Model construction device

Abstract

A model construction device comprises a control unit that performs machine learning using disaster damage information (50) indicating whether or not first equipment attached to a first bridge was damaged at the time of a flood disaster, first equipment information (51) indicating characteristics of the first equipment, first water amount information (52) indicating the water amount observed at an observation station corresponding to the first bridge at the time of the flood disaster, and first environmental information (53) indicating the environment in which the first bridge is installed, and that thus constructs a prediction model (11) which, upon receipt of input of second equipment information indicating characteristics of second equipment attached to a second bridge, second water amount information indicating the water amount observed at an observation station corresponding to the second bridge, and second environmental information indicating the environment in which the second bridge is installed, predicts the probability that the second equipment will be damaged.

Description

Model Building Equipment

[0001] The present disclosure relates to a model construction device.

[0002] Research is known that attempts to analyze damage to bridges affected by floods and quantitatively evaluate their vulnerability. Non-Patent Document 1 proposes a method for estimating the current risk of damage and future risk of scouring for railway bridges through multivariate analysis based on various parameters obtained from routine inspection work and past damage data, and extracting bridge piers that are at risk of damage when the river floods. Non-Patent Document 2 focuses on the Misasa River in Hiroshima Prefecture, Japan, where many bridges were damaged by heavy rain in July 2018, and proposes a method for investigating bridge damage along the river and classifying bridge damage using specific indicators based on the results of statistical analysis.

[0003] Patent Documents 1 and 2 disclose techniques for predicting damage to facilities due to natural disasters.

[0004] International Publication No. 2021 / 240650 Japanese Patent Application Laid-Open No. 2023-160598

[0005] Masahiko Samizo and three others, "Statistical Method for Identifying Bridge Piers at Risk of Damage During Flooding of Railway Bridges," Journal of the Japan Society of Civil Engineers, D3 (Civil Engineering Planning), Vol. 69, No. 3, pp. 237-249, 2013. Takashi Inoue and two others, "Study on Bridge Damage Risk Assessment Indicators During Heavy Rains - Using Longitudinal Distribution Data of Damage on the Misasa River Bridge in the 2018 Western Japan Heavy Rains," Journal of the Japan Society of Civil Engineers, B1 (Hydraulic Engineering), Vol. 75, No. 2, I_1417-I_1422, 2019.

[0006] The methods proposed in Non-Patent Documents 1 and 2 are intended to predict bridge damage caused by floods, and cannot be applied to predicting damage to pipelines attached to bridges. The technologies disclosed in Patent Documents 1 and 2 are also intended to predict damage to facilities caused by natural disasters other than floods, such as earthquakes, and are difficult to apply to predicting damage to facilities caused by floods.

[0007] The purpose of this disclosure, made in consideration of the above circumstances, is to predict flood damage to equipment attached to bridges.

[0008] A model building device according to one embodiment acquires damage information indicating whether a first facility attached to a first bridge was damaged during a flood, first facility information indicating the characteristics of the first facility, first water volume information indicating the water volume observed at an observation station corresponding to the first bridge during the flood, and first environmental information indicating the environment in which the first bridge is installed; and comprises a control unit that performs machine learning using the acquired damage information, first facility information, first water volume information, and first environmental information to construct a predictive model that predicts the possibility of the second facility being damaged when second facility information indicating the characteristics of a second facility attached to a second bridge, second water volume information indicating the water volume observed at an observation station corresponding to the second bridge, and second environmental information indicating the environment in which the second bridge is installed is input.

[0009] According to the present disclosure, it is possible to predict damage caused by flooding to equipment attached to a bridge.

[0010] 4 is a block diagram showing the configuration of a system according to one embodiment. FIG. 5 is a schematic diagram showing an example of equipment attached to a bridge. FIG. 6 is a diagram showing the flow of constructing a prediction model. FIG. 7 is a table showing example values ​​of damage presence / absence, equipment characteristics, excess water level, and accumulated rainfall. FIG. 8 is a table showing an example of categorization and standardization of the values ​​shown in FIG. 4. FIG. 9 is a diagram showing an example of constructing a prediction model using a neural network. FIG. 10 is a diagram showing the flow of predicting damage probability and damage presence / absence. FIG. 11 is a diagram showing an example of damage prediction results.

[0011] An embodiment will be described below with reference to the drawings.

[0012] In each drawing, the same or corresponding parts are denoted by the same reference numerals. In the description of this embodiment, the description of the same or corresponding parts will be omitted or simplified as appropriate.

[0013] The configuration of a system 10 according to this embodiment will be described with reference to FIG.

[0014] The system 10 according to this embodiment includes a model construction device 20 and a damage prediction device 30 .

[0015] The model construction device 20 is a device that constructs the prediction model 11. The model construction device 20 may be a general-purpose computer such as a PC, a server computer such as a cloud server, or a dedicated computer. "PC" is an abbreviation for personal computer.

[0016] The model construction device 20 may be able to communicate with the disaster prediction device 30 via a network. The network may include, for example, the Internet, at least one WAN, at least one MAN, or any combination thereof. "WAN" is an abbreviation for wide area network. "MAN" is an abbreviation for metropolitan area network. The network may include at least one wireless network, at least one optical network, or any combination thereof. The wireless network may be, for example, an ad hoc network, a cellular network, a wireless LAN, a satellite communication network, or a terrestrial microwave network. "LAN" is an abbreviation for local area network.

[0017] The damage prediction device 30 is a device that predicts flood damage to facilities attached to a bridge 40, such as a pipeline 44 as shown in Fig. 2, using the prediction model 11 constructed by the model construction device 20. The damage prediction device 30 may be a general-purpose computer such as a PC, a server computer such as a cloud server, or a dedicated computer.

[0018] The bridge 40 is, for example, a road bridge spanning a river. As shown in Fig. 2, the bridge 40 includes a deck 41, a bridge girder 42, and a support member 43 formed of steel or the like. The pipeline 44 is attached to the bridge girder 42 below the deck 41 via the support member 43. The pipeline 44 is, for example, a structure for transporting water or gas, or a structure for laying communication cables or electric wires.

[0019] Predicting damage to a bridge 40 requires detailed information about the bridge 40 and the river, but it is difficult for the operator managing the pipeline 44 to collect similar information for numerous facilities scattered across a wide area. Furthermore, the pipeline 44 differs in material, structure, and dimensions from the bridge 40 itself, meaning that the pipeline 44 may be damaged even if the bridge 40 itself is not. Therefore, in this embodiment, the model construction device 20 uses information about the facilities owned by the operator managing the pipeline 44 and information such as river water levels, rainfall, and topography published by government ministries and agencies or local governments to build a model for predicting flood damage to the pipeline 44. The damage prediction device 30 uses this model to predict and output the probability of flood damage or the presence or absence of flood damage to the pipeline 44. Therefore, this embodiment makes it possible to predict flood damage to the pipeline 44.

[0020] The configuration of a model building device 20 according to this embodiment will be described with reference to FIG.

[0021] The model building device 20 includes a control unit 21 , a storage unit 22 , a communication unit 23 , and an input unit 24 .

[0022] The control unit 21 includes at least one processor, at least one programmable circuit, at least one dedicated circuit, or any combination thereof. The processor is a general-purpose processor such as a CPU or GPU, or a dedicated processor specialized for specific processing. "CPU" is an abbreviation for central processing unit. "GPU" is an abbreviation for graphics processing unit. An example of the programmable circuit is an FPGA. "FPGA" is an abbreviation for field-programmable gate array. An example of the dedicated circuit is an ASIC. "ASIC" is an abbreviation for application specific integrated circuit. The control unit 21 controls each part of the model construction device 20 and executes processing related to the operation of the model construction device 20.

[0023] The storage unit 22 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or any combination thereof. The semiconductor memory is, for example, a RAM, a ROM, or a flash memory. "RAM" is an abbreviation for random access memory. "ROM" is an abbreviation for read only memory. RAM is, for example, an SRAM or a DRAM. "SRAM" is an abbreviation for static random access memory. "DRAM" is an abbreviation for dynamic random access memory. ROM is, for example, an EEPROM. "EEPROM" is an abbreviation for electrically erasable programmable read only memory. Flash memory is, for example, an SSD. "SSD" is an abbreviation for solid-state drive. Magnetic memory is, for example, an HDD. "HDD" is an abbreviation for hard disk drive. The storage unit 22 functions, for example, as a main storage device, an auxiliary storage device, or a cache memory. The storage unit 22 stores information used in the operation of the model construction device 20 and information obtained by the operation of the model construction device 20 .

[0024] The communication unit 23 includes at least one communication module. The communication module is, for example, a module compatible with a wired LAN communication standard such as Ethernet (registered trademark), a wireless LAN communication standard such as IEEE 802.11, or a mobile communication standard such as LTE, 4G, or 5G. "IEEE" is an abbreviation for Institute of Electrical and Electronics Engineers. "LTE" is an abbreviation for Long Term Evolution. "4G" is an abbreviation for 4th generation. "5G" is an abbreviation for 5th generation. The communication unit 23 may communicate with the damage prediction device 30. The communication unit 23 receives information used in the operation of the model construction device 20 and transmits information obtained by the operation of the model construction device 20. As shown in FIG. 1, the communication unit 23 may transmit the prediction model 11 to the damage prediction device 30.

[0025] The input unit 24 is, for example, a physical key, a capacitance key, a pointing device, a touch screen integrated with a display, a camera, or a microphone. The input unit 24 accepts operations to input information used in the operation of the model construction device 20. The input unit 24 may be connected to the model construction device 20 as an external input device, instead of being provided in the model construction device 20. As a connection interface, an interface compatible with standards such as USB, HDMI (registered trademark), or Bluetooth (registered trademark) can be used. "USB" is an abbreviation for Universal Serial Bus. "HDMI (registered trademark)" is an abbreviation for High-Definition Multimedia Interface.

[0026] The functions of the model construction device 20 are realized by executing a model construction program according to this embodiment on a processor serving as the control unit 21. That is, the functions of the model construction device 20 are realized by software. The model construction program causes a computer to execute the operations of the model construction device 20, thereby causing the computer to function as the model construction device 20. That is, the computer functions as the model construction device 20 by executing the operations of the model construction device 20 in accordance with the model construction program.

[0027] The program can be stored on a non-transitory computer-readable medium. Examples of the non-transitory computer-readable medium include flash memory, magnetic recording devices, optical disks, magneto-optical recording media, and ROMs. The program can be distributed by selling, transferring, or lending portable media such as SD cards, DVDs, or CD-ROMs that store the program. "SD" is an abbreviation for Secure Digital. "DVD" is an abbreviation for digital versatile disc. "CD-ROM" is an abbreviation for compact disc read only memory. The program can also be distributed by storing it in the storage of a server and transferring it from the server to another computer. The program can also be provided as a program product.

[0028] A computer temporarily stores a program stored on a portable medium or transferred from a server in its main storage device. The computer then reads the program stored in the main storage device with its processor and executes processing in accordance with the read program. The computer may also read the program directly from the portable medium and execute processing in accordance with the program. The computer may also execute processing in accordance with the received program each time a program is transferred from the server to the computer. Processing may also be executed using a so-called ASP-type service that realizes functions simply by issuing execution instructions and obtaining results, without transferring the program from the server to the computer. "ASP" is an abbreviation for application service provider. A program is information used for processing by a computer and includes something equivalent to a program. For example, data that is not a direct instruction to a computer but has properties that define computer processing falls under the category of "something equivalent to a program."

[0029] Some or all of the functions of the model construction device 20 may be realized by a programmable circuit or a dedicated circuit as the control unit 21. In other words, some or all of the functions of the model construction device 20 may be realized by hardware.

[0030] The configuration of a disaster prediction device 30 according to this embodiment will be described with reference to FIG.

[0031] The damage prediction device 30 includes a control unit 31 , a memory unit 32 , a communication unit 33 , an input unit 34 , and a display unit 35 .

[0032] The control unit 31 includes at least one processor, at least one programmable circuit, at least one dedicated circuit, or any combination thereof. The processor is a general-purpose processor such as a CPU or GPU, or a dedicated processor specialized for specific processing. The programmable circuit is, for example, an FPGA. The dedicated circuit is, for example, an ASIC. The control unit 31 controls each part of the damage prediction device 30 and executes processing related to the operation of the damage prediction device 30.

[0033] The memory unit 32 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or any combination thereof. The semiconductor memory is, for example, a RAM, a ROM, or a flash memory. The RAM is, for example, an SRAM or a DRAM. The ROM is, for example, an EEPROM. The flash memory is, for example, an SSD. The magnetic memory is, for example, an HDD. The memory unit 32 functions as, for example, a main memory device, an auxiliary memory device, or a cache memory. The memory unit 32 stores information used in the operation of the damage prediction device 30 and information obtained by the operation of the damage prediction device 30.

[0034] The communication unit 33 includes at least one communication module. The communication module is, for example, a module compatible with a wired LAN communication standard such as Ethernet (registered trademark), a wireless LAN communication standard such as IEEE 802.11, or a mobile communication standard such as LTE, 4G, or 5G. The communication unit 33 may communicate with the model construction device 20. The communication unit 33 receives information used in the operation of the damage prediction device 30 and transmits information obtained by the operation of the damage prediction device 30. As shown in FIG. 1 , the communication unit 33 may receive the prediction model 11 from the model construction device 20.

[0035] The input unit 34 is, for example, a physical key, a capacitance key, a pointing device, a touch screen integrated with a display, a camera, or a microphone. The input unit 34 accepts an operation to input information used in the operation of the damage prediction device 30. The input unit 34 may be connected to the damage prediction device 30 as an external input device instead of being provided in the damage prediction device 30. The connection interface may be an interface compatible with standards such as USB, HDMI (registered trademark), or Bluetooth (registered trademark).

[0036] The display unit 35 is, for example, an LCD or organic EL display. "LCD" is an abbreviation for liquid crystal display. "EL" is an abbreviation for electroluminescent. The display unit 35 displays information obtained by the operation of the damage prediction device 30. The display unit 35 may be connected to the damage prediction device 30 as an external display device instead of being provided in the damage prediction device 30. The connection interface may be an interface compatible with standards such as USB, HDMI (registered trademark), or Bluetooth (registered trademark).

[0037] The functions of the damage prediction device 30 are realized by executing a damage prediction program according to this embodiment on a processor serving as the control unit 31. That is, the functions of the damage prediction device 30 are realized by software. The damage prediction program causes a computer to execute the operations of the damage prediction device 30, thereby causing the computer to function as the damage prediction device 30. That is, the computer functions as the damage prediction device 30 by executing the operations of the damage prediction device 30 in accordance with the damage prediction program.

[0038] Some or all of the functions of the damage prediction device 30 may be realized by a programmable circuit or a dedicated circuit as the control unit 31. In other words, some or all of the functions of the damage prediction device 30 may be realized by hardware.

[0039] The operation of the model building device 20 according to this embodiment will be described with reference to Fig. 3. The operation described below corresponds to the model building method according to this embodiment.

[0040] The control unit 21 of the model construction device 20 acquires damage information 50, first equipment information 51, first water volume information 52, and first environment information 53. The damage information 50 is information indicating whether the first equipment attached to the first bridge was damaged when the flood occurred. The first equipment information 51 is information indicating the characteristics of the first equipment. The first water volume information 52 is information indicating the water volume observed at the observation station corresponding to the first bridge when the flood occurred. The first environment information 53 is information indicating the environment in which the first bridge is installed.

[0041] Specifically, if some or all of the information among the disaster information 50, the first equipment information 51, the first water volume information 52, and the first environmental information 53 is pre-stored in the storage unit 22, the control unit 21 reads the information from the storage unit 22. Alternatively, the control unit 21 may receive some or all of the information among the disaster information 50, the first equipment information 51, the first water volume information 52, and the first environmental information 53 from another device via the communication unit 23. Alternatively, the control unit 21 may accept input of some or all of the information among the disaster information 50, the first equipment information 51, the first water volume information 52, and the first environmental information 53 from the user via the input unit 24.

[0042] In this embodiment, the first facility is a pipeline 44 as shown in FIG.

[0043] The damage information 50 classifies the state of the pipeline 44 into two categories, "damaged" and "not damaged," based on the results of surveys of the pipeline 44 in past floods.

[0044] The definition of "damaged" is determined by the person conducting the damage prediction based on the required performance of the pipeline 44. If damage occurs in the pipeline 44 that does not penetrate the interior, a response is required for pipes transporting flammable materials or pipes subject to internal pressure, but a pipe containing cables may require observation. In this embodiment, damage to the first facility refers to a state in which the first facility cannot be continued to be used unless repair or reinforcement measures are taken due to changes in the river's water level. Examples of forms of flood damage to the pipeline 44 include scratches, flattening, bending, or joint separation on the pipeline surface due to the impact of water currents or drifting objects. The pipeline 44 will also be damaged if the bridge girders 42 of the bridge 40 are washed away or if the bridge piers tilt or collapse. However, because it is difficult to predict damage to the bridge 40 without detailed information about the bridge 40 and the river, damage to the pipeline 44 when the bridge 40 itself is damaged is not subject to prediction.

[0045] In this embodiment, the damage information 50 includes the presence or absence of damage, as well as the equipment name and number for uniquely identifying the pipeline 44 .

[0046] The first equipment information 51 includes the pipe type, the installation position, or the section length.

[0047] The pipe type is the type of the pipeline 44. The pipe type indirectly indicates, for example, the material of the pipeline 44 and the type or strength of the joint. When there are multiple combinations of outer diameter and inner diameter of the pipeline 44, a pipe type is determined for each combination of material and outer diameter and inner diameter.

[0048] The attached position is the position where the conduit 44 is installed relative to the bridge girder 42. The attached position is, for example, the upstream or downstream side of the river on which the bridge 40 is built, or the back surface between two main girders. In particular, if the conduit 44 is attached to the side facing upstream, the conduit 44 is susceptible to damage due to the action of water currents or drifting debris.

[0049] The section length is the length of the section over which the pipeline 44 is installed. If the pipeline 44 is installed over the entire length of the bridge 40, the section length is the bridge length. The bridge length can be considered to be the width of the river over which the bridge 40 is built.

[0050] The first equipment information 51 may further include the installation method of the pipeline 44 or the spacing of the support members 43. Examples of installation methods include a double-end supported beam, a cantilever beam, or a suspended structure. If the first equipment includes multiple pipelines 44, the first equipment information 51 may further include the total number of pipelines 44 or the total number of pipelines 44 arranged horizontally and vertically. The first equipment information 51 may further include, as items related to the strength deterioration of the pipelines 44, the year the pipelines 44 were installed on the bridge 40, the number of years since the pipelines 44 were installed on the bridge 40, the distance from the coast to the bridge 40, the amount of snow-melting agent sprayed on the pipelines 44, or the hours of sunshine at the location where the bridge 40 is located.

[0051] In this embodiment, the first equipment information 51 includes the pipe type, installation position, and section length, as well as the equipment name and number, and equipment coordinates for identifying the position of the pipeline 44.

[0052] The first water volume information 52 includes observed values ​​of the river water level and the rainfall.

[0053] The river water level is the water level of the river over which the bridge 40 is built. The river water level is, for example, the maximum water level from the start to the end of rainfall, or the water depth calculated from the estimated inundation depth in a hazard map published by an administrative agency, local government, or other organization. The river water level may be, for example, the water level measured from a predetermined water level reference at an observation station corresponding to the location where the bridge 40 is located. The observation station is, for example, a location nearest to the location of the bridge 40 where a measuring instrument capable of measuring the water level is installed. In this embodiment, the difference between the water level at an observation station near the bridge 40 and the flood danger water level at that observation station is defined as the exceedance water level, and this value is used as the river water level. For the water level and flood danger water level, for example, data from a hydrological and water quality database published by the Ministry of Land, Infrastructure, Transport and Tourism of Japan or data published by local governments is used. It is considered that the pipeline 44 will not be damaged by flooding unless subjected to water currents or floating debris. If the pier or abutment tilts, overturns, or moves due to scouring, the pipeline 44 will also be affected, but as mentioned above, it is difficult for a bridge contractor without detailed information about the bridge 40 and the river to assess whether the bridge 40 has been damaged. Therefore, damage to the pipeline 44 in such cases will not be predicted. The water pressure acting on the pipeline 44 or the impact of floating debris is greater the faster the river's flow velocity, so it is an important indicator, but at present there is little data made public by government ministries or local governments. In the future, there is a possibility that this will be used if more data on flow velocity or flow rate becomes available.

[0054] The first water volume information 52 may further include the water pressure of the river, the amount of drifting debris in the river, the flow velocity of the river, or the flow rate of the river at the point where the bridge 40 is located.

[0055] The rainfall is the rainfall at the point where the bridge 40 is located. In this embodiment, the rainfall is the accumulated rainfall at an observation station near the bridge 40 or the accumulated rainfall in 1-km mesh units published by the Japan Meteorological Agency. The accumulated time is thought to vary depending on conditions such as the topography and soil of the river basin. Therefore, it is desirable to input rainfall for multiple different accumulated times in the process of building the prediction model 11 using a machine learning method and select a rainfall value taking into consideration the contribution or performance of damage prediction. While the above-mentioned excess water level represents the water level at a certain point in time, the accumulated rainfall is thought to relate to the water level several hours to several days later.

[0056] In this embodiment, the first water quantity information 52 includes the excess water level and the observation station coordinates for identifying the location of the observation station that measures the water level. The first water quantity information 52 also includes the accumulated rainfall and the observation station coordinates for identifying the location of the observation station that measures the rainfall.

[0057] The first environmental information 53 includes one or more items such as river width, levee height, cross-sectional area, and riverbed gradient. These items affect the water level, flow rate, and flow velocity depending on the amount of water flowing into the river. Some of these items can be obtained by making an information disclosure request to the river administrator. The river width is the width of the river on which the bridge 40 is built. Qualitatively, a narrow river is more likely to have a high water level and a high flow velocity, while a wide river is more likely to have a high flow velocity. The levee height is the height of the levee on which the bridge 40 is built. The cross-sectional area is the area of ​​the area on which the bridge 40 is built. The riverbed gradient is the slope of the riverbed in the direction of the river flow of the river on which the bridge 40 is built.

[0058] The control unit 21 of the model construction device 20 constructs a prediction model 11 by performing machine learning using the acquired damage information 50, first facility information 51, first water volume information 52, and first environment information 53. The prediction model 11 is a model that predicts the possibility of the second facility being damaged when second facility information 61, second water volume information 62, and second environment information 63 as shown in FIG. 7 are input. The second facility information 61 is information indicating the characteristics of the second facility attached to the second bridge. The second water volume information 62 is information indicating the water volume observed at the observation station corresponding to the second bridge. The second environment information 63 is information indicating the environment in which the second bridge is installed.

[0059] In this embodiment, the second facility is the pipeline 44 shown in Fig. 2, similar to the first facility. Damage to the second facility refers to a state in which, similar to the damage to the first facility, the second facility cannot be used continuously unless repair or reinforcement measures are taken due to a change in the water level of the river.

[0060] Details of the second equipment information 61, the second water volume information 62, and the second environmental information 63 are similar to those of the first equipment information 51, the first water volume information 52, and the first environmental information 53, respectively, and therefore will not be explained here.

[0061] In this embodiment, the control unit 21 of the model construction device 20 uses the various information described above to construct a prediction model 11 in order to calculate the predicted damage probability Re of any attached pipeline. The damage information 50 includes whether or not the attached pipeline has been damaged in past floods caused by heavy rain or typhoons.

[0062] Examples of values ​​to be used are shown in Figure 4. The objective variable, damage presence, is represented by 1, and the non-damaged variable by 0. Among the values ​​used as explanatory variables, pipe type and rack location are qualitative data, while section length, excess water level, and accumulated rainfall are quantitative data. As shown in Figure 5, the qualitative data, pipe type and rack location, are divided into several categories, and dummy variables are assigned a value of 1 if the equipment corresponds to that category and 0 if it does not. The quantitative data, section length, excess water level, and accumulated rainfall, are different physical quantities with different value ranges, so they are standardized to have an average of 0 and a variance of 1. That is, when performing machine learning, the control unit 21 categorizes qualitative data values, such as pipe type or rack location, included in the first equipment information 51. When performing machine learning, the control unit 21 standardizes quantitative data values, such as section length, included in the first equipment information 51. When performing machine learning, the control unit 21 standardizes observed values, such as section length, included in the first equipment information 51. When performing machine learning, the control unit 21 standardizes observed values ​​included in the first water volume information 52. When performing machine learning, the control unit 21 may standardize the values ​​of the items included in the first environmental information 53.

[0063] The prediction model 11 may have any structure as long as it can predict the probability of damage Re or whether or not any attached pipeline is damaged, and may be, for example, a neural network. A neural network is a machine learning method that attempts to artificially realize information processing in the brain. An example of constructing a prediction model 11 using a neural network is shown in Figure 6.

[0064] The output damage probability Re is a value between 0 and 1. When predicting the binary value of whether or not a damage has occurred from the damage probability Re, the threshold value T of the damage probability Re is set according to the purpose. For example, if you want to predict the presence or absence of damage as thoroughly as possible, you can set the threshold value T to be small, and if you want to extract only facilities that are highly likely to be damaged, you can set the threshold value T to be large.

[0065] 6 shows the concept of model construction using a neural network, and the number of intermediate or hidden layers may be one or more. The number of input explanatory variables and intermediate layer nodes is not limited to those shown in the figure.

[0066] The operation of the damage prediction device 30 according to this embodiment will be described with reference to Fig. 7. The operation described below corresponds to the damage prediction method according to this embodiment.

[0067] The control unit 31 of the damage prediction device 30 acquires the second equipment information 61, the second water volume information 62, and the second environmental information 63. Specifically, if some or all of the second equipment information 61, the second water volume information 62, and the second environmental information 63 are pre-stored in the storage unit 32, the control unit 31 reads the information from the storage unit 32. Alternatively, the control unit 31 may receive some or all of the second equipment information 61, the second water volume information 62, and the second environmental information 63 from another device via the communication unit 33. Alternatively, the control unit 31 may accept input of some or all of the second equipment information 61, the second water volume information 62, and the second environmental information 63 from a user via the input unit 34.

[0068] The second facility information 61 is, for example, prepared in advance and stored in a database. When a heavy rain forecast or a flood warning or other alert is issued, forecast and real-time rainfall and water level data are obtained from government ministries or agencies or local governments. When predictions are used in disaster prevention planning, such as identifying weak points, data such as estimated rainfall and estimated flood depth used in flood damage estimates or hazard maps of the local government where the facility is located is used.

[0069] The control unit 31 of the damage prediction device 30 extracts information used to predict damage to the second facility from the second water volume information 62 by associating the second facility with the nearest water level and rainfall observation station using the facility coordinates included in the second facility information 61 and the observation station coordinates included in the second water volume information 62. If the second environmental information 63 also includes coordinates, the control unit 31 uses those coordinates to extract information used to predict damage to the second facility from the second environmental information 63, as with the second water volume information 62. If the second environmental information 63 does not include coordinates, the control unit 31 may extract information used to predict damage to the second facility from the second environmental information 63 using any other method.

[0070] The control unit 31 of the damage prediction device 30 acquires the prediction model 11 constructed by the model construction device 20. Specifically, the control unit 31 receives the prediction model 11 from the model construction device 20 via the communication unit 33. Alternatively, if the prediction model 11 is stored in advance in an external storage medium, the control unit 31 may read the prediction model 11 from the storage medium.

[0071] The control unit 31 of the damage prediction device 30 inputs the acquired second equipment information 61, second water volume information 62, and second environment information 63 into the acquired prediction model 11, and predicts the possibility that the second equipment will be damaged using the prediction model 11. Specifically, the control unit 31 inputs variables into the prediction model 11 to predict the damage probability Re of each attached pipeline. The control unit 31 compares the predicted damage probability Re with a threshold value T to predict whether each attached pipeline will be damaged.

[0072] The control unit 31 of the damage prediction device 30 outputs the predicted result of the possibility that the second facility will be damaged. Specifically, the control unit 31 displays the prediction result on the display unit 35 as a list in tabular form or as information on a map of a geographic information system. To make the prediction result easier to understand visually, the prediction result may be displayed in different colors according to the damage probability Re or whether or not the facility is damaged.

[0073] As described above, in this embodiment, the model construction device 20 combines information such as whether the pipeline 44 has been damaged by past floods, water volume information such as rainfall, water level, and flow velocity during floods, equipment information such as the type of pipe and installation location of the pipeline 44, and environmental information such as the topography, geology, and vegetation at the equipment installation location, and uses the combined information and machine learning techniques to construct the prediction model 11. When heavy rain is forecast or a flood warning is issued, the damage prediction device 30 uses the prediction model 11 to predict damage to equipment located in the relevant area.

[0074] According to this embodiment, it is possible to predict the probability of flood damage to the conduits 44 attached to the bridge 40. When the conduits 44 are installed in an area where rivers are expected to flood due to typhoons or heavy rain, applying this embodiment makes it possible to identify the attached conduits that are likely to be damaged, enabling the following measures: (1) Preventing service outages by switching lines, cable routes, flow paths, etc. in advance (2) Formulating an efficient emergency inspection plan in which inspections are prioritized starting with facilities that are more likely to be damaged (3) Preparing and deploying restoration materials and personnel in quantities estimated based on damage predictions.

[0075] For facilities that are predicted to have a high probability of being damaged even under normal circumstances, a highly reliable network can be constructed by implementing measures such as relocating them to a safe location, securing alternative routes, securing alternative means, or reinforcing them.

[0076] As an example, predictions were made using the prediction model 11 according to this embodiment for past disaster cases.

[0077] As examples of past disasters, we looked at Typhoon Hagibis (the "East Japan Typhoon") in 2019 and the heavy rains of July 2020. The East Japan Typhoon brought heavy rain to a wide area of ​​Japan, from Shizuoka to Iwate, from October 10 to 13, 2019. The heavy rains of July 2020 were a disaster caused by a front that remained stationary near Japan, resulting in heavy rains in southern and northern Kyushu, Tokai, and Tohoku from July 3 to 31 of the same year.

[0078] Of the various data from approximately 580 facilities in 10 prefectures at the time, 4 / 5 was used as learning data to build a prediction model 11, and the remaining 1 / 5 was input into the prediction model 11 as evaluation data. As a result, the prediction success rate was over 80% overall. Figure 8 shows an excerpt of the prediction results for Kyushu. The recall rate, which is the percentage of correctly predicted damage out of the actual damage that was detected, was also over 80%.

[0079] The following additional notes are provided regarding the above-described embodiments.

[0080] (Supplementary paragraph 1) A model construction device comprising a control unit that acquires damage information indicating whether a first facility attached to a first bridge was damaged when a flood occurred, first facility information indicating the characteristics of the first facility, first water volume information indicating the water volume observed at an observation station corresponding to the first bridge when the flood occurred, and first environmental information indicating the environment in which the first bridge is installed, and performs machine learning using the acquired damage information, first facility information, first water volume information, and first environmental information to construct a predictive model that predicts the possibility of the second facility being damaged when second facility information indicating the characteristics of a second facility attached to a second bridge, second water volume information indicating the water volume observed at an observation station corresponding to the second bridge, and second environmental information indicating the environment in which the second bridge is installed is input. (Supplementary Item 2) The model construction device according to Supplementary Item 1, wherein the first facility and the second facility are each a pipeline, and the first facility information and the second facility information each include the type of the pipeline, the position where the pipeline is attached, or the length of the section where the pipeline is attached. (Supplementary Item 3) The model construction device according to Supplementary Item 1 or Supplementary Item 2, wherein the first water volume information includes observed values ​​of river water level and rainfall, and the control unit standardizes the observed values ​​included in the first water volume information when performing the machine learning. (Supplementary Item 4) The model construction device according to any one of Supplementary Items 1 to 3, wherein the first environmental information and the second environmental information each include river width, levee height, cross-sectional area, or riverbed gradient.

[0081] The present disclosure is not limited to the above-described embodiments. For example, two or more blocks shown in the block diagram may be integrated, or one block may be divided. Two or more steps shown in the flowchart may be executed in parallel or in a different order, instead of being executed in chronological order as described, depending on the processing capabilities of the device executing each step, or as needed. Other modifications are possible within the scope of the present disclosure.

[0082] REFERENCE SIGNS LIST 10 System 11 Prediction model 20 Model construction device 21 Control unit 22 Memory unit 23 Communication unit 24 Input unit 30 Disaster prediction device 31 Control unit 32 Memory unit 33 Communication unit 34 Input unit 35 Display unit 40 Bridge 41 Deck slab 42 Bridge girder 43 Support member 44 Pipeline 50 Disaster information 51 First facility information 52 First water volume information 53 First environmental information 61 Second facility information 62 Second water volume information 63 Second environmental information

Claims

1. A model construction device comprising a control unit that acquires damage information indicating whether a first facility attached to a first bridge was damaged during a flood, first facility information indicating the characteristics of the first facility, first water volume information indicating the water volume observed at an observation station corresponding to the first bridge during the flood, and first environmental information indicating the environment in which the first bridge is installed, and performs machine learning using the acquired damage information, first facility information, first water volume information, and first environmental information to construct a predictive model that predicts the possibility of the second facility being damaged when second facility information indicating the characteristics of a second facility attached to a second bridge, second water volume information indicating the water volume observed at an observation station corresponding to the second bridge, and second environmental information indicating the environment in which the second bridge is installed is input.

2. The model construction device of claim 1, wherein the first equipment and the second equipment are each a pipeline, and the first equipment information and the second equipment information each include the type of the pipeline, the position where the pipeline is installed, or the length of the section where the pipeline is installed.

3. The model construction device described in claim 1, wherein the first water volume information includes observed values ​​of river water level and rainfall, and the control unit standardizes the observed values ​​included in the first water volume information when performing the machine learning.

4. The model construction device according to claim 1, wherein the first environmental information and the second environmental information each include a river width, a bank height, a cross-sectional area, or a riverbed gradient.

Images