Learning device, estimation device, learning method, estimation method, and program

Abstract

To improve the technique for predicting damage to equipment due to a disaster.SOLUTION: A learning apparatus 10 includes: a first model generation unit 112 which generates a first model for calculating a degree of deterioration of equipment by machine learning that uses, as training data, deterioration data including information on the equipment, an environmental condition of the equipment, and a degree of deterioration of the equipment; and a second model generation unit 114 which generates a second model for calculating a degree of damage to the equipment by machine learning that uses, as training data, damage data including a disaster condition of a disaster on the equipment, a degree of damage due to the disaster, and the degree of deterioration calculated by the first model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a learning device, an estimation device, a learning method, an estimation method, and a program. [Background technology]

[0002] Conventionally, there are methods for predicting damage to infrastructure facilities in the event of a disaster. Regarding the prediction of damage in the event of an earthquake, there is a method for making predictions based on a statistical function of the maximum ground surface velocity (earthquake damage rate function) (see, for example, Non-Patent Documents 1 and 2). Regarding the prediction of damage in the event of a disaster other than an earthquake, a method has been proposed for predicting damage to utility poles, power lines, etc. by inputting typhoon information (see, for example, Non-Patent Document 3). It has also been suggested that deterioration of facilities may make them more susceptible to damage (see, for example, Non-Patent Document 4). [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Yoshihisa Maruyama and 1 other collaborators, "Improvement of Macroscopic Water Distribution Pipe Damage Prediction Formulas Taking into Account Recent Earthquake Damage Data," Journal of the Japan Society of Civil Engineers, A1 (Structural and Earthquake Engineering), Vol. 65, No. 1 (Journal of Earthquake Engineering, Vol. 30), pp. 565-574, 2009. [Non-patent document 2] Takeshi Yamamura and three others, "A Method for Creating Earthquake Damage Rate Functions for Road Bridges," Journal of Earthquake Engineering, Japan Society of Civil Engineers, pp.674-679, 2007 [Non-patent document 3] Chubu Electric Power, "Typhoon Wind Speed ​​Prediction System (RAMP-T)", [online], [Retrieved March 22, 2022], Internet <URL: https: / / www.chuden.co.jp / resource / seicho_kaihatsu / kaihatsu / techno / techno_naiyou2013 / techno_gaiyou_16.pdf> [Non-patent document 4] Akira Ito and two others, "Quantification and Prediction of Seismic Performance Degradation Due to Corrosion in Underground Telecommunication Pipes," Materials Testing Technology, Vol. 59, No. 4, pp. 188-193, 2014 Summary of the Invention [Problem to be solved by the invention]

[0004] Deterioration of equipment may make it more susceptible to damage in the event of a disaster, but conventional technology has not been able to quantitatively evaluate the degree of susceptibility to damage or reflect equipment deterioration in damage predictions. Unless data on deterioration is available for all equipment, it is not possible to quantitatively predict damage to that equipment.

[0005] In view of the above circumstances, an object of the present invention is to improve the technology for predicting damage to facilities due to disasters. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems, the learning device of the present invention includes a first model generation unit that generates a first model that calculates the degree of deterioration of the equipment through machine learning using deterioration data as training data, the deterioration data including information about the equipment, the environmental conditions of the equipment, and the degree of deterioration of the equipment, and a second model generation unit that generates a second model that calculates the degree of damage to the equipment through machine learning using damage data as training data, the damage conditions of a disaster to the equipment, the degree of damage caused by the disaster, and the degree of deterioration calculated by the first model.

[0007] The estimation device according to the present invention also includes a model acquisition unit that acquires a first model that calculates the degree of deterioration of the equipment, the first model being learned by machine learning using as training data deterioration data including information about the equipment, environmental conditions of the equipment, and the degree of deterioration of the equipment, and a second model that calculates the degree of damage to the equipment, the second model being learned by machine learning using as training data damage data including disaster conditions of a disaster to the equipment, the degree of damage caused by the disaster, and the degree of deterioration calculated by the first model, and an estimation unit that applies target data including information about the equipment of a target equipment, the environmental conditions of the target equipment, and disaster conditions of the target equipment to the first model and the second model, and estimates the degree of damage to the target equipment.

[0008] Furthermore, the learning method of the present invention is a learning method executed by a learning device, and includes a first model generation step of generating a first model that calculates the degree of deterioration of the equipment through machine learning using, as training data, deterioration data including information about the equipment, environmental conditions of the equipment, and the degree of deterioration of the equipment, and a second model generation step of generating a second model that calculates the degree of damage to the equipment through machine learning using, as training data, damage data including disaster conditions of a disaster to the equipment, the degree of damage caused by the disaster, and the degree of deterioration calculated by the first model.

[0009] Furthermore, the estimation method according to the present invention is an estimation method executed by an estimation device, and includes an acquisition step of acquiring: a first model that calculates the degree of deterioration of the equipment, which is learned by machine learning using, as training data, deterioration data including information about the equipment, environmental conditions of the equipment, and the degree of deterioration of the equipment; and a second model that calculates the degree of damage to the equipment, which is learned by machine learning using, as training data, disaster conditions of a disaster to the equipment, the degree of damage caused by the disaster, and damage data including the degree of deterioration calculated by the first model; and an estimation step of applying target data including information about the equipment of a target equipment, the environmental conditions of the target equipment, and disaster conditions of the target equipment to the first model and the second model, and estimating the degree of damage to the target equipment.

[0010] Furthermore, the program according to the present invention causes a computer to function as the learning device according to the present invention. [Effects of the Invention]

[0011] According to the present invention, it is possible to improve the technology for predicting damage to facilities due to disasters. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a diagram showing a schematic configuration of a damage prediction system according to an embodiment of the present invention. [Figure 2] FIG. 10 is a diagram showing an example of degradation data according to the embodiment; [Figure 3] FIG. 10 is a diagram illustrating an example of a portion of a gradient boosting decision tree as an example of a learning algorithm for a first model. [Figure 4] FIG. 10 is a diagram showing an example of damage data according to the present embodiment. [Figure 5] FIG. 10 is a diagram illustrating an example of explanatory variables used in generating a first model. [Figure 6] FIG. 10 is a diagram illustrating an example of target data according to the present embodiment. [Figure 7A] FIG. 2 is a diagram illustrating the operation of the damage prediction system according to the present embodiment. [Figure 7B] FIG. 2 is a diagram illustrating the operation of the damage prediction system according to the present embodiment. [Figure 8] FIG. 10 is a diagram showing a schematic configuration of a damage prediction system according to a modified example. [Figure 9] FIG. 10 is a diagram illustrating an example of target data according to a modified example. [Figure 10A] FIG. 10 is a diagram illustrating an example of divided target data. [Figure 10B] FIG. 10 is a diagram illustrating an example of divided target data. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. The embodiments described below are examples of the configuration of the present invention, and the present invention is not limited to the following embodiments.

[0014] <Outline of Damage Prediction System 1> FIG. 1 is a diagram showing the configuration of a damage prediction system 1 according to this embodiment. As shown in FIG. 1, the damage prediction system 1 includes a learning device 10 and an estimation device 20. The learning device 10 and the estimation device 20 are communicably connected to a network 30, which may include, for example, the Internet and a mobile communication network, via wired or wireless communication. There are no particular limitations on the communication method for transmitting and receiving information between the devices. The learning device 10 and the estimation device 20 may be integrated.

[0015] The learning device 10 and the estimation device 20 are computers such as servers that belong to a cloud computing system or other computing systems.

[0016] The network 30 may include the Internet, at least one wide area network (WAN), at least one metropolitan area network (MAN), or any combination thereof. The network 30 may also 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 local area network (LAN), a satellite communication network, or a terrestrial microwave network.

[0017] First, an overview of this embodiment will be described, and details will be provided later. The learning device 10 generates a first model that calculates the degree of deterioration of the facility through machine learning using, as training data, deterioration data that includes information about the facility, environmental conditions of the facility, and the degree of deterioration of the facility. The learning device 10 generates a second model that calculates the degree of damage to the facility through machine learning using, as training data, damage data that includes disaster conditions of a disaster to the facility, the degree of damage caused by the disaster, and the degree of deterioration calculated by the first model. The learning device 10 outputs the generated model to the estimation device 20.

[0018] The estimation device 20 estimates the degree of damage to the target facility using the first model and the second model generated by the learning device 10. The estimation device 20 applies target data including information about the facility of the target facility, environmental conditions of the target facility, and disaster conditions of the target facility to the first model and the second model, and estimates the degree of damage to the target facility.

[0019] According to this embodiment, it is possible to predict the degree of deterioration of the target equipment and then accurately predict the degree of damage to the target equipment due to a disaster, thereby improving the technology for predicting damage to equipment due to a disaster.

[0020] <Configuration of learning device 10> As shown in FIG. 1, the learning device 10 includes a control unit 11, a storage unit 12, a communication unit 13, an input unit 14, and an output unit 15.

[0021] The storage unit 12 includes one or more memories, and may include, for example, semiconductor memory, magnetic memory, optical memory, etc. Each memory included in the storage unit 12 may function, for example, as a main memory device, an auxiliary memory device, or a cache memory. The storage unit 12 stores any information used in the operation of the learning device 10. The storage unit 12 does not necessarily need to be provided inside the learning device 10, but may be configured to be provided externally to the learning device 10.

[0022] Communication unit 13 includes one or more communication interfaces that connect to network 30. The communication interfaces may be compatible with, for example, a mobile communication standard, a wired LAN standard, or a wireless LAN standard, but are not limited to these and may be compatible with any communication standard. Communication unit 13 receives information used in the operation of learning device 10 and transmits information obtained by the operation of learning device 10.

[0023] Input unit 14 includes at least one input interface. The input interface may be, for example, a physical key, a capacitive key, a pointing device, a touch screen integrated with a display, or a microphone. Input unit 14 accepts operations to input information used in the operation of study device 10. Input unit 14 may be connected to study device 10 as an external input device instead of being provided in study device 10. Any connection method may be used, for example, Universal Serial Bus (USB), High-Definition Multimedia Interface (HDMI) (registered trademark), or Bluetooth (registered trademark).

[0024] Output unit 15 includes at least one output interface. The output interface is, for example, a display or a speaker. The display is, for example, an LCD (Liquid Crystal Display) or an organic EL (Electro Luminescence) display. Output unit 15 outputs information obtained by the operation of learning device 10. Output unit 15 may be connected to learning device 10 as an external output device instead of being provided in learning device 10. Any connection method can be used, for example, USB, HDMI (registered trademark), or Bluetooth (registered trademark).

[0025] The control unit 11 is realized by a control and arithmetic circuit (controller). The control and arithmetic circuit may be configured with dedicated hardware such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field-Programmable Gate Array), or may be configured with a processor, or may be configured to include both. The control unit 11 executes processes related to the operation of the learning device 10 while controlling each part of the learning device 10. The control unit 11 can send and receive information to and receive information from external devices via the communication unit 13 and the network 30.

[0026] The control unit 11 includes a degradation data acquisition unit 111, a first model generation unit 112, a damage data acquisition unit 113, and a second model generation unit 114.

[0027] (Generation of the first model) The degradation data acquisition unit 111 acquires degradation data. The degradation data acquisition unit 111 may acquire the degradation data by reading it from the storage unit 12, or may acquire degradation data input by the user via the input unit 14.

[0028] Fig. 2 shows an example of deterioration data. In Fig. 2, the deterioration data is shown in a table format, but the format of the deterioration data is not limited to this. The deterioration data for each piece of equipment includes information about the equipment, the environmental conditions of the equipment, and the degree of deterioration of the equipment.

[0029] The term "facilities" includes infrastructure facilities such as pipelines and utility poles. Pipes are, for example, communication pipes that protect communication cables, and include underground pipes buried underground. Pipes may be water pipes, gas pipes, electricity pipes, etc. Utility poles are installed above ground, and power lines, communication lines, etc. are installed thereon. However, the term "facilities" is not limited to these, and may include underground facilities such as manholes and cable tunnels, and above-ground facilities such as bridges, roads, tunnels, and buildings. Referring to FIG. 2, the degradation data includes two types of facilities: pipelines and utility poles.

[0030] The information about the facility includes information indicating the material of the facility, the number of years since the facility was constructed, etc. There may be multiple pieces of information about the facility, and it may include any information such as the repair history of the facility, the size of the facility, the characteristics of the facility, etc. For example, the information about the facility may include the length of the pipeline as the facility, the material of the joints, the pipe diameter, the buried depth, etc.

[0031] The environmental conditions include information indicating the environment of the location where the equipment is installed. There may be multiple environmental conditions, and may include, for example, the topographical classification of the area where the equipment is located, the ground, the type of rock or soil in the ground, climate, latitude, longitude, altitude, and other information about the location where the equipment is installed. The topography includes lowlands, highlands, plains, plateaus, hills, mountains, mountain ranges, valleys, and the like. The ground includes bedrock layers, hard gravel layers, and the like. The climate includes average temperature, average humidity, average hours of sunshine, average amount of solar radiation, average wind speed, amount of airborne salt, annual precipitation, and the like. Without being limited to these, the environmental conditions may also include the components of the soil on the land where the equipment is installed, underground temperature, water content, acidity (pH), and the like.

[0032] The degree of deterioration includes information indicating the results of the equipment inspection. The degree of deterioration may be a continuous value such as the length, depth, or area of ​​the damaged part of the equipment, or may be a discrete value indicating the presence or absence or degree of deterioration in stages.

[0033] Referring to Figure 2, in the line below the title line showing deterioration data for the pipeline, the information about the equipment is that the material of the equipment is "steel pipe" and the age is "3 years." The environmental conditions for the pipeline are that the ground is a "hard gravel layer" and the annual precipitation is "XX mm." As for the degree of deterioration of the pipeline, the inspection results show that the degree of deterioration is "high." In the next line showing deterioration data for the utility pole, the information about the equipment is that the material of the equipment is "concrete" and the age is "5 years." The environmental conditions for the pole are that the ground is a "rock layer" and the annual precipitation is "YY mm." As for the degree of deterioration of the utility pole, the inspection results show that the degree of deterioration is "medium."

[0034] The information about the facility and the environmental conditions included in the deterioration data may differ for each facility. For example, an underground pipeline may include the material of the joints as information about the facility and the amount of water content in the ground as an environmental condition, while a utility pole may include the number of distribution lines as information about the facility and the annual average wind speed as an environmental condition.

[0035] The degradation data acquisition unit 111 can input any value for any item in the degradation data that does not have a value. The degradation data acquisition unit 111 may acquire any value from the storage unit 12 or an external device via the communication unit 13. For example, suppose a value indicating the ground condition of a pipeline is unknown in the degradation data. In this case, the degradation data acquisition unit 111 may acquire information indicating the location of the pipeline, identify other equipment within a predetermined distance from the location, and input the value indicating the ground condition of the identified equipment as the value indicating the ground condition of the pipeline. In this case, information indicating the location of each equipment may be included in the degradation data or may be included in any other database. Alternatively, if there is a missing value in the acquired degradation data, the degradation data acquisition unit 111 may display a message on the output unit 15 prompting the user to input it, acquire a value entered by the user in the input unit 14, and reflect the value in the degradation data. This improves the accuracy of the generated first model even when there is a missing value in the degradation data.

[0036] The degradation data acquisition unit 111 outputs the acquired degradation data to the first model generation unit 112.

[0037] The first model generation unit 112 generates a first model based on the deterioration data acquired by the deterioration data acquisition unit 111, in which information about the equipment and environmental conditions are quantified as explanatory variables and the degree of deterioration is quantified as a target variable. The first model generation unit 112 may generate the first model by any machine learning method such as a neural network or regression analysis, a deep learning method, or a statistical method. The first model calculates the degree of deterioration of the equipment using any discrete or continuous value. The first model may calculate the degree of deterioration using two values, 0 and 1, representing the presence or absence of deterioration, or using a categorical value. The categorical value may be a value other than a numerical value, such as A, B, or C.

[0038] The following equation 1 expresses the function of the first model according to this embodiment as the inner product of coefficient vector a and vector x. As shown in equations 2 and 3, coefficient vector a and vector x each have m elements. The number of elements, m, represents the total number of pieces of information about the equipment and environmental conditions. The first model calculates the degree of deterioration f(x) by multiplying vector x, which serves as an explanatory variable, by coefficient vector a. The coefficient vector a can be obtained by applying any fitting method. The first model generation unit 112 can generate a first model for each type of equipment.

[0039]

number

number

number

[0040] The first model generated by the first model generation unit 112 is not limited to the above. For example, the first model generation unit 112 may generate the first model by ensemble learning such as gradient boosting. FIG. 3 shows an example of a portion of a gradient boosting decision tree as an example of the learning algorithm of the first model. The decision tree finally classifies the target equipment into categories A to E by comparing with the values ​​of average temperature, average altitude, average precipitation, and topography. A to E are category values ​​indicating the degree of deterioration calculated by the first model.

[0041] The first model generation unit 112 stores the generated first model in the storage unit 12. The first model generation unit 112 outputs to the damage data acquisition unit 113 information indicating the degree of damage calculated for each piece of equipment.

[0042] (Generation of the second model) The damage data acquisition unit 113 acquires damage data. The damage data acquisition unit 113 can acquire the damage data by generating damage data by adding the degree of deterioration of each piece of equipment output from the first model generation unit 112 to the disaster conditions and the degree of damage for each piece of equipment.

[0043] Fig. 4 shows an example of damage data. Fig. 4 shows the damage data in a table format, but the format of the damage data is not limited to this. As shown in Fig. 4, the damage data includes, for each piece of equipment, the disaster conditions, the extent of damage caused by the disaster, and the degree of deterioration calculated by the first model.

[0044] The disaster conditions include information indicating various measurement values ​​related to a disaster that has occurred to the facility. There may be multiple disaster conditions, and for example, if the disaster is seismic motion, the conditions may include the maximum ground surface velocity, maximum ground surface acceleration, and average velocity of waves propagating through the ground at the location where the facility is installed. Without being limited to these, the disaster conditions may also include maximum ground surface displacement, measured seismic intensity, real-time seismic intensity, SI (Spectral Intensity) values, and the like. If the disaster is liquefaction, the disaster conditions may include liquefaction history and liquefaction index, which indicate whether liquefaction has occurred in the past. If the disaster is a typhoon, the disaster conditions may include wind speed, rainfall, and flying object history, which indicates whether any flying objects have been detected near the facility.

[0045] The extent of damage caused by the disaster includes information indicating the results of inspections of the equipment after the disaster. The extent of damage may be a continuous value such as the angle of tilt of the equipment, the amount of subsidence of the land where the equipment is located, the length, depth, or area of ​​the damaged area, or may be a discrete value indicating the presence or absence or degree of damage in stages.

[0046] The degree of deterioration is a value calculated by the first model for the corresponding equipment. The damage data acquisition unit 113 may identify the corresponding equipment by any method. The damage data acquisition unit 113 may identify, from the equipment in the damage data, equipment whose characteristics, such as the type of equipment, are similar to those of the equipment for which the first model has calculated the degree of deterioration, and may add the degree of deterioration calculated by the first model to the damage data of the identified equipment.

[0047] Referring to Figure 4, in the line showing damage data for pipelines under the title line, the maximum ground velocity as a disaster condition for earthquake motion to pipelines is "K cm / s" and the maximum ground acceleration is "M cm / s" 2 ". The inspection results show that the degree of damage to the pipeline was "small." The degree of deterioration of the pipeline calculated by the first model, "large," is added to the rightmost column. In the next row showing damage data for utility poles, the maximum ground velocity as a disaster condition for utility poles is "L cm / s," and the maximum ground acceleration is "N cm / s 2 ". The inspection results show that the degree of damage to the utility pole was "large." The degree of deterioration of the utility pole calculated by the first model, "medium," is added to the right-most column.

[0048] The disaster conditions included in the damage data may differ depending on the type of facility. For example, if the facility is an above-ground structure such as a utility pole, the disaster conditions for when the disaster is a typhoon may include a flying object history indicating the presence or absence of flying objects, whereas if the facility is an underground pipeline, the flying object history may not be included as a disaster condition.

[0049] The damage data acquisition unit 113 can input any value for any item of the damage data that does not have a value entered. The damage data acquisition unit 113 may be able to acquire any value from the storage unit 12 or an external device via the communication unit 13. For example, suppose that a value indicating the maximum ground surface velocity of a pipeline is unknown among the damage data. In this case, the damage data acquisition unit 113 may acquire information indicating the location of the pipeline, identify other facilities within a predetermined distance from the location, and input the value indicating the maximum ground surface velocity of the identified facilities as the value indicating the maximum ground surface velocity of the pipeline. In this case, information indicating the location of each facility may be included in the damage data or may be included in any other database. Without being limited to this, the damage data acquisition unit 113 may be able to acquire a value entered by a user and input the value into the damage data, similar to the deterioration data acquisition unit 111.

[0050] The damage data acquisition unit 113 outputs the acquired damage data to the second model generation unit 114.

[0051] The second model generation unit 114 generates a second model in which the disaster conditions and the degree of deterioration contained in the acquired damage data are quantified as explanatory variables, and the degree of damage is quantified as a target variable. Similar to the first model generation unit 112, the second model generation unit 114 may generate the second model using any machine learning method such as a neural network or regression analysis, a deep learning method, or a statistical method. The second model calculates the degree of damage to the equipment using any discrete or continuous value. The second model may calculate the degree of damage using two values, 0 and 1, representing the presence or absence of damage, or using a categorical value. The categorical value may be a value other than a numerical value, such as A, B, or C.

[0052] The following equation 4 expresses the function of the second model according to this embodiment as the inner product of coefficient vector b and vector p. As shown in equations 5 and 6, coefficient vector b and vector p each have n+1 elements. The number of elements, n, represents the total number of disaster conditions. The n+1th element represents the degree of deterioration calculated by the first model. The second model calculates the degree of damage g(p) by multiplying vector p, which serves as an explanatory variable, by coefficient vector b. The coefficient vector b can be obtained by applying any fitting method. The second model generation unit 114 can generate second models for the number of combinations of equipment types and disasters.

[0053]

number

number

number

[0054] The second model generated by the second model generation unit 114 is not limited to the above. For example, like the first model generation unit 112, the second model generation unit 114 may generate the second model by ensemble learning such as gradient boosting.

[0055] The second model generation unit 114 stores the generated second model in the storage unit 12.

[0056] The control unit 11 reads out the first model and the second model from the storage unit 12. The control unit 11 transmits the read out first model and second model to the estimation device 20 via the communication unit 13 in response to a request from the estimation device 20 or automatically.

[0057] The first model generation unit 112 may display the explanatory variables used for the generated first model, and the second model generation unit 114 may display the explanatory variables used for the generated second model, via the output unit 15. In this case, the first model generation unit 112 or the second model generation unit 114 may display the explanatory variables for all generated models in descending order of coefficient value. FIG. 5 is a diagram showing environmental conditions as explanatory variables of the first model, displayed by the first model generation unit 112 via the output unit 15, in descending order of coefficient value. Referring to FIG. 5, it can be seen that the average amount of solar radiation, as an environmental condition, contributed greatly to the calculation of the degree of deterioration, and the environmental condition with the next highest contribution was the "lowland" topography classification. Being able to check the explanatory variables allows the user to consider the accuracy of the first and second models and use them as information to determine whether to adopt the calculation results.

[0058] <Configuration of Estimation Device 20> Referring back to FIG. 1, the estimation device 20 includes a control unit 21, a storage unit 22, a communication unit 23, an input unit 24, and an output unit 25.

[0059] The storage unit 22 includes one or more memories, and may include, for example, a semiconductor memory, a magnetic memory, an optical memory, etc. Each memory included in the storage unit 22 may function, for example, as a main storage device, an auxiliary storage device, or a cache memory. The storage unit 22 stores any information used in the operation of the estimation device 20. The storage unit 22 does not necessarily need to be provided inside the estimation device 20, and may be configured to be provided outside the estimation device 20.

[0060] The communication unit 23 includes one or more communication interfaces that connect to the network 30. The communication interfaces may be compatible with, for example, a mobile communication standard, a wired LAN standard, or a wireless LAN standard, but are not limited to these and may be compatible with any communication standard. The communication unit 23 receives information used in the operation of the estimation device 20 and transmits information obtained by the operation of the estimation device 20.

[0061] The input unit 24 includes at least one input interface. The input interface is, for example, a physical key, a capacitance key, a pointing device, a touch screen integrated with a display, or a microphone. The input unit 24 accepts an operation to input information used in the operation of the estimation device 20. The input unit 24 may be connected to the estimation device 20 as an external input device instead of being provided in the estimation device 20. Any connection method can be used, for example, USB, HDMI (registered trademark), Bluetooth (registered trademark), or the like.

[0062] The output unit 25 includes at least one output interface. The output interface is, for example, a display or a speaker. The display is, for example, an LCD or an organic EL display. The output unit 25 outputs information obtained by the operation of the estimation device 20. The output unit 25 may be connected to the estimation device 20 as an external output device instead of being provided in the estimation device 20. Any connection method can be used, for example, USB, HDMI (registered trademark), or Bluetooth (registered trademark).

[0063] The control unit 21 is realized by a control and arithmetic circuit (controller). The control and arithmetic circuit may be configured by dedicated hardware such as an ASIC or FPGA, or may be configured by a processor, or may be configured by including both. The control unit 21 executes processing related to the operation of the estimation device 20 while controlling each part of the estimation device 20. The control unit 21 can send and receive information to and receive information from external devices via the communication unit 23 and the network 30.

[0064] The control unit 21 includes a model acquisition unit 211 , a target data acquisition unit 212 , and an estimation unit 213 .

[0065] (Estimation of the extent of damage to the target facilities) The model acquisition unit 211 acquires the first model generated by the first model generation unit 112 and the second model generated by the second model generation unit 114 by receiving them from the learning device 10 via the communication unit 23. The model acquisition unit 211 outputs the acquired first model and second model to the estimation unit 213.

[0066] The target data acquisition unit 212 acquires target data including information about the equipment of the target equipment, the environmental conditions of the target equipment, and the disaster conditions of the target equipment. Any method may be used to acquire the target data. For example, the target data acquisition unit 212 may display a message via the output unit 25 prompting the user to input the environmental conditions of the target equipment and the disaster conditions of the target equipment, and acquire values ​​input by the user via the input unit 24 as target data. The target data acquisition unit 212 may communicate with a terminal device held by the user, receive values ​​input by the user to the terminal device from the terminal device, and acquire the values ​​as target data.

[0067] FIG. 6 is an example of target data. In FIG. 6, the target data is shown in table format, but the format of the target data is not limited to this. Referring to FIG. 6, in the row below the title row showing target data for the pipeline as the target facility, the material of the facility as information about the facility is "vinyl pipe" and the age is "5 years." The ground as the environmental condition of the pipeline is "rock layer" and the annual precipitation is "VV mm." The maximum ground surface velocity as the disaster condition of earthquake motion for the pipeline is "P cm / s" and the maximum ground surface acceleration is "R cm / s." 2 ". In the next line showing the target data for the utility pole, the material of the equipment as information about the facility is "concrete" and the age is "10 years". The ground as environmental conditions for the utility pole is "rock layer" and the annual precipitation is "WW mm". The maximum ground surface velocity as disaster conditions of earthquake motion for the utility pole is "Q cm / s" and the maximum ground surface acceleration is "S cm / s".2 The target data acquisition unit 212 outputs the acquired target data to the estimation unit 213.

[0068] The estimation unit 213 applies the target data acquired by the target data acquisition unit 212 to the first model and the second model acquired by the model acquisition unit 211, and estimates the degree of damage to the target equipment. Specifically, the estimation unit 213 first quantifies information about the equipment of the target equipment contained in the target data and the environmental conditions of the target equipment, and inputs these into a first model for the equipment corresponding to the target equipment. The estimation unit 213 acquires the degree of deterioration output by the first model, quantifies the degree of deterioration and the disaster conditions of the target equipment contained in the target data, and inputs these into a second model for the target equipment and the disaster to the target equipment. The estimation unit 213 acquires the degree of damage calculated by the second model.

[0069] For example, suppose the estimation unit 213 estimates the degree of damage to a pipeline from the target data in FIG. 6 . In this case, the estimation unit 213 first quantifies the values ​​of information about the pipeline, such as “material,” “age,” etc., and environmental conditions, such as “ground,” “annual precipitation,” etc., and inputs these values ​​into a first model for the pipeline acquired from the learning device 10. The degree of deterioration calculated by the first model is “medium.” The estimation unit 213 quantifies the calculated degree of deterioration and the values ​​of the pipeline disaster conditions included in the target data, such as “maximum ground surface velocity,” “maximum ground surface acceleration,” etc., and inputs these values ​​into a second model for the pipeline and seismic motion acquired from the learning device 10. The degree of damage calculated by the second model is “large.” The estimation unit 213 acquires the calculated degree of damage. The estimation unit 213 similarly estimates the degree of damage to utility poles. In this way, the estimation unit 213 estimates the degree of damage to the target facility.

[0070] The estimation unit 213 displays the estimated extent of damage via the output unit 25. Any method may be used for the display. For example, the estimation unit 213 may display the position of the target equipment on a map via the output unit 25, and display the estimated extent of damage above the position of the equipment.

[0071] <Program> A computer capable of executing program instructions can also be used to function as the learning device 10 or the estimation device 20. Here, the computer may be a general-purpose computer, a dedicated computer, a workstation, a PC (Personal Computer), an electronic notepad, etc. The program instructions may be program code, code segments, etc. for performing the necessary tasks.

[0072] The computer includes a processor, a storage unit, an input unit, an output unit, and a communication interface. The processor may be a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an SoC (System on a Chip), or the like, and may be configured with multiple processors of the same or different types. The processor reads and executes programs from the storage unit to control the above components and perform various arithmetic processing. Note that at least a portion of these processing contents may be implemented by hardware. The input unit is an input interface that accepts user input operations and acquires information based on the user operations, such as a pointing device, keyboard, or mouse. The output unit is an output interface that outputs information, such as a display or speaker. The communication interface is an interface for communicating with external devices.

[0073] The program may be recorded on a computer-readable recording medium. Using such a recording medium, the program can be installed on a computer. Here, the recording medium on which the program is recorded may be a non-transitory recording medium. The non-transitory recording medium is not particularly limited, and may be, for example, a CD-ROM, a DVD-ROM, or a USB memory. Furthermore, the program may be downloaded from an external device via a network.

[0074] <Operation of Damage Prediction System 1> 7A and 7B, the operation of the damage prediction system 1 including the learning device 10 and the estimation device 20 according to this embodiment will be described. Among the operations of the damage prediction system 1, the operation of the learning device 10 corresponds to the learning method according to this embodiment, and the operation of the estimation device 20 corresponds to the estimation method according to this embodiment.

[0075] 7A, the degraded data acquisition unit 111 of the learning device 10 acquires degraded data. Any method may be employed to acquire the degraded data.

[0076] In step S2, the degradation data acquisition unit 111 outputs the acquired degradation data to the first model generation unit 112.

[0077] In step S3, the first model generation unit 112 generates a first model in which information about the equipment and environmental conditions are quantified as explanatory variables and the degree of deterioration is quantified as a response variable, based on the deterioration data acquired by the deterioration data acquisition unit 111. The first model generation unit 112 may generate the first model by any machine learning method such as a neural network or regression analysis, a deep learning method, or a statistical method.

[0078] In step S4, the first model generation unit 112 stores the generated first model in the storage unit 12, and outputs to the damage data acquisition unit 113 information indicating the degree of damage calculated for each facility.

[0079] In step S5, the damage data acquisition unit 113 acquires damage data. The damage data acquisition unit 113 can acquire the damage data by generating damage data by adding information indicating the degree of deterioration of each piece of equipment output from the first model generation unit 112 to information indicating the disaster conditions and the degree of damage for each piece of equipment.

[0080] In step S6, the damage data acquisition unit 113 outputs the acquired damage data to the second model generation unit 114.

[0081] In step S7, the second model generation unit 114 generates a second model in which the disaster conditions and the degree of deterioration included in the acquired damage data are quantified as explanatory variables, and the degree of damage is quantified as a target variable. As with the first model generation unit 112, the second model generation unit 114 may generate the second model by any machine learning method such as a neural network or regression analysis, a deep learning method, or a statistical method.

[0082] In step S8, the second model generation unit 114 stores the generated second model in the storage unit 12.

[0083] In step S9, the control unit 11 reads out the first model and the second model from the storage unit 22 and transmits them to the estimation device 20 via the communication unit 13.

[0084] In step S10, the model acquisition unit 211 of the estimation device 20 acquires the first model generated by the first model generation unit 112 and the second model generated by the second model generation unit 114 from the learning device 10 by receiving them via the communication unit 23.

[0085] In step S11 of FIG. 7B, the model acquisition unit 211 outputs the acquired first model and second model to the estimation unit 213.

[0086] In step S12, the target data acquisition unit 212 acquires the target data. Any method may be employed to acquire the target data.

[0087] In step S13 , the target data acquisition unit 212 outputs the acquired target data to the estimation unit 213 .

[0088] In step S14, the estimation unit 213 applies the target data acquired by the target data acquisition unit 212 to the first model and the second model acquired by the model acquisition unit 211, and estimates the degree of damage to the target equipment. Specifically, the estimation unit 213 quantifies the information about the target equipment contained in the target data and the environmental conditions of the target equipment, and inputs them into a first model for the equipment corresponding to the target equipment. The estimation unit 213 acquires the degree of deterioration output by the first model, quantifies the degree of deterioration and the disaster conditions of the target equipment contained in the target data, and inputs them into a second model for the target equipment and the disaster to the target equipment. The estimation unit 213 acquires the degree of damage calculated by the second model. In this way, the estimation unit 213 estimates the degree of damage to the target equipment.

[0089] In step S15, the estimation unit 213 displays the estimated extent of damage via the output unit 25. Any method may be used for the display. For example, the output unit 25 may output the estimated extent of damage as audio or video. Thereafter, the operation of the damage prediction system 1 ends.

[0090] As described above, the learning device 10 of this embodiment includes a first model generation unit 112 that generates a first model that calculates the degree of deterioration of the equipment through machine learning using, as training data, deterioration data that includes information about the equipment, the environmental conditions of the equipment, and the degree of deterioration of the equipment, and a second model generation unit 114 that generates a second model that calculates the degree of damage to the equipment through machine learning using, as training data, damage data that includes disaster conditions of a disaster to the equipment, the degree of damage caused by the disaster, and the degree of deterioration calculated by the first model.

[0091] The learning device 10 of this embodiment can easily generate a model that predicts the extent of damage caused by a disaster, reflecting the results of predictions of the degree of deterioration of the equipment. This makes it possible to accurately predict damage caused by a disaster by taking into account the degree of deterioration of the target equipment, thereby improving the technology for predicting damage to equipment caused by a disaster.

[0092] As described above, in the learning device 10 of this embodiment, the facilities include infrastructure facilities including underground pipes or utility poles, and the disasters include at least one of earthquakes, liquefaction, and typhoons.

[0093] The learning device 10 of this embodiment can estimate damage to infrastructure facilities, which are important facilities related to the public interest, caused by earthquakes, liquefaction, and typhoons, which are disasters that are expected to have a high frequency and serious impact, thereby improving the technology for predicting damage to facilities caused by disasters.

[0094] As described above, the estimation device 20 of this embodiment includes a model acquisition unit 211 that acquires a first model that calculates the degree of deterioration of equipment, which is learned by machine learning using, as training data, deterioration data including information about the equipment, environmental conditions of the equipment, and the degree of deterioration of the equipment, and a second model that calculates the degree of damage to the equipment, which is learned by machine learning using, as training data, disaster conditions of a disaster to the equipment, the degree of damage caused by the disaster, and damage data including the degree of deterioration calculated by the first model, and an estimation unit 213 that applies target data including information about the equipment of the target equipment, the environmental conditions of the target equipment, and disaster conditions of the target equipment to the first model and the second model, and estimates the degree of damage to the target equipment.

[0095] According to the estimation device 20 of this embodiment, the degree of deterioration of the target facility can be predicted without inspecting the degree of deterioration of the target facility in advance, thereby reducing inspection costs. Since the degree of damage in the event of a disaster can be accurately estimated by taking into account the predicted degree of deterioration, the technology for predicting damage to facilities due to disasters can be improved.

[0096] Although the present invention has been described based on the drawings and embodiments, it should be noted that those skilled in the art can easily make various modifications and alterations based on the present invention, and therefore, it should be noted that these modifications and alterations are included within the scope of the present invention.

[0097] (Variation) A modified example of the present disclosure will be described.

[0098] Fig. 8 is a diagram showing the configuration of a damage prediction system 1 according to this modification. As shown in Fig. 8, a control unit 21 of an estimation device 20 according to this modification further includes a division unit 214 in addition to a model acquisition unit 211, a target data acquisition unit 212, and an estimation unit 213. The configurations of the units other than the control unit 21 of the estimation device 20 according to this modification are the same as those of the above-described embodiment, and therefore description thereof will be omitted.

[0099] The target data acquisition unit 212 of the estimation device 20 according to this modification acquires target data including information about the target equipment, environmental conditions of the target equipment, and disaster conditions for each of multiple disasters that the target equipment may experience. Any method may be used to acquire the target data. For example, similar to the above-described embodiment, the target data acquisition unit 212 may display a message via the output unit 25 prompting the user to input the environmental conditions of the target equipment and disaster conditions for each of multiple disasters, and acquire values ​​input by the user via the input unit 24 as target data. The target data acquisition unit 212 may communicate with a terminal device held by the user, receive values ​​input by the user to the terminal device from the terminal device, and acquire the values ​​as target data. The target data acquisition unit 212 outputs the acquired target data to the division unit 214.

[0100] An example of target data is shown in Figure 9. Referring to Figure 9, the equipment information, environmental conditions, and earthquake motion disaster conditions for the pipeline as the target equipment are the same as those in the above-mentioned embodiment. In addition, the target data of this modified example further includes a liquefaction disaster condition as another disaster. Referring to Figure 9, the liquefaction history as a liquefaction disaster condition for the pipeline is "Yes," and the liquefaction index is "TT." In the next row, the equipment information, environmental conditions, and earthquake motion disaster conditions for the utility pole as the target equipment are the same as those in the above-mentioned embodiment, and the liquefaction history as a liquefaction disaster condition is "No," and the liquefaction index is "UU." In this way, the target data acquired in this modified example includes disaster conditions for each of multiple disasters.

[0101] The dividing unit 214 divides the target data acquired by the target data acquisition unit 212 by disaster. Figures 10A and 10B show the results of dividing the target data in Figure 9 by the dividing unit 214. Figure 10A shows target data including disaster conditions of seismic motion for pipelines and utility poles, and Figure 10B shows target data including disaster conditions of liquefaction for pipelines and utility poles. The dividing unit 214 outputs the divided target data to the estimation unit 213.

[0102] The estimation unit 213 applies each of the target data divided by the division unit 214 to the first model and the second model acquired by the model acquisition unit 211 in the same manner as in the above-mentioned embodiment, and estimates the degree of damage to each of the target facilities.

[0103] The estimation unit 213 displays the degree of damage estimated for each of the divided target data via the output unit 25. Any method may be used for the display. For example, the estimation unit 213 may display the degree of damage estimated for each disaster separately for each disaster, or may display the degree of damage for all disasters together. In this case, the estimation unit 213 may integrate the estimated degree of damage by summing up category values ​​indicating the degree of damage. For example, the estimation unit 213 may display the location of the target equipment on a map via the output unit 25, and may display the estimated degree of damage for each disaster in a list above the location of the equipment.

[0104] As described above, in this modification, the target data includes disaster conditions for each of a plurality of disasters. The estimation device 20 according to this modification further includes a division unit 214 that divides the target data for each disaster. The estimation unit 213 applies the target data divided by the division unit 214 to the first model and the second model, and estimates the degree of damage to the target equipment for each of the plurality of disasters.

[0105] According to the estimation device 20 of this modification, it is possible to estimate the degree of damage caused by multiple disasters for one piece of target equipment at once using target data in which disaster conditions for multiple disasters have been input. Since the estimated results of the degree of damage for various assumed disasters can be viewed in a list, it becomes easier for users to plan maintenance for the target equipment. Therefore, it is possible to improve the technology for predicting damage to equipment caused by disasters. [Explanation of symbols]

[0106] 1. Damage prediction system 10 Learning Device 11 Control section 12 Storage section 13 Communications Department 14 Input section 15 Output section 20 Estimation device 20 21 Control section 22 Memory section 23 Communications Department 24 Input section 25 Output section 30 Network 111 Deterioration data acquisition unit 112 First model generation unit 113 Damage Data Acquisition Department 114 Second Model Generation Unit 211 Model Acquisition Department 212 Target Data Acquisition Unit 213 Estimation Department 214 Division

Claims

1. a first model generation unit that generates a first model that calculates the degree of deterioration of the facility by machine learning using deterioration data that includes information about the facility, environmental conditions of the facility, and the degree of deterioration of the facility as training data; a second model generation unit that generates a second model that calculates the degree of damage to the facility by machine learning using damage data that includes disaster conditions of a disaster to the facility, the degree of damage caused by the disaster, and the degree of deterioration calculated by the first model as training data; and A learning device comprising:

2. The learning device according to claim 1 , wherein the facility includes infrastructure facilities including underground pipes or utility poles, and the disaster includes at least one of an earthquake, liquefaction, and a typhoon.

3. a model acquisition unit that acquires a first model that calculates the degree of deterioration of the facility, the first model being trained by machine learning using as training data information about the facility, environmental conditions of the facility, and deterioration data including the degree of deterioration of the facility; and a second model that calculates the degree of damage to the facility, the second model being trained by machine learning using as training data damage data including disaster conditions of a disaster to the facility, the degree of damage caused by the disaster, and the degree of deterioration calculated by the first model; an estimation unit that applies information about the target equipment and environmental conditions of the target equipment to the first model, acquires the degree of deterioration of the target equipment output by the first model, and applies the acquired degree of deterioration of the target equipment and disaster conditions of the target equipment to the second model to estimate the degree of damage to the target equipment; An estimation device comprising:

4. The method further comprises a division unit that divides target data including disaster conditions for each of a plurality of disasters as disaster conditions for the target equipment, The estimation device according to claim 3 , wherein the estimation unit applies the target data divided by the division unit to the second model, and estimates the degree of damage to the target equipment for each of the plurality of disasters.

5. A learning method executed by a learning device, a first model generation step of generating a first model that calculates the degree of deterioration of the facility by machine learning using deterioration data, including information about the facility, environmental conditions of the facility, and the degree of deterioration of the facility, as training data; a second model generation step of generating a second model that calculates the degree of damage to the facility by machine learning using damage data as training data, the damage data including the disaster conditions of the disaster to the facility, the degree of damage caused by the disaster, and the degree of deterioration calculated by the first model; Learning methods including.

6. An estimation method executed by an estimation device, an acquisition step of acquiring a first model for calculating the degree of deterioration of the facility, the first model being trained by machine learning using as training data information about the facility, environmental conditions of the facility, and deterioration data including the degree of deterioration of the facility, and a second model for calculating the degree of damage to the facility, the second model being trained by machine learning using as training data damage data including disaster conditions of a disaster to the facility, the degree of damage caused by the disaster, and the degree of deterioration calculated by the first model; an estimation step of applying information about the equipment of the target equipment and environmental conditions of the target equipment to the first model, acquiring the degree of deterioration of the target equipment output by the first model, and applying the acquired degree of deterioration of the target equipment and disaster conditions of the target equipment to the second model, and estimating the degree of damage to the target equipment; An estimation method, including:

7. A program for causing a computer to function as the learning device according to claim 1 or 2.

Images