Disaster risk evaluation device, disaster risk evaluation method, and program

Abstract

To provide a technique for predicting occurrence of earthquakes, and evaluating a disaster risk for various facilities due to earthquake occurrence.SOLUTION: A disaster risk evaluation device 1 includes as functions thereof: an acquisition section 31 for acquiring earthquake occurrence past record map data which indicates that an earthquake has occurred in a range of a specific region, in a first period of the past; an inference section 32 for reasoning that an earthquake will occur in the future, using the earthquake occurrence past record map data as input, on the basis of a mechanical learning model 23 which uses past record data indicating past earthquake occurrence regions in time series, and foundation data for each region, as teacher data; an earthquake prediction output section 33 for outputting earthquake occurrence prediction map data which indicates that an earthquake will occur in a specific region in a second period in the future, on the basis of an inference result; and a disaster risk evaluation section 34 for evaluating a disaster risk of a specific facility in the second period, in accordance with an earthquake occurrence of the inference result, and for outputting an evaluation result.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a disaster risk assessment device, a disaster risk assessment method, and a program for predicting the occurrence of an earthquake and assessing the risk of damage to a specified facility due to the occurrence of an earthquake. [Background technology]

[0002] Earthquakes are disasters that cause widespread damage to the general public, business operators of various businesses, etc., and large-scale earthquakes can cause enormous damage and endanger lives. Therefore, various research and development has been carried out on technology to predict future earthquakes based on various data obtained from past earthquakes, and this technology is used for earthquake emergency alerts, etc.

[0003] One known earthquake prediction technology is a seismic motion evaluation model that incorporates new observational data and simulation results obtained each time an earthquake occurs into a database, making it possible to effectively utilize big data for evaluating and predicting seismic motion (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2022-047650 Summary of the Invention [Problem to be solved by the invention]

[0005] While predicting the occurrence of earthquakes is important, it is also important for business operators to assess the degree of risk of damage, etc. to the various facilities they manage and use due to the occurrence of an earthquake.

[0006] Therefore, an object of the present invention is to provide a disaster risk assessment device, a disaster risk assessment method, and a program for predicting the occurrence of an earthquake and assessing the risk of damage to various facilities due to the occurrence of an earthquake. [Means for solving the problem]

[0007] In order to solve the above problem, the invention of claim 1 is a damage risk assessment device that predicts the occurrence of an earthquake and evaluates the risk of damage to a specified facility due to the occurrence of an earthquake, characterized in that it comprises: an acquisition unit that acquires earthquake occurrence record map data showing on a map that an earthquake occurred in a specified area during a first period, which is a specified period in the past; an inference unit that uses the earthquake occurrence record map data as an input and makes an inference about the occurrence of an earthquake in the specified area in the future based on a machine learning model that uses as training data the record data showing areas where past earthquakes have occurred in time series and ground data for each area; an earthquake prediction output unit that outputs earthquake occurrence prediction map data showing on a map the prediction result of an earthquake occurring in the specified area during a second period, which is a specified period in the future, based on the result of the inference; and a damage risk assessment unit that evaluates the risk of damage to facilities installed in the specified area during the second period due to the occurrence of an earthquake based on the result of the inference and outputs the evaluation result.

[0008] The invention of claim 2 is characterized in that, in the disaster risk assessment device described in claim 1, it is provided with a learning unit that uses the earthquake occurrence record map data acquired in time series as teacher data, performs machine learning corresponding to the time series, and generates the machine learning model based on the results of the machine learning.

[0009] The invention of claim 3 is characterized in that, in the disaster risk assessment device described in claim 2, the learning unit performs machine learning using the earthquake occurrence prediction map data for the second period and the earthquake occurrence record map data for the second period as teacher data, and updates the machine learning model based on the results of the machine learning.

[0010] The invention of claim 4 is characterized in that, in the disaster risk assessment device described in claim 1, the disaster risk assessment unit evaluates the disaster risk of facilities installed in the specified area during the second period based on facility data that is pre-stored as data regarding facilities installed in the specified area and the result of the inference.

[0011] The invention of claim 5 is characterized in that, in the disaster risk assessment device described in any of claims 1 to 4, the earthquake occurrence history map data is image data showing, on a map showing the specified area, the locations where earthquakes occurred during the first period and data related to the earthquakes.

[0012] The invention of claim 6 is characterized in that, in the disaster risk assessment device described in claim 5, the earthquake occurrence history map data is image data that shows the magnitude of the earthquake or the degree of shaking caused by the earthquake as a size of a circle on a map showing the specified area, with the location where an earthquake occurred during the first period as its center.

[0013] The invention of claim 7 is characterized in that, in the disaster risk assessment device described in any of claims 1 to 4, the ground data includes information regarding one or more of the geology, topography, active faults, trenches, and plate structures of each region.

[0014] The invention of claim 8 is a damage risk assessment method executed by a computer having a processor and a memory, for predicting the occurrence of an earthquake and assessing the risk of damage to a specified facility due to the occurrence of an earthquake, characterized in that the processor executes the following steps: acquiring earthquake occurrence record map data showing on a map that an earthquake occurred within a specified area during a first period, which is a specified period in the past; using the earthquake occurrence record map data as input, inferring the occurrence of an earthquake within the specified area in the future based on a machine learning model using record data showing areas where earthquakes have occurred in the past in chronological order and ground data for each area as training data; outputting earthquake occurrence forecast map data showing on a map the predicted results of an earthquake occurring within the specified area during a second period, which is a specified period in the future, based on the result of the inference; and assessing the risk of damage due to an earthquake occurring to facilities located in the specified area during the second period, based on the result of the inference, and outputting the evaluation result.

[0015] The invention of claim 9 is a program for predicting the occurrence of an earthquake and evaluating the risk of damage to a specified facility due to an earthquake, which is executed by a computer having a processor and a memory, characterized in that the processor is caused to execute the following steps: acquire earthquake occurrence record map data showing on a map that an earthquake occurred within a specified area during a first period, which is a specified period in the past; use the earthquake occurrence record map data as an input to make an inference about the occurrence of an earthquake within the specified area in the future based on a machine learning model using record data showing areas where earthquakes have occurred in the past in chronological order and ground data for each area as training data; output earthquake occurrence prediction map data showing on a map the predicted results of an earthquake occurring within the specified area during a second period, which is a specified period in the future, based on the result of the inference; and evaluate the risk of damage due to an earthquake occurring to facilities located in the specified area during the second period, based on the result of the inference, and output the evaluation result. Effect of the Invention

[0016] According to the inventions described in claims 1, 8, and 9, based on a machine learning model using historical data showing past earthquake occurrence areas in chronological order and ground data for each area as training data, earthquake occurrence record map data for a first period, which is a predetermined period in the past, is input, and earthquake occurrence prediction map data for a second period, which is a predetermined period in the future, is output. In addition, the damage risk of facilities in the second period due to earthquakes based on the results of the inference is evaluated.

[0017] That is, based on the machine learning model, earthquake occurrence record map data showing on a map the occurrence of an earthquake within a specified area is used to output earthquake occurrence forecast map data showing on a map the predicted results of an earthquake occurring within that area within a desired period. This makes it possible to predict the occurrence of an earthquake with a simple configuration and grasp the predicted results for a desired period. In addition, the damage risk of a facility due to an earthquake occurrence based on the result of the inference is evaluated. This makes it possible to grasp the damage risk of the facility, and by using the mid- to long-term prediction results, to grasp the safety measures that should be taken and the timing of those measures, thereby improving the safety of operations at the facility, streamlining the costs required for safety measures, and contributing to disaster prevention in the area.

[0018] According to the invention described in claim 2, a learning unit is provided that performs machine learning in response to the time series using the earthquake occurrence record map data acquired in time series as teacher data. This makes it possible to generate a machine learning model based on a larger amount of the latest earthquake occurrence data.

[0019] According to the invention described in claim 3, machine learning is performed using the earthquake occurrence forecast map data for a predetermined second period and the earthquake occurrence record map data for the second period as training data.

[0020] In other words, by performing machine learning on the prediction data for the second period based on the machine learning model and the actual data for that period, it is possible to obtain correct data from the discrepancy between the prediction and the actual data, and then update the machine learning model by performing further machine learning. This makes it possible to update the machine learning model based on more recent earthquake occurrence data.

[0021] According to the invention described in claim 4, the risk of damage to a facility due to an earthquake occurring during a second period is evaluated based on facility data of the facility located in a specified area and the result of inference by the machine learning model. This makes it possible to evaluate the risk of damage to a facility due to an earthquake occurring according to the situation of the facility.

[0022] According to the inventions described in claims 5 and 6, the earthquake occurrence record map data is image data showing the location where an earthquake has occurred and data related to the earthquake on a map showing a specified region. The earthquake occurrence record map data may also be image data showing the scale of the earthquake or the degree of shaking caused by the earthquake as a circle size centered on the location where the earthquake has occurred on a map showing a specified region. This makes it possible to predict the occurrence of earthquakes with a simple configuration.

[0023] According to the invention described in claim 7, the ground data includes data on one or more of the geology, topography, active faults, ocean trenches, and plate structures of each region. This makes it possible to make more precise predictions of earthquake occurrence. [Brief description of the drawings]

[0024] [Figure 1] 3 is an image diagram showing an example of earthquake occurrence record map data used in the disaster risk assessment device in accordance with the first embodiment of the present invention. FIG. [Diagram 2] 1 is a functional block diagram showing functions of a disaster risk assessment device 1 according to a first embodiment of the present invention. [Diagram 3] 3 is an image diagram showing an example of earthquake occurrence record map data acquired in time series by the learning unit 35 of FIG. 2. FIG. [Figure 4] 3 is a flowchart showing a procedure for a disaster risk assessment output process by the control unit 30 of FIG. 2. [Diagram 5] 3 is a flowchart showing a procedure of a learning process performed by a control unit 30 of FIG. 2. [Figure 6] FIG. 3 is a block diagram showing the procedure of machine learning and inference by the learning unit 35 of FIG. 2. [Figure 7] FIG. 7 is a functional block diagram showing a computer 700 according to a second embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0025] Hereinafter, the present invention will be described based on the illustrated embodiment.

[0026] (overview) A damage risk assessment device according to an embodiment of the present invention is a device for predicting the occurrence of an earthquake and assessing the damage risk of a specific facility due to the occurrence of an earthquake. As a specific embodiment, this damage risk assessment device holds a machine learning model in which past earthquake occurrence areas are shown in chronological order and ground data for each area are used as teacher data. Based on this machine learning model, the damage risk assessment device receives earthquake occurrence history map data showing on a map that an earthquake occurred in a certain area during a specified period in the past (first period), performs inference regarding the occurrence of an earthquake in that area during a specified period in the future (second period), and outputs the inference result as earthquake occurrence forecast map data. In addition, this damage risk assessment device evaluates the damage risk of a facility installed in the certain area due to the occurrence of an earthquake, which is the inference result, and outputs the evaluation result.

[0027] To infer whether an earthquake will occur, i.e., to predict earthquakes, this disaster risk assessment device uses earthquake occurrence record map data, which is image data showing locations where earthquakes have occurred in a specified period of time in the past and data related to the earthquakes on a map showing a certain region. Data related to earthquakes is, for example, the scale (magnitude) of the earthquake or the degree of shaking caused by the earthquake (seismic intensity), and the earthquake occurrence record map data shows this value as the size of a circle on the map. Note that earthquake occurrence forecast map data, which is the result of earthquake prediction, is also image data that shows the same.

[0028] Fig. 1 is an image diagram showing an example of earthquake occurrence record map data used in a disaster risk assessment device according to embodiment 1 of the present invention. The earthquake occurrence record map 211 shown in Fig. 1 is a map shown by the above earthquake occurrence record map data, and on a map M1 of a certain predetermined area, circles C1, C2, and C3 showing the scale (magnitude) of the earthquake or the degree of shaking caused by the earthquake (seismic intensity) by their size are displayed as data related to earthquakes that occurred in a predetermined period (first period) in the past.

[0029] The damage risk assessment device according to the first embodiment of the present invention performs earthquake prediction using earthquake occurrence record map data, which is image data obtained by digitizing the earthquake occurrence record map 211. The damage risk assessment device then outputs the results of the earthquake prediction as data on earthquakes predicted to occur in a predetermined future period (second period) using a map similar to the earthquake occurrence record map 211, for example, by indicating the scale (magnitude) of the earthquake or the degree of shaking caused by the earthquake (seismic intensity) by its size, as in circles C1, C2, and C3.

[0030] Furthermore, the damage risk assessment device according to the first embodiment of the present invention stores data on specific facilities installed on the map M1, specifically facilities such as buildings, dams, bridges, and steel towers. The damage risk assessment device then assesses the damage risk of the specific facilities in the event of an earthquake according to the above prediction, and outputs the assessment result. The damage risk includes, for example, the risk of direct damage to the facility due to an earthquake, and indirect risks that occur due to damage to surrounding facilities and infrastructure equipment, even if the facility is not damaged.

[0031] This configuration makes it possible to predict earthquakes without analyzing huge amounts of past earthquake data as big data. This makes it possible to predict earthquake occurrences and understand the results of predictions with a simple configuration. It also makes it possible to understand the damage risk of a facility, as well as the safety measures that should be taken and the timing of their implementation, thereby improving the safety of operations at the facility, streamlining the costs required for safety measures, and contributing to regional disaster prevention.

[0032] (Embodiment 1) <Configuration> 2 to 6 show this embodiment, and FIG. 2 is a functional block diagram showing functions of a damage risk assessment device 1 according to the first embodiment of the present invention. This damage risk assessment device 1 is a device that holds a machine learning model using performance data showing past earthquake occurrence areas in chronological order and ground data for each area as training data, performs earthquake prediction using earthquake occurrence performance map data shown in FIG. 1 as input, and outputs the prediction result as earthquake occurrence prediction map data similar to that in FIG. 1. The damage risk assessment device 1 may be configured, for example, by a computer (desktop, laptop, tablet, etc.) such as a stationary PC (Personal Computer) or a laptop PC, or may be configured by a server device such as a Web server (including a cloud server).

[0033] When configured as a server device, the disaster risk assessment device 1 may be configured to operate by communicating with a terminal device via a network. In this case, the terminal device is connected to the network by communicating with a communication device such as a wireless base station compatible with communication standards such as 4G, 5G, and LTE (Long Term Evolution), and a wireless LAN router compatible with wireless LAN (Local Area Network) standards such as IEEE (Institute of Electrical and Electronics Engineers) 802.11.

[0034] The disaster risk assessment device 1 includes, as its functional components, a communication unit 10, a storage unit 20, and a control unit 30. As shown in Fig. 2, the communication unit 10, the storage unit 20, and the control unit 30 are electrically connected to each other via a bus or the like.

[0035] The communication unit 10 is a communication interface for performing wired or wireless communication, and may use any communication protocol as long as mutual communication can be performed. The communication unit 10 performs communication using a communication protocol such as TCP / IP (Transmission Control Protocol / Internet Protocol).

[0036] The storage unit 20 stores programs and input data for executing various control processes and functions in the control unit 30, and is composed of memories including a RAM (Random Access Memory), a ROM (Read Only Memory), etc., and storage including a HDD (Hard Disk Drive), an SSD (Solid State Drive), a DRAM (Dynamic Random Access Memory), etc. The storage unit 20 also stores an earthquake occurrence record map database 21, a facility database 22, and a machine learning model 23.

[0037] The earthquake occurrence record map database 21 stores earthquake occurrence record map data, which is acquired by the acquisition unit 31 described later and shows on a map that earthquakes occurred within a predetermined area during a predetermined period in the past (first period). The earthquake occurrence record map data is image data showing locations where earthquakes occurred during a predetermined period in the past and data related to the earthquakes on a map showing a certain area, as in the example shown in FIG. 1, and the data related to the earthquakes is, for example, the scale (magnitude) of the earthquake or the degree of shaking caused by the earthquake (seismic intensity) shown by the size of a circle on the map. The predetermined area shown in the earthquake occurrence record map data is not limited to a specific area, and may be a prefecture, a city, a town, a village, a region (Tohoku region, Kanto region, etc.), the whole of Japan, or a foreign country, as in the map M1 shown in FIG. 1. Furthermore, the predetermined area shown in the earthquake occurrence record map data may be a part of a mesh obtained by dividing the earth into a predetermined area (for example, 1 square kilometer) unit.

[0038] The earthquake occurrence record map database 21 may also store ground data for each predetermined region. The ground data includes, for example, information on one or more of geology, topography, active faults, ocean trenches, and plate structures. The earthquake occurrence record map database 21 may also store ground data for each predetermined region, specifically, for each region such as the prefectures, cities, wards, towns, and villages. The earthquake occurrence record map database 21 may also store ground data for each area unit obtained by dividing the ground into a mesh-like shape in a predetermined area unit as described above. The earthquake occurrence record map database 21 may also store information on the start and end times of a predetermined period in the past, linked to the earthquake occurrence record map data.

[0039] The facility database 22 stores data on facilities installed in a given area, specifically facilities such as buildings, dams, bridges, and steel towers. The data on facilities is various data necessary for evaluating the risk of damage in the event of an earthquake of a certain scale, and the risk of damage includes, for example, the risk of direct damage to the facility due to an earthquake, and the indirect risk that occurs even if the facility is not damaged due to damage to surrounding facilities and infrastructure equipment. The facility database 22 stores information including, for example, the structure of the building, the date of construction, the earthquake resistance standard applied at the time of construction, information on the current damage and repair status of the facility, information on the infrastructure equipment used, and information on evacuation such as evacuation routes. This makes it possible to make more precise predictions of earthquake occurrence.

[0040] The machine learning model 23 is a machine learning model that uses earthquake occurrence record map data within a specified area as input and infers whether an earthquake will occur within the specified area in the future, and is an inference algorithm that is the result of training (learning) using record data showing past earthquake occurrence areas in chronological order and ground data for each area as teacher data. The machine learning model 23 is a machine learning model that has learned by recognizing patterns such as correlations between data on past earthquake occurrence areas and ground data for each area, and data on earthquakes that have occurred thereafter. The machine learning model 23 may be generated by machine learning performed as a function of the learning unit 35 described later, or may be a machine learning model generated as a result of machine learning performed by another device. Machine learning for generating a machine learning model will be described later.

[0041] The control unit 30 controls the overall operation of the disaster risk assessment device 1 by executing the programs stored in the storage unit 20, and is composed of devices including a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), a microprocessor, a processor core, a multiprocessor, an ASIC (Application-Specific Integrated Circuit), and an FPGA (Field Programmable Gate Array). The functions of the control unit 30 include an acquisition unit 31, an inference unit 32, an earthquake prediction output unit 33, a disaster risk assessment unit 34, and a learning unit 35. The acquisition unit 31, the inference unit 32, the earthquake prediction output unit 33, the disaster risk assessment unit 34, and the learning unit 35 are started by the programs stored in the storage unit 20 and executed by the disaster risk assessment device 1.

[0042] The acquisition unit 31 controls a process of acquiring earthquake occurrence record map data showing on a map that earthquakes occurred within a specified area during a first period, which is a specified period in the past. The earthquake occurrence record map data acquired by the acquisition unit 31 is image data showing locations where earthquakes occurred during a specified period in the past and data related to the earthquakes on a map showing a specified area such as the example shown in FIG. 1, as described above, and the data related to the earthquakes is, for example, the scale (magnitude) of the earthquake or the degree of shaking caused by the earthquake (seismic intensity) shown by the size of a circle on the map. This earthquake occurrence record map data may be used for machine learning to generate or update the machine learning model 23 in the learning unit 35 described later.

[0043] The acquisition unit 31 stores the acquired earthquake occurrence history map data in the earthquake occurrence history map database 21, for example.

[0044] The inference unit 32 controls a process of inferring that an earthquake will occur in a predetermined area in the future, using the earthquake occurrence record map data acquired by the acquisition unit 31 as an input, based on the machine learning model 23 that uses the achievement data showing the areas where earthquakes have occurred in the past in a time series and the ground data for each area as teacher data. Specifically, the inference unit 32 performs inference based on an inference algorithm of the machine learning model 23 that is a result of training (learning) using the achievement data showing the areas where earthquakes have occurred in the past in a time series and the ground data for each area as teacher data. The earthquake occurrence record map data input to the inference unit 32 is, for example, as shown in the example of FIG. 1, image data showing locations where earthquakes have occurred in a predetermined period of time in the past on a map showing a certain area, and data related to the earthquake, and the data related to the earthquake is, for example, the scale (magnitude) of the earthquake or the degree of shaking caused by the earthquake (seismic intensity) shown by the size of a circle on the map.

[0045] Specifically, the inference unit 32 infers that an earthquake will occur in a specified area in the future based on the machine learning model 23 as follows. The inference unit 32 recognizes patterns such as correlations between data on areas where earthquakes have occurred in the past and ground data for each area, and data on earthquakes that have occurred subsequently, stored in the machine learning model 23, from the earthquake occurrence record map data acquired by the acquisition unit 31. The inference unit 32 then determines whether an earthquake will occur in the area in the pattern, and if so, when and where it will occur. At this time, if it is determined that the pattern corresponds to a specific earthquake occurrence pattern, the inference unit 32 may weight the data on areas where earthquakes have occurred in the past or the ground data for each area.

[0046] The earthquake prediction output unit 33 controls a process of outputting earthquake occurrence prediction map data showing the prediction result of an earthquake occurring in a predetermined area in a second period, which is a predetermined period in the future, on a map based on the result of inference by the inference unit 32. Specifically, the earthquake prediction output unit 33 outputs data on whether an earthquake will occur in the area determined by the inference unit 32, and when and where it will occur if it will occur, as earthquake occurrence prediction map data similar to the earthquake occurrence record map data. The earthquake prediction output unit 33 may output data on when an earthquake will occur in the area in a time series as earthquake occurrence prediction map data. The earthquake occurrence prediction map data is, for example, image data showing locations where an earthquake is predicted to occur in a predetermined period in the future and data related to the predicted earthquake on a map showing a predetermined area such as the example shown in FIG. 1, and the data related to the earthquake is, for example, the scale (magnitude) of the earthquake or the degree of shaking caused by the earthquake (seismic intensity) shown by the size of a circle on the map.

[0047] The damage risk assessment unit 34 controls the process of evaluating the damage risk of facilities installed in a predetermined area in a second period, which is a predetermined period in the future, due to the occurrence of an earthquake based on the result of inference by the inference unit 32, and outputting the evaluation result. Specifically, the damage risk assessment unit 34 refers to the facility database 22 and reads data on predetermined facilities installed on the map M1 shown in FIG. 1, specifically facilities such as buildings, dams, bridges, and steel towers. Then, the damage risk assessment unit 34 evaluates the damage risk of the above facilities in the event of the occurrence of a predicted earthquake, which is the result of inference by the inference unit 32, and outputs the evaluation result. The damage risk includes, for example, the risk of direct damage to the facility due to an earthquake, and the indirect risk that occurs due to damage to surrounding facilities and infrastructure equipment even if the facility is not damaged.

[0048] The learning unit 35 performs machine learning on the earthquake occurrence record map data acquired in time series by the acquisition unit 31 as teacher data, corresponding to the time series, and controls the process of generating or updating a machine learning model based on the results of the machine learning. At this time, the learning unit 35 may update the existing machine learning model 23 by re-learning, or may generate a new machine learning model.

[0049] The learning unit 35 performs training (learning) using the past earthquake occurrence record map data and the ground data for each region as teacher data, and records patterns such as correlations between data on regions where earthquakes have occurred in the past and the ground data for each region, and data on earthquakes that have occurred thereafter.Then, the learning unit 35 generates, by machine learning, an inference algorithm that infers whether an earthquake will occur within a specified region in the future.

[0050] In addition, the learning unit 35 may perform machine learning using earthquake occurrence prediction map data for a second period, which is a predetermined period in the future, and earthquake occurrence record map data for the second period as training data, and update the machine learning model 23 based on the results of the machine learning.

[0051] FIG. 3 is an image diagram showing an example of earthquake occurrence record map data acquired in a time series by the learning unit 35 in FIG. 2. The earthquake occurrence record map 212 shown in FIG. 3 is a map shown by earthquake occurrence record map data in a first period, which is a predetermined period in the past, and circles C4, C5, and C6 showing the scale (magnitude) of an earthquake or the degree of shaking (seismic intensity) caused by an earthquake are displayed on a map M2 of a certain predetermined area as data related to an earthquake that occurred in a predetermined period in the past (first period). The earthquake occurrence prediction map 213 shown in FIG. 3 is a map shown by earthquake occurrence prediction map data predicted to occur in a predetermined period (second period) in the future, and circles C7 and C8 showing the scale (magnitude) of an earthquake or the degree of shaking (seismic intensity) caused by an earthquake are displayed on a map M3 of an area similar to the map M2 of the predetermined area as data related to an earthquake that is predicted to actually occur in the predetermined period (second period). The earthquake occurrence prediction map 213 shown in FIG. 3 may have multiple earthquake occurrence prediction map data in a time series.

[0052] The learning unit 35 performs machine learning using the earthquake occurrence prediction map data, which is prediction data for the second time period, and the earthquake occurrence record map data, which is actual data for that time period, as teacher data. That is, since the earthquake occurrence record map 213, which is actual data, is the correct answer data for the earthquake occurrence record map 212, which is prediction data, further machine learning is performed based on the deviation between the prediction data and the actual data, thereby updating the machine learning model 23. This makes it possible to predict earthquake occurrence with a simple configuration, and to update the machine learning model based on more and more recent earthquake occurrence data.

[0053] The machine learning in the learning unit 35 may be performed by supervised machine learning using the past earthquake occurrence record map data and the ground data for each region as teacher data, or may be performed by deep learning. The learning unit 35 may also perform data processing (annotation) on the teacher data, that is, the past earthquake occurrence record map data and the ground data for each region, by adding, for example, an interpretation indicated by each data as tag information. Furthermore, when updating the machine learning model 23, the learning unit 35 may update by re-learning, or may be configured to update only when the performance of the machine learning model 23 has been improved by performing a performance evaluation of the machine learning model 23.

[0054] 3, the configuration in which the size of the circles C4 to C8 indicates the scale (magnitude) of an earthquake or the degree of shaking caused by an earthquake (seismic intensity) is merely an example, and the configuration in which the scale (magnitude) of an earthquake or the degree of shaking caused by an earthquake (seismic intensity) is not limited to this. For example, the area shown on the map M2 may be divided into mesh-like sections at predetermined intervals, and each divided area may be shown in a different color (including different shades of color).

[0055] <Processing flow> An example of the flow of control processing (disaster risk assessment method) by the control unit 30 of the disaster risk assessment device 1 will be described with reference to Fig. 4 to Fig. 6. Fig. 4 is a flowchart showing the procedure of disaster risk assessment output processing by the control unit 30 of Fig. 2.

[0056] As the process of step S101, the acquisition unit 31 of the control unit 30 acquires earthquake occurrence record map data showing on a map that earthquakes occurred within a specified area during a first period, which is a specified period in the past. The earthquake occurrence record map data acquired in step S101 is image data showing locations where earthquakes occurred during a specified period in the past and data related to the earthquakes on a map showing a specified area such as the example shown in Fig. 1, and the data related to the earthquakes is, for example, the scale (magnitude) of the earthquake or the degree of shaking caused by the earthquake (seismic intensity) shown by the size of a circle on the map.

[0057] In the process of step S102, the inference unit 32 of the control unit 30 uses the earthquake occurrence record map data acquired in step S101 as input and infers whether an earthquake will occur within a predetermined area in the future, based on the machine learning model 23 that uses the achievement data showing the areas where past earthquakes have occurred in time series and the ground data for each area as teacher data. Specifically, in step S102, inference is made based on the inference algorithm of the machine learning model 23, which is the result of training (learning) using the achievement data showing the areas where past earthquakes have occurred in time series and the ground data for each area as teacher data.

[0058] In the process of step S103, the earthquake prediction output unit 33 of the control unit 30 outputs earthquake occurrence prediction map data showing on a map the predicted result of an earthquake occurring in a predetermined area in a second period, which is a predetermined period in the future, based on the result of the inference in step S102. In step S103, for example, on a map showing a predetermined area such as the example shown in Fig. 1, the earthquake occurrence prediction map data is output, which is image data showing locations where earthquakes are predicted to occur in a predetermined period in the future and data related to the predicted earthquake, and in which the data related to the earthquake is, for example, the scale (magnitude) of the earthquake or the degree of shaking caused by the earthquake (seismic intensity) shown by the size of a circle on the map.

[0059] In the process of step S104, the damage risk assessment unit 34 of the control unit 30 assesses the damage risk of facilities installed in a predetermined area in a second period, which is a predetermined period in the future, due to the occurrence of an earthquake based on the result of the inference in step S102, and outputs the assessment result. In step S104, specifically, the facility database 22 is referenced to read data on predetermined facilities installed on the map M1 shown in FIG. 1, specifically facilities such as buildings, dams, bridges, and steel towers, and the damage risk of the above facilities in the event of the predicted earthquake resulting from the inference in step S102 occurs, and the assessment result is output. This makes it possible to grasp the damage risk of the facility, as well as to grasp the safety measures that should be taken and the timing of those measures, thereby improving the safety of operations at the facility, streamlining the costs required for safety measures, and contributing to disaster prevention in the region.

[0060] Fig. 5 is a flowchart showing the procedure of the learning process by the control unit 30 in Fig. 2. The learning process shown in Fig. 5 may be performed automatically at a predetermined period (quarterly, once a year, etc.) or may be performed whenever required by a user's instruction operation.

[0061] As part of the processing of step S201, the acquisition unit 31 of the control unit 30 acquires earthquake occurrence history map data showing on a map that earthquakes occurred within a specified area during a first period, which is a specified period in the past, in a manner similar to the processing of step S101.

[0062] As part of the processing of step S202, the control unit 30 acquires earthquake occurrence prediction map data that shows on a map the predicted results of an earthquake occurring within a specified area during a second period, which is a specified period in the future, output in step S103.

[0063] In the process of step S203, the learning unit 35 of the control unit 30 performs machine learning on the time series of the earthquake occurrence record map data acquired in step S201 as teacher data, and generates or updates a machine learning model based on the results of the machine learning. Specifically, in step S203, training (learning) is performed using the past earthquake occurrence record map data and ground data for each region as teacher data, and patterns such as correlations between data on past earthquake occurrence regions and ground data for each region and data on earthquakes that have occurred thereafter are recorded. Then, the learning unit 35 generates an inference algorithm by machine learning that infers the occurrence of an earthquake within a specified region in the future.

[0064] In the process of step S204, the learning unit 35 of the control unit 30 updates the machine learning model 23 based on the result of the machine learning performed in step S203. Note that, when generating a new machine learning model different from the machine learning model 23, the process of step S204 does not need to be performed. As a result, the machine learning model 23 is updated to the latest learning state.

[0065] The procedure of machine learning and inference by the learning unit 35 of the disaster risk assessment device 1 will be described with reference to Fig. 6. Fig. 6 is a block diagram showing the procedure of machine learning and inference by the learning unit 35 of Fig. 2.

[0066] In the process of step S301, the acquisition unit 31 of the control unit 30 collects teacher data for machine learning. Examples of teacher data include earthquake occurrence record map data showing past earthquake occurrence areas in chronological order, earthquake occurrence prediction map data predicted in the past, and ground data for each area. The earthquake occurrence record map data and earthquake occurrence prediction map data are image data showing locations on a map where an earthquake has occurred or where an earthquake is predicted to occur, and data related to the earthquake. The ground data includes information on one or more of geology, topography, active faults, ocean trenches, and plate structures, for example.

[0067] In the process of step S302, the learning unit 35 of the control unit 30 performs supervised machine learning using the training data collected in step S301, and generates or updates the machine learning model 23.

[0068] As processing in step S303, the acquisition unit 31 of the control unit 30 acquires earthquake occurrence history map data, which is input data for inferring whether an earthquake will occur in a specified area in the future, based on the machine learning model 23.

[0069] In the process of step S304, the inference unit 32 of the control unit 30 performs inference regarding the occurrence of an earthquake in a predetermined area in the future, based on the input data acquired in step S303.

[0070] In the process of step S305, the earthquake prediction output unit 33 of the control unit 30 outputs the inference result of step S304.

[0071] <Effects> According to the disaster risk assessment device 1 and disaster risk assessment method of the first embodiment, based on the machine learning model 23 using the past record data showing the past earthquake occurrence areas in a time series and the ground data for each area as teacher data, an inference is made about the occurrence of an earthquake in a predetermined area in the future from the earthquake occurrence record map data showing on a map the occurrence of an earthquake in a predetermined area in a first period, which is a predetermined period in the past. Then, based on the result of the inference, the earthquake occurrence prediction map data showing on a map the prediction result of the occurrence of an earthquake in the area in the second period, which is a predetermined period in the future, is output. This makes it possible to predict the occurrence of an earthquake and grasp the prediction result with a simple configuration.

[0072] Moreover, according to the damage risk assessment device 1 and the damage risk assessment method, the damage risk of a facility installed in a specified area in a second period, which is a specified period in the future due to an earthquake based on the result of the inference, is assessed and the assessment result is output. For this purpose, a facility database 22 is provided for storing data on the facility. This makes it possible to assess the damage risk of a facility due to an earthquake according to the facility's situation, and therefore it is possible to grasp the damage risk of the facility as well as the safety measures that should be taken and the timing of those measures, thereby contributing to improving the safety of operations at the facility, streamlining the costs required for safety measures, and contributing to disaster prevention in the region.

[0073] Furthermore, the damage risk assessment device 1 and the damage risk assessment method include a learning unit 35 that updates the machine learning model 23. This makes it possible to generate a machine learning model based on a larger amount of the latest earthquake occurrence data.

[0074] Furthermore, according to the disaster risk assessment device 1 and the disaster risk assessment method, the learning unit 35 performs machine learning using the earthquake occurrence prediction map data for a second period, which is a predetermined period in the future, and the earthquake occurrence record map data for the second period as teacher data, and updates the machine learning model 23 based on the machine learning results. Therefore, by performing machine learning from the prediction data for the second period based on the machine learning model and the record data for that period, correct answer data can be obtained from the discrepancy between the prediction and the record, and the machine learning model is updated by performing further machine learning. This makes it possible to update the machine learning model based on more and more recent earthquake occurrence data.

[0075] (Embodiment 2 (program)) 7 is a functional block diagram showing an example of the configuration of a computer (electronic calculator) 700 according to the second embodiment of the present invention. The computer 700 includes a CPU 701, a main memory device 702, an auxiliary memory device 703, and an interface 704.

[0076] Here, the details of the programs for realizing the functions constituting the acquisition unit 31, the inference unit 32, the earthquake prediction output unit 33, the damage risk assessment unit 34, and the learning unit 35 according to the first embodiment will be described. These functional blocks are implemented in a computer 700. The operations of these components are stored in the auxiliary storage device 703 in the form of a program. The CPU 701 reads the program from the auxiliary storage device 703 and loads it in the main storage device 702, and executes the above-mentioned processing according to the program. The CPU 701 also secures a storage area corresponding to the above-mentioned storage unit in the main storage device 702 according to the program.

[0077] Specifically, the program is a program that, on a computer 700, realizes the following steps: acquiring earthquake occurrence history map data showing on a map that earthquakes occurred within a specified area during a first period, which is a specified time period in the past; using the earthquake occurrence history map data as input and making inferences about the occurrence of an earthquake within a specified area in the future based on a machine learning model that uses history data showing areas where earthquakes have occurred in the past in chronological order and ground data for each area as training data; outputting earthquake occurrence prediction map data showing on a map the predicted results of an earthquake occurring within a specified area during a second period, which is a specified time period in the future, based on the results of the inference; and evaluating the risk of damage due to an earthquake occurring to facilities located in the specified area during the second period, based on the results of the inference, and outputting the evaluation results.

[0078] The CPU 701 is hardware for executing an instruction set written in a program, and is composed of an arithmetic unit, a register, peripheral circuits, etc. The CPU 701 is composed of at least one processor, and is typically a microprocessor, but may be other types of processors including an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), a microprocessor, a processor core, and a multiprocessor. The at least one processor may be a single core or a multicore. The at least one processor may also be a processor in the broad sense, such as a hardware circuit (for example, an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit)) that performs part or all of the processing.

[0079] The main memory device 702 is for temporarily storing programs and data processed by the programs, and is, for example, a memory such as a dynamic random access memory (DRAM). The main memory device 702 is one or more memories, and may typically be configured by a main storage device, and at least one of the memories may be a volatile memory or a non-volatile memory.

[0080] The auxiliary storage device 703 is an example of a non-transitory tangible medium. Other examples of non-transitory tangible media include a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, and a semiconductor memory connected via the interface 704. In addition, when this program is distributed to the computer 700 via a network, the computer 700 that receives the program may load the program into the main storage device 702 and execute the above-mentioned process.

[0081] The program may be for realizing part of the above-mentioned functions. Furthermore, the program may be a so-called difference file (difference program) that realizes the above-mentioned functions in combination with another program already stored in the auxiliary storage device 703.

[0082] Although the embodiment of the present invention has been described in detail above, the specific configuration is not limited to this embodiment, and even if there are design changes and the like within the scope of the gist of the present invention, they are included in the present invention. [Explanation of symbols]

[0083] 1: Disaster risk assessment device 10: Communications Department 20: Storage section 21: Earthquake occurrence record map database 22: Facility Database 23: Machine learning models 30: Control section 31: Acquisition part 32: Reasoning part 33: Earthquake prediction output section 34: Disaster Risk Assessment Department 35: Learning Department 700: Computer 701: CPU 702: Main memory 703 :Auxiliary storage device 704: Interface

Claims

1. A disaster risk assessment device that predicts the occurrence of an earthquake and assesses the risk of damage to a specified facility due to the occurrence of an earthquake, an acquisition unit that acquires earthquake occurrence record map data that shows on a map that earthquakes have occurred within a specified area during a first period that is a specified period in the past; an inference unit that uses the earthquake occurrence record map data as input and infers the occurrence of an earthquake in a specified area in the future based on a machine learning model that uses the historical data showing the areas where earthquakes have occurred in the past in chronological order and the ground data for each area as training data; an earthquake prediction output unit that outputs earthquake occurrence prediction map data showing a prediction result of an earthquake occurring in a specified area in a second period, which is a specified period in the future, on a map based on the result of the inference; and a damage risk assessment unit that assesses a damage risk of facilities installed in the specified area during the second period due to an earthquake occurrence based on a result of the inference and outputs an assessment result. A disaster risk assessment device comprising:

2. a learning unit that performs machine learning on the earthquake occurrence record map data acquired in time series as teacher data, and generates the machine learning model based on the results of the machine learning; The disaster risk assessment device according to claim 1 .

3. the learning unit performs machine learning using the earthquake occurrence forecast map data for the second time period and the earthquake occurrence record map data for the second time period as teacher data, and updates the machine learning model based on a result of the machine learning.

3. The disaster risk assessment device according to claim 2.

4. The damage risk assessment unit assesses a damage risk of facilities installed in the specified area during the second period based on facility data stored in advance as data related to facilities installed in the specified area and the result of the inference. The disaster risk assessment device according to claim 1 .

5. The earthquake occurrence record map data is image data showing locations where earthquakes occurred during the first period and data related to the earthquakes on a map showing the specified region.

5. A disaster risk assessment device according to claim 1, wherein:

6. The earthquake occurrence record map data is image data on a map showing the specified region, with the size of a circle representing the scale of the earthquake or the degree of shaking caused by the earthquake at the center of the location where the earthquake occurred during the first period, or image data on which the specified region is divided into a mesh at predetermined intervals and the locations where earthquakes occurred during the first period are represented in different colors. The disaster risk assessment device according to claim 5 .

7. The ground data includes data on any one or more of the geology, topography, active faults, ocean trenches, and plate structures of each region; 5. A disaster risk assessment device according to claim 1, wherein:

8. A damage risk assessment method for predicting an earthquake occurrence and assessing a damage risk of a predetermined facility due to the occurrence of an earthquake, the method being executed by a computer having a processor and a memory, the method comprising: The processor, A step of acquiring earthquake occurrence record map data showing on a map that earthquakes have occurred within a predetermined area during a first period, which is a predetermined period in the past; A step of inferring future earthquakes occurring within a specified area using the earthquake occurrence map data as input, based on a machine learning model that uses historical data showing areas where earthquakes have occurred in the past in chronological order and ground data for each area as training data; outputting earthquake occurrence prediction map data showing on a map the predicted results of an earthquake occurring within a predetermined area during a second period, which is a predetermined period in the future, based on the result of the inference; and evaluating a damage risk due to an earthquake occurring at a facility located in the specified area during the second time period based on a result of the inference, and outputting an evaluation result. A disaster risk assessment method comprising:

9. A program for predicting the occurrence of an earthquake and assessing a risk of damage to a predetermined facility due to the occurrence of an earthquake, the program being executed by a computer having a processor and a memory, The processor, A step of acquiring earthquake occurrence record map data showing on a map that earthquakes have occurred within a predetermined area during a first period, which is a predetermined period in the past; A step of inferring future earthquakes occurring within a specified area using the earthquake occurrence map data as input, based on a machine learning model that uses historical data showing areas where earthquakes have occurred in the past in chronological order and ground data for each area as training data; outputting earthquake occurrence prediction map data showing on a map the predicted results of an earthquake occurring within a predetermined area during a second period, which is a predetermined period in the future, based on the result of the inference; and evaluating a damage risk due to an earthquake occurring at a facility located in the specified area during the second time period based on a result of the inference, and outputting an evaluation result. A program characterized by:

Images