addition, horizontal analysis of change correlation can predict the physical range and direction in which the fire spreads using the difference and change trend of the measured values measured by each of the temperature sensors that are separated from each other.

In addition, when it is determined that a disaster has occurred in the disaster occurrence detection step (S2) (s140), the judgment result in the disaster occurrence detection step (S2) and the disaster spread prediction step (S3) using a risk prediction model ) and the data collected from each of the heterogeneous multi-sensors 100 are analyzed, and the expected range of disaster spread according to the type of disaster, the place of occurrence, the spread speed, and the spread direction, and the It is desirable to infer the situation and to perform the risk prediction step (S4) of estimating and learning the risk of the facility accordingly (s160).

Here, it is preferable that the facility risk level expresses the level of diffusion according to the type of disaster, and the facility risk level is defined using slope change prediction data based on the response level for each type of disaster defined by the government. It is preferable to define the steps by dividing the risk level of the facility into quantitative risk judgment and qualitative risk judgment. It is desirable to converge the risk level by mapping . In addition, it is desirable to determine the risk level by mapping the risk response manual defined in the standard operation manual for each type of disaster to determine the qualitative risk level.

And, when the occurrence symptom of the disaster, the type of the disaster with the risk of occurrence, and the expected place of occurrence are predicted in the disaster symptom prediction step (S1), the disaster event propagation step of providing the disaster event to the management system (S5 ) While performing (s190), it is preferable to generate an alarm and simultaneously transmit location information or CCTV screens for places with a risk of disaster to the management system 400 to induce intensive monitoring.

In addition, when the disaster occurrence detection step (S2) determines whether a disaster has occurred, the type of disaster that has occurred, and the location of the disaster, and the disaster spread prediction step (S3) predicts the spread speed of the disaster and the direction of the disaster Disaster event propagation step (S5 ) is preferably performed (s190).

And, as shown in FIG. 6, the results predicted or determined in the disaster symptom prediction step (S1), the disaster occurrence detection step (S2), the disaster spread prediction step (S3), and the risk prediction step (S4) are described above. When there is confirmation of the management system 400, that is, when the correctness and erroneousness of the prediction and judgment result is confirmed (s200), the result is the disaster symptom prediction model, the disaster type judgment model, and the disaster spread prediction model. And it is preferable to perform a procedure for learning by feeding back to the risk prediction model.

7 shows another embodiment of a system configuration diagram for performing the disaster situation prediction and diagnosis method based on heterogeneous multi-sensors according to the present invention. 8 shows a flow chart of a procedure by which this embodiment is performed. Hereinafter, the present invention will be described with reference to FIGS. 7 and 8 .

As shown in FIG. 7, this embodiment is an embodiment further including an ontology-based disaster knowledge database 500 in addition to the configuration according to the above-described embodiment. Therefore, in this embodiment, as shown in FIG. 8, the information system 300 uses the ontology-based disaster knowledge database 500 to perform the disaster symptom prediction step (S1), the disaster occurrence detection step (S2), A decision-making information providing step (S6) of providing information for decision-making for each situation to the management system 400 with respect to the results predicted or judged in the disaster spread prediction step (S3) and the risk prediction step (S4). It is desirable to further include (s310 to s320).

Accordingly, the information system 300 uses the ontology-based disaster knowledge database 500 to determine the type of disaster that has a risk of occurrence, the type of disaster that has occurred, the place of occurrence, the rate of spread, the direction of spread, the expected range of spread, and the future Information for decision-making for each disaster situation according to the situation to be developed or the risk level of the facility can be provided to the disaster manager through the management system 400 .

9 and 10 show conceptual diagrams of the disaster knowledge database 500 included in the present invention. As shown in FIG. 9, the disaster knowledge database 500 for determining the disaster situation and presenting predicted information step by step constructs knowledge information for each situation, step, and predicted response of the disaster into an ontology system to predict and determine the situation. It is desirable to be able to provide information. This is because a situation analysis and supporting knowledge system that is responsible for the safety and job performance of workers performing rescue and recovery work and management in a state of high understanding of the site is required. Therefore, it is desirable to provide the function of defining the ontology specification logic configuration and reasoning relationship that minimizes risk by providing quick situational judgment and knowledge information in a high-risk situation in an unsafe disaster site.

On the other hand, OWL-DL assigns symbolic names to components to infer the expression of actions and configures specification logic including basic hierarchical relationships. As shown in FIG. 10, the ontology disaster anomaly situation reasoning knowledge model uses OWL-DL, which configures standard operation manual-based description logic to classify actions from data of heterogeneous multi-sensors at disaster sites, to provide detailed responders and related persons. It is desirable to allow knowledge information to be inferred. At this time, the class should have an expression range for determining the relationship between the horizontal abnormal situation prediction evaluated for the disaster anomaly situation processed above, and the instance preferably configures the prediction distribution for the vertical abnormal situation prediction as a time series distribution.

And, in the present invention, it is also preferable that the management system 400 further include a digital twin generation step (S7) of implementing the disaster event as a digital twin model. In this way, the predicted data produced can confirm the expected spread time and size for each type of disaster, and through this, it is possible to simulate changes according to the type of disaster with a digital twin model. In addition, it is desirable that the generated digital twin data be linked to the on-site situation room and smart terminal using heterogeneous multiple devices to provide a service linked to spatial data that provides a 3D digital twin model for disaster safety management functions. Then, by using the smart terminal, it is possible to quickly determine the personal safety and abnormal situation of workers at the disaster site on the disaster communication network (PS-LTE), and it can be linked with the national disaster safety management system.

Although the present invention has been described with the various examples described above, the present invention is not necessarily limited to these examples, and may be variously modified and implemented without departing from the technical spirit of the present invention. Therefore, the examples disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain, and the scope of the technical idea of the present invention is not limited by these examples. The protection scope of the present invention should be construed according to the following claims, and all technical ideas within the equivalent scope should be construed as being included in the scope of the present invention.

10 Facilities
100 Heterogeneous multi-sensors
101 sensor network
200 Multisensor Gateway
300 Information Systems
400 management system
500 Disaster Knowledge Database

Claims (9)

The information system deep learning analyzes the big data collected from heterogeneous multi-sensors, which are of various types and each type consists of a plurality of sensors, to predict signs of disasters that may occur within the facility, and to determine whether or not a disaster has occurred and what occurred As a method of diagnosing the type of disaster, the rate of spread and the situation to be developed in stages,
Detection intervals collected from each of the heterogeneous multi-sensors using a disaster symptom prediction model

Disaster symptom prediction step (S1) of predicting and learning the signs of a disaster, the type of disaster with a risk of occurrence, and the expected place of occurrence by selecting and analyzing the data generated by the types of sensors related to the precursor symptoms of a disaster among the data. ;
The detection section collected from each of the heterogeneous multi-sensors using a disaster type determination model

By horizontally clustering the correlation of changes among the data of the data, whether a disaster has occurred, the type of disaster that has occurred, and the location of the occurrence are judged and learned,
- Detection section in the disaster symptom prediction step

In the case of predicting the occurrence symptom for , the first model among the disaster type determination models is used,
- If the occurrence symptom is not predicted, a disaster occurrence detection step (S2) using a second model among the disaster type judgment models;
When it is determined that a disaster has occurred in the disaster occurrence detection step (S2), a vertical analysis of time series changes for each of the data collected from each of the heterogeneous multi-sensors and a horizontal analysis of the correlation of the changes are performed using a disaster diffusion prediction model. Through the detection zone

A disaster spread prediction step (S3) of predicting and learning the spread speed and spread direction of disasters in;
When it is determined that a disaster has occurred in the disaster occurrence detection step (S2), a risk prediction model is used to predict the spread of the disaster according to the type of the disaster, the place of occurrence, the spread rate, and the spread direction, and the spread of the disaster in stages. A risk prediction step (S4) of inferring the situation to be and predicting and learning the risk of the facility accordingly; and
The symptom of occurrence of the disaster, the type of disaster with a risk of occurrence, the expected location, whether or not the disaster occurred, the type of the disaster, the place of occurrence, the spread rate, the spread direction, the expected spread range, the Disaster event propagation step (S5) of providing a disaster event including the situation to be developed and / or the risk of the facility to the management system; characterized in that it consists of a heterogeneous multi-sensor based disaster situation prediction and diagnosis method According to claim 1,
Using the ontology-based disaster knowledge database, the type of disaster with the risk of occurrence, the type of disaster that occurred, the place of occurrence, the spread rate, the spread direction, the expected range of spread, the situation to be developed, or the risk level of the facility A method for predicting and diagnosing a disaster situation based on heterogeneous multi-sensors, characterized in that it further includes; According to claim 2,
Characterized in that the management system further comprises a digital twin generation step (S7) of implementing the disaster event as a digital twin model; heterogeneous multi-sensor based disaster situation prediction and diagnosis method According to claim 2 or 3,
The disaster event propagation step (S5) and the decision-making information providing step (S5) include the disaster symptom prediction step (S1), the disaster occurrence detection step (S2), the disaster spread prediction step (S3) or the risk prediction step Disaster situation prediction and diagnosis method based on heterogeneous multi-sensors, characterized in that each is performed when the disaster event is predicted or determined in step (S4) According to any one of claims 1 to 3,
When the occurrence symptom is predicted in the disaster symptom prediction step (S1),
- At the same time as generating an alarm in the disaster event propagation step (S5), location information or CCTV screens for places with a risk of disaster occurrence are transmitted to the management system to induce intensive monitoring,
- The above detection section

Disaster situation prediction and diagnosis method based on heterogeneous multi-sensors, characterized in that the disaster detection step (S2) for delete delete According to any one of claims 1 to 3,
Disaster situation prediction and diagnosis method based on heterogeneous multi-sensors, characterized in that the first model has a lower criterion for the change correlation than the second model According to any one of claims 1 to 3,
A data normalization step (S0) of normalizing through pre-processing of the data collected from each of the heterogeneous multi-sensors prior to the disaster symptom prediction step (S1); further comprising,
The data normalization step (S0),
- Determining the type of each of the heterogeneous multi-sensors and standardizing the structure and terms of data collected from each of the heterogeneous multi-sensors; and
- A unit standardization process of defining a number type having the most types as a standard after classifying heterogeneous integers or real numbers through dimensionality reduction for each type of the heterogeneous multi-sensor; characterized in that it comprises a heterogeneous multi-sensor based Disaster situation prediction and diagnosis method

Images