JP7444811B2 - Building damage estimation method, building damage estimation system, building damage learning method, building damage learning system and program - Google Patents

Description

The present disclosure relates to a building disaster degree estimation method, a building disaster degree estimation system, a building disaster degree learning method, a building disaster degree learning system, and a program.

Patent Document 1 discloses a building damage estimation system and method. This building damage estimation system is composed of a building damage estimation device, an earthquake sensor installed on the foundation ground surface of the building, and an earthquake sensor installed at representative points within the building. The building damage estimation device includes a virtual building model derivation unit that derives a virtual building model that mathematically represents the movement of a building due to an earthquake, based on known building information about the building, and a virtual building model derivation unit that derives a virtual building model that formulates the movement of the building due to an earthquake, and a ground motion acceleration measured by an earthquake sensor during an earthquake. an earthquake response analysis section that inputs information into a virtual building model and performs earthquake response analysis of the virtual building model; It includes a motion information calculation section and a disaster degree estimation section that estimates the degree of damage to a building using earthquake information and movement information after an earthquake.

In the building damage estimation system and method described in Patent Document 1, there is no need to install acceleration sensors on all layers of the building, and earthquake sensors are installed only on the foundation ground surface of the building and at representative points within the building. Bye. Therefore, the degree of damage to a building after an earthquake can be estimated at a lower cost than before.

JP 2016-197013 Publication

However, in the building damage estimation system and method described in Patent Document 1, it is necessary to install earthquake sensors at least at representative points within the building, so compared to the case where acceleration sensors are installed on all layers of the building. Although it would be cheaper to do so, there was still the problem that it was difficult to reduce costs.

Furthermore, the seismic response analysis performed by the seismic response analysis department is specifically a time history response analysis that takes into account the nonlinearity of building rigidity, so there is a problem in that the analysis takes time.

The present disclosure was invented in view of the above-mentioned conventional problems, and includes a method and system for estimating the degree of damage to a building, which facilitates cost reduction and the time required for estimating the degree of disaster. , our objective is to provide a method for learning the degree of damage to buildings, a system and program for learning the degree of damage to buildings.

A disaster degree estimation method for a building according to one aspect of the present disclosure includes an acquisition step and an estimation step. The acquisition step is a step of acquiring input information including maximum acceleration of the earthquake motion and information on spectrum values in response spectrum analysis of the time-series waveform of the earthquake motion. The estimation step is a step of estimating the output information based on the input information using a learned model built in advance. The learned model uses the input information as input data, and is constructed by machine learning using output information regarding the damage level of the building due to the earthquake motion as training data.

A disaster degree estimation system for a building according to one aspect of the present disclosure includes an acquisition unit and an estimation unit. The acquisition unit acquires input information including information on a maximum acceleration of an earthquake motion and a spectrum value in a response spectrum analysis of a time-series waveform of the earthquake motion. The estimation unit estimates the output information based on the input information using a learned model built in advance. The learned model uses the input information as input data, and is constructed by machine learning using output information regarding the damage level of the building due to the earthquake motion as training data.

A building disaster degree learning method according to one aspect of the present disclosure includes an acquisition step and a learning step. The acquisition step is a step of acquiring input information including maximum acceleration of the earthquake motion and information on spectrum values in response spectrum analysis of the time-series waveform of the earthquake motion. The learning step is a step of constructing a trained model using the input information as input data, and using machine learning to construct the trained model using output information regarding the damage level of the building due to the earthquake motion as teacher data.

A building disaster degree learning system according to one aspect of the present disclosure includes an acquisition unit and a learning unit. The acquisition unit acquires input information including information on a maximum acceleration of an earthquake motion and a spectrum value in a response spectrum analysis of a time-series waveform of the earthquake motion. The learning unit is a learned model that uses the input information as input data, and constructs the learned model by machine learning using output information regarding the damage level of the building due to the earthquake motion as teacher data.

A program according to an aspect of the present disclosure causes one or more processors to execute the method for estimating the degree of disaster for a building or the method for learning the degree of disaster for a building.

According to the building damage estimation method, building damage estimation system, building damage learning method, building damage learning system, and program of one aspect of the present disclosure, costs can be easily reduced and disaster damage estimation It is easy to shorten the time required.

FIG. 1 is a block diagram of a building damage estimation system and a disaster damage learning system according to an embodiment of the present disclosure. FIG. 2 is a configuration diagram of a neural network in the trained model of the embodiment described above. FIG. 3 is a configuration diagram of an analytical model of a building in the embodiment same as above. FIG. 4 is a flowchart of a method for learning the degree of damage to a building in the embodiment described above. FIG. 5 is a diagram showing the relationship between the average value of the absolute value of the weight W1 and the node in the neural network of the analytical model of method A according to the above embodiment. FIG. 6 is a diagram showing the relationship between the average absolute value of the weight W1 and the node in the neural network of the analytical model of method B according to the embodiment. FIG. 7 is a flowchart of a method for estimating the degree of damage to a building in the embodiment described above. FIG. 8A is a relationship diagram between the estimated value of the degree of disaster and the correct value using the training data of method A according to the above embodiment. FIG. 8B is a relationship diagram between the estimated value of the degree of disaster and the correct value using the test data of construction method A according to the above embodiment. FIG. 9A is a relationship diagram between the estimated value of the degree of disaster and the correct value using the training data of method B of the embodiment. FIG. 9B is a diagram showing the relationship between the estimated value of the degree of disaster and the correct value using the test data of method B of the embodiment. FIG. 10A is a relationship diagram between the estimated value of the degree of disaster and the correct value using the test data of method B of the embodiment same as above. FIG. 10B is a relationship diagram between the estimated value of the degree of disaster and the correct value using the test data of method B of Example 1. FIG. 11A is a relationship diagram between the estimated value of the degree of disaster and the correct value using the test data of construction method B of the embodiment same as above. FIG. 11B is a relationship diagram between the estimated value of the degree of disaster and the correct value using the test data of the B construction method of Example 2.

(overview)
The present disclosure relates to a building damage estimation method, a building damage estimation system, a building damage learning method, a building damage learning system, and a program. The primary purpose is to estimate the degree of A second objective of the present disclosure is to quickly estimate the degree of damage to a building at low cost.

Human casualties and damage to buildings caused by earthquakes have become bigger problems than before, and various efforts have recently been made to reduce these damages. Examples of human damage include injuries caused by people falling down or colliding with falling objects, and injuries caused by damage to buildings during and immediately after an earthquake. Furthermore, even if no human casualties occur during an earthquake, if buildings are damaged, human casualties may occur due to further damage or collapse of the damaged buildings. In any case, building damage is not only damage to the building, but also causes human damage, and in order to minimize human damage and damage to the building, it is necessary to determine the degree of damage to the building (disaster damage level). ) needs to be quickly understood.

For this reason, some companies, including housing manufacturers, install earthquake sensors in buildings to directly measure the vibrations of the building due to earthquake motion, thereby ascertaining the degree of damage to the building. However, although this method makes it easy to accurately assess the degree of damage to a building, it requires installing earthquake sensors in many residential units and creating an environment for transmitting measurement data from the earthquake sensors, which is costly. . Furthermore, the accuracy of estimating the degree of damage is not good if only the seismic intensity is determined from the seismic intensity distribution published by public institutions.

In the building damage estimation method, building damage estimation system, building damage learning method, building damage learning system and program disclosed herein, the input information is obtained from a large number of earthquake sensors installed in actual dwelling units. The information on the time-series waveform of the seismic motion is not required, and instead, the input information is the maximum acceleration of the seismic motion and information on the spectrum value in the response spectrum analysis of the time-series waveform of the seismic motion. This eliminates the need to install any or almost no seismic sensors in buildings, making it easier to reduce costs, and making it possible to estimate the degree of damage to buildings without performing time history response analysis that takes into account the nonlinearity of building stiffness. , it becomes easier to shorten the time required to estimate the degree of damage to a building.

(detail)
Hereinafter, a building damage estimation method, a building damage estimation system, a building damage learning method, a building damage learning system, and a program according to an embodiment of the present disclosure will be described in detail with reference to FIG. 11B. do. The building disaster degree estimation method of this embodiment can be executed by the building disaster degree estimation system 1 shown in FIG. The disaster damage estimation system 1 can be connected to the seismic motion waveform server 9 via the communication network 8 .

(1) Disaster damage estimation system As described above, the disaster damage estimation system 1 is a system for executing the method for estimating the disaster damage of a building according to the present embodiment. The disaster damage estimation system 1 includes a communication section 11, a storage section 12, and a processing section 13. The disaster damage estimation system 1 can be realized by one or more servers.

The communication unit 11 is a communication interface. The communication unit 11 is connectable to the communication network 8 and has a function of communicating through the communication network 8. The communication unit 11 includes, for example, a transmitter and a receiver. The communication unit 11 complies with a predetermined communication protocol. The predetermined communication protocol may be selected from a variety of well-known wired and wireless communication standards.

The storage unit 12 is used to store information used by the processing unit 13. The storage unit 12 includes one or more storage devices. The storage device is, for example, a RAM (Random Access Memory) or an EEPROM (Electrically Erasable Programmable Read Only Memory).

The processing unit 13 is a control circuit that controls the operation of the disaster damage estimation system 1. The processing unit 13 can be realized, for example, by a computer system including one or more processors (microprocessors) and one or more memories. That is, one or more processors function as the processing unit 13 by executing one or more (computer) programs (applications) stored in one or more memories. Although the program is pre-recorded in the memory of the processing unit 13 here, it may also be provided via a telecommunications line such as the Internet or by being recorded on a non-temporary recording medium such as a memory card.

The processing unit 13 includes a preprocessing unit 131, an acquisition unit 132, an estimation unit 133, and a presentation unit 134. The preprocessing unit 131, the acquisition unit 132, the estimation unit 133, and the presentation unit 134 do not represent actual configurations, but represent functions realized by the processing unit 13.

(1-1) Preprocessing unit The preprocessing unit 131 acquires input information necessary for the disaster degree estimation method executed by the disaster degree estimation system 1. The process of obtaining input information is executed based on the time series data of or the time series data of acceleration (these are collectively referred to as the time series waveform of seismic motion).

The time-series waveform of the seismic motion can be obtained from the seismic motion waveform server 9 via the communication network 8 . The seismic motion waveform server 9 may be, for example, a server on the Internet operated by the National Research Institute for Earth Science and Disaster Resilience. It is possible to obtain data such as "net)". The Strong Motion Observation Network (K-NET, KiK-net) allows you to obtain data from many observation points across the country.

The preprocessing unit 131 obtains input information necessary for the disaster severity estimation method executed by the disaster severity estimation system 1 from the time-series waveform of the earthquake motion acquired from the earthquake motion waveform server 9 . The input information includes information on the maximum acceleration of the seismic motion, the predetermined acceleration duration, and the spectrum value.

The maximum acceleration of seismic motion is the so-called peak ground acceleration (PGA), which is a measured value of the maximum acceleration on the earth's surface at the time of seismic motion. PGA can be obtained from the time-series waveform of seismic motion.

The predetermined acceleration duration is the time during which the acceleration of the seismic motion exceeds the predetermined acceleration obtained by multiplying PGA by a positive value less than 1, which occupies the predetermined duration of the seismic motion. The positive value by which PGA is multiplied is 0.8 in this embodiment. The predetermined acceleration duration can be determined from the time-series waveform of the seismic motion.

The spectral value is obtained by a response spectrum analysis (in this embodiment, an acceleration response spectrum analysis as shown below) of a time-series waveform of acceleration of the seismic motion. Specifically, the maximum response acceleration is obtained by inputting the target seismic motion waveform into a linear one-mass system shear model with a constant damping constant (5% in this embodiment) and performing time history response analysis. By repeating this analysis by changing the natural period of the one-mass shear model, an acceleration response spectrum showing the relationship between the natural period and the maximum response acceleration is expressed.

(1-2) Acquisition Unit The acquisition unit 132 acquires input information required in the disaster severity estimation method executed by the disaster severity estimation system 1. The input information is the above-mentioned PGA, the predetermined acceleration duration time, and the spectrum value in the acceleration response spectrum analysis of the time-series waveform of the seismic motion.

(1-3) Estimating Unit The estimating unit 133 estimates output information regarding the degree of damage caused by earthquake motion of the building based on the input information acquired by the acquiring unit 132 using a trained model built in advance. In this embodiment, the estimation unit 133 outputs information on the maximum amount of deformation of the building as output information.

A trained model is constructed by machine learning using input information as input data and output information as training data. In this embodiment, the learned model uses a neural network. The trained model is constructed by the building damage level learning system 2. The disaster severity learning system 2 will be described later.

(1-4) Presentation Unit The presentation unit 134 presents the output information estimated by the estimation unit 133. In this embodiment, by sending the output information estimated by the estimation unit 133 to the terminal device 7 through the communication network 8, the output information can be presented at the input/output unit 71 of the terminal device 7.

(2-1) Communication Network The communication network 8 may include the Internet. The communication network 8 may include not only a network based on a single communication protocol but also a plurality of networks based on different communication protocols. The communication protocol may be selected from a variety of well-known wired and wireless communication standards. Communication network 8 may include data communication equipment such as repeater hubs, switching hubs, bridges, gateways, routers, and the like.

(2-2) Terminal Device The terminal device 7 can be used to display information from the building disaster degree estimation system 1 and to transmit safety information to the disaster degree estimation system 1, for example. The terminal device 7 includes an input/output section 71, a communication section 72, and a processing section 73. The terminal device 7 can be realized by a desktop computer, a laptop computer, or a mobile terminal (smartphone, tablet terminal, wearable terminal, etc.).

The input/output unit 71 includes an input device for operating the terminal device 7. Input devices may include, for example, a keyboard, mouse, touch pad, and the like. The input/output unit 71 also includes an image display device for displaying information. The image display device may include a thin display device such as a liquid crystal display or an organic EL display.

The communication unit 72 is a communication interface and is similar to the communication unit 11 described above, so a detailed description thereof will be omitted.

The processing unit 73 can be realized, for example, by a computer system including one or more processors (microprocessors) and one or more memories.

(3) Disaster damage learning system The disaster damage learning system 2 includes a communication section 21, a storage section 22, and a processing section 23. The disaster severity learning system 2 can be realized by one or more servers.

The communication unit 21 is a communication interface and is similar to the communication unit 11 described above, so a detailed description thereof will be omitted. The storage unit 22 is used to store information used by the processing unit 23, and is similar to the storage unit 12 described above, so a detailed explanation will be omitted. The processing unit 23 is a control circuit that controls the operation of the disaster severity learning system 2, and has the same hardware configuration as the storage unit 12 described above.

The processing unit 23 includes a preprocessing unit 231, an acquisition unit 232, a learning unit 233, and a presentation unit 234. The preprocessing unit 231, the acquisition unit 232, the learning unit 233, and the presentation unit 234 do not represent actual configurations, but represent functions realized by the processing unit 23.

(3-1) Pre-processing unit The pre-processing unit 231 uses the time-series waveform of the seismic motion to acquire input information that serves as teacher data required in the disaster severity learning method executed by the disaster severity learning system 2. Executes a process that requests input information. The time-series waveform of the seismic motion, which is the basis of the training data, can be obtained from the seismic motion waveform server 9 via the communication network 8, similarly to the disaster degree estimation system 1.

The preprocessing unit 231 obtains input information necessary for the disaster severity learning method executed by the disaster severity learning system 2 from the time-series waveform of the earthquake motion acquired from the earthquake motion waveform server 9. The input information includes information on the PGA, predetermined acceleration duration, and spectrum value similar to the disaster damage estimation system 1.

(3-2) Acquisition Unit The acquisition unit 232 acquires input information necessary for the disaster severity learning method executed by the disaster severity learning system 2. The input information is the above-mentioned PGA, predetermined acceleration duration time, and spectrum value information.

(3-3) Learning Unit The learning unit 233 uses machine learning to construct a trained model that uses input information as input data, and uses output information regarding the degree of damage caused by earthquake motion of buildings as training data. The trained model uses the neural network 4 shown in FIG. 2. In this embodiment, as input data, measured seismic intensity data is input to the node 41000 of the input layer 41, PGA data is input to the node 41001, and data of the predetermined acceleration duration is input to the node 41002 (a positive value is 0.8 ). Furthermore, the acceleration response spectrum values of earthquake motion are input to nodes 41003 to 41102 of the input layer 41. The spectrum value is the maximum value of the period band 0.0 (sec) or more and less than 0.1 (sec) for node 41003, and the maximum value of the period band 0.1 (sec) or more and less than 0.2 (sec) for node 41004. The maximum value of . The input layer 41 has 103 nodes, node 41000 to node 41102.

The neural network 4 has two intermediate layers, an intermediate layer 42 and an intermediate layer 43. The middle layer 42 has 64 nodes, node 4200 to node 4263. The intermediate layer 43 has 64 nodes, node 4300 to node 4363. Output layer 44 has one node 440. The output information output from the node 440 is information on the maximum amount of deformation of the building. The maximum amount of deformation of the building as training data uses data obtained by time-history response analysis of the time-series waveform of seismic motion acquired by the preprocessing unit 231 for the analytical model of the building. The analytical model of the building will be described later.

(3-4) Presentation Unit As shown in FIG. 1, the presentation unit 234 presents various output results from the processing unit 23. In this embodiment, by sending various output results from the processing section 23 to the terminal device 7 through the communication network 8, the output information can be presented at the input/output section 71 of the terminal device 7.

(4) Analytical model of building As shown in FIG. 3, in this embodiment, the analytical model of building 3 is an n-mass point system shear model having nonlinear shear stiffness and linear viscous damping. It is assumed that the disturbance input to the building 3 is only the vibration of the ground 30, and there is no influence of the rocking spring of the ground 30.

As shown in FIG. 1, the mass of the i (i=1 to n) layer (story) of the building 3 is expressed as m _i , the layer stiffness of the i layer is k _i , the layer damping coefficient of the i layer is c _i , and the time When the absolute acceleration of the i layer at time t is expressed as u'' _i (t), the layer shear force Q _i (t) at time t in the i layer is expressed by the formula shown in [Equation 1] below based on the equation of motion. In equation (1), u' _i (t) is the absolute velocity at time t, and u _i (t) is the absolute displacement at time t.

Interlayer displacement D _i (t)=u _i (t)−u _i −1(t) of i layer at time t, interlayer velocity D _i ′(t)=u _i ′(t)−u _i −1′( t), interlayer acceleration D _i ″(t)=u _i ″(t)−u _i −1″(t), then equation (1) is expressed by the equation shown in [Equation 2] below.

The absolute acceleration u ₁ ″(t), absolute velocity u ₁ ′(t), and absolute displacement u ₁ (t) of the 1st (i=1) layer are obtained from the time-series waveform of the earthquake motion. Through time history response analysis of the analytical model, all absolute accelerations u _i ″(t), absolute velocities u _i ′(t), and absolute displacements u _i (t), all interstory displacements D _i (t), and interstory speeds D _i ′(t), interlayer acceleration D _i ″, and the maximum deformation amount of each layer.

In this embodiment, analytical models of the building 3 are set for two patterns of construction methods A and B, which are so-called different construction methods.

(5) Operation (5-1) Building disaster level learning method As shown in FIG. 4, in the disaster level learning system 2, a preprocessing step is first performed (S1). In step S1, the preprocessing unit 231 executes processing to obtain input information. Specifically, in step S1, measured seismic intensity data is acquired. Furthermore, in the preprocessing step, a time-series waveform of the seismic motion is acquired, and PGA data and data of a predetermined acceleration duration are extracted. Furthermore, in step S1, the time-series waveform of the seismic motion is subjected to acceleration response spectrum analysis to obtain spectrum values in each period band.

In this embodiment, the time-series waveforms of 1250 earthquakes that have occurred in Japan in recent years are acquired from the seismic motion waveform server 9 operated by the National Research and Development Agency, National Research Institute for Earth Science and Disaster Prevention, to learn the degree of disaster.

Next, an acquisition step is executed (S2). In step S2, the acquisition unit 232 acquires input information.

Next, a learning step is executed (S3). In step S3, the learning unit 233 constructs a trained model. In this embodiment, among the data based on the time-series waveforms of 1250 earthquake motions, data based on the time-series waveforms of 937 earthquake motions, which is 75%, is used as training data, and the time-series waveforms of 313 earthquake motions, which are 25%, are used as training data. The data based on this was used as test data to confirm generalization ability.

Next, a presentation step is executed (S4). In step S4, the presenting unit 234 presents the weights W1 to W3 of the neural network 4.

FIG. 5 shows an average value of the weight W1 in the trained model of the neural network 4 of the analytical model of method A--a node relationship diagram. From this result, node 41000 (measured seismic intensity), node 41001 (PGA), node 41002 (predetermined acceleration duration), node 41003 (spectral value in period band 0.0 (sec) or more and less than 0.1 (sec)) ~The value of weight W1 multiplied by the value of node 41007 (spectral value of period band 0.4 (sec) or more and less than 0.5 (sec)) is greater than the value of weight W1 multiplied by the value of nodes 41008 to 41102. It can be seen that the values of nodes 41000 to 41107 have a large influence on the output information.

Similarly, FIG. 6 shows an average value of the weight W1 in the trained model of the neural network 4 of the analytical model of method B--a node relationship diagram. From this result, it can be seen that in Method B, the values of nodes 41000 to 41107 have a large influence on the output information, which is larger than the value of weight W1 multiplied by the values of nodes 41008 to 41102, as in FIG. 5. .

(5-2) Method for estimating disaster degree of building As shown in FIG. 7, in the disaster degree estimation system 1, first, a preprocessing step is executed (S11). In step S11, the preprocessing unit 131 executes processing to obtain input information. Specifically, in step S11, measured seismic intensity data is acquired. Furthermore, in step S11, a time-series waveform of the seismic motion is acquired, and PGA data and data of a predetermined acceleration duration are extracted. Further, in step S11, the time-series waveform of the seismic motion is subjected to acceleration response spectrum analysis, and the spectral value in each periodic band and the maximum value of the spectral value in each periodic band are obtained.

Next, an acquisition step is executed (S12). In step S12, the acquisition unit 132 acquires input information.

Next, an estimation step is executed (S13). In step S13, the estimation unit 133 estimates the degree of disaster using the trained model. In this embodiment, estimation was made for training data based on time-series waveforms of 937 earthquake motions and test data based on time-series waveforms of 313 earthquake motions.

Next, a presentation step is executed (S14). In step S14, the presentation unit 234 presents the maximum amount of deformation of the building as output information.

Figure 8A shows a relationship diagram between the estimated value of the degree of disaster using the training data of method A - the correct value (teacher data), and Figure 8B shows the estimated value of the degree of disaster using the test data of method A - the correct value. A relationship diagram is shown. Note that the unit is an arbitrary unit (au). From this result, the accuracy of the disaster damage estimation value using the test data is comparable to the accuracy of the disaster damage estimation value using the training data, confirming the generalization ability of the disaster damage estimation method.

Figure 9A shows the relationship diagram between the estimated value of the disaster degree using the training data of method B and the correct value, and Figure 9B shows the relationship diagram between the estimated value of the degree of disaster and the correct value using the test data of method B. show. This result also shows that the accuracy of disaster damage estimation using test data is comparable to the accuracy of disaster damage estimation using training data, and the generalization ability of the disaster damage estimation method is You can check it.

(6) Summary of the present embodiment In the above-described disaster degree estimation system 1, disaster degree estimation method, disaster degree learning system 2, and disaster degree learning method, information on time-series waveforms of earthquake motion is unnecessary as input information. By using a trained model, we only need information on the maximum acceleration of earthquake motion and the spectrum value in the acceleration response spectrum analysis of the time-series waveform of earthquake motion. This eliminates the need to perform time history response analysis as in the past, making it easier to shorten the time required to estimate the degree of disaster. Specifically, when performing time history response analysis, the time required for the analysis would be enormous, but the time required for estimation in the estimator 133 is now only a few minutes. Note that although some time is required for the processing in the preprocessing unit 131 that performs acceleration response spectrum analysis, the time required for the disaster degree estimation method is 10 minutes in total, compared to the time history response analysis. 1 or less, and it is possible to significantly shorten the time compared to the conventional method.

Furthermore, there is no need to install any or almost no earthquake sensors in the building, resulting in cost reduction.

(6-1) Example 1
FIG. 10A shows a relationship diagram (same as FIG. 9B) of the estimated degree of disaster using the test data of Method B of the above embodiment (same as FIG. 9B), and FIG. 10B shows the test data of Method B of Example 1. A relationship diagram between the estimated value of the degree of disaster used and the correct value is shown.

Example 1 is an example in which data on the predetermined acceleration duration time is not included as input data. The degree of agreement between the estimated value of Example 1 shown in FIG. 10B and the correct value is lower than the degree of agreement between the estimated value of this embodiment and the correct value shown in FIG. 10A, and the degree of agreement between the estimated value of Example 1 shown in FIG. The degree of accuracy is inferior to the degree of accuracy of the estimated value of the degree of disaster according to the present embodiment shown in FIG. 10A. From this, the effectiveness of including data on the predetermined acceleration duration time as input data can be confirmed.

(6-2) Example 2
FIG. 11A shows a relationship diagram (same as FIG. 9B) of the estimated degree of disaster using the test data of Method B of the above embodiment and the correct value, and FIG. 11B shows the test data of Method B of Example 2. A relationship diagram between the estimated value of the degree of disaster used and the correct value is shown.

Example 2 is an example in which the input data does not include data of the maximum value of the spectrum value in a period band of 3.0 (sec) or more and less than 10.0 (sec). That is, in this embodiment, the input layer 41 has 103 nodes from node 41000 to node 41102, whereas in Example 2, the input layer 41 has only 33 nodes from node 41000 to node 41032.

The degree of agreement between the estimated value of Example 2 shown in FIG. 11B and the correct value is lower than the degree of agreement between the estimated value of this embodiment and the correct value shown in FIG. 11A, and the degree of agreement between the estimated value of Example 2 shown in FIG. The degree of accuracy is inferior to the degree of accuracy of the estimated value of the degree of disaster according to the present embodiment shown in FIG. 11A. From this, it can be confirmed that it is effective to set the periodic band of the input data spectrum value to a range of 0.0 (sec) or more and less than 10.0 (sec).

(7) Modifications Embodiments of the present disclosure are not limited to the above embodiments. The embodiments described above can be modified in various ways depending on the design, etc., as long as the objects of the present disclosure can be achieved. Modifications of the above embodiment are listed below. The modified examples described below can be applied in combination as appropriate.

The disaster damage estimation system 1 and the disaster damage learning system 2 in the present disclosure include, for example, a computer system. A computer system mainly consists of a processor and a memory as hardware. When a processor executes a program recorded in the memory of the computer system, the functions of the building damage estimation method and disaster damage learning method in the present disclosure are realized. The program may be pre-recorded in the memory of the computer system, may be provided through a telecommunications line, or may be recorded on a non-transitory storage medium readable by the computer system, such as a memory card, optical disc, hard disk drive, etc. may be provided. A processor in a computer system is comprised of one or more electronic circuits including semiconductor integrated circuits (ICs) or large scale integrated circuits (LSIs). The integrated circuits such as IC or LSI referred to herein have different names depending on the degree of integration, and include integrated circuits called system LSI, VLSI (Very Large Scale Integration), or ULSI (Ultra Large Scale Integration). Furthermore, FPGAs (Field-Programmable Gate Arrays), which are programmed after the LSI is manufactured, or logic devices that can reconfigure the connections inside the LSI or reconfigure the circuit sections inside the LSI, may also be used as processors. I can do it. The plurality of electronic circuits may be integrated into one chip, or may be provided in a distributed manner over a plurality of chips. A plurality of chips may be integrated into one device, or may be distributed and provided in a plurality of devices. The computer system herein includes a microcontroller having one or more processors and one or more memories. Therefore, the microcontroller is also composed of one or more electronic circuits including semiconductor integrated circuits or large-scale integrated circuits.

Further, a plurality of functions in the disaster damage estimation system 1 and the disaster damage learning system 2 may be integrated into one housing. Furthermore, at least some functions of the disaster damage estimation system 1 and the disaster damage learning system 2, for example, some functions of the processing unit 23, may be realized by a so-called cloud (cloud computing) or the like.

As the time-series waveform of seismic motion, the time-series waveform of seismic motion published by a public institution is used, but the time-series waveform of seismic motion may be obtained by measuring it by yourself.

The time-series waveform of seismic motion does not necessarily need to be acquired via the communication network 8. For example, the disaster degree estimation system 1 may directly acquire the time-series waveform of the earthquake motion from the earthquake motion waveform server 9 without going through the communication network 8 . Moreover, the disaster degree estimation system 1 may be directly equipped with a seismometer.

The positive value by which PGA is multiplied does not have to be 0.8, and can be set as appropriate, such as 0.9, 0.7, 0.6, 0.5, etc.

The output information from the estimation unit 133 is not limited to information on the maximum amount of deformation of the building.

As a machine learning algorithm, an algorithm other than one using a neural network may be used, such as a supervised learning algorithm such as a support vector machine.

The analytical model of the building 3 is not limited to an n-mass point system shear model having nonlinear shear stiffness and linear viscous damping.

It is preferable that the input information used for estimation in the estimation unit 133 includes information on measured seismic intensity. This makes it easier to improve the accuracy of estimating the degree of damage to a building.

The output information estimated by the estimation unit 133 preferably includes information on the maximum deformation amount for a plurality of layers. This makes it easier to improve the accuracy of estimating the degree of damage to a building.

It is preferable that the output information estimated by the estimation unit 133 includes information on the maximum amount of interlayer deformation. This makes it easier to improve the accuracy of estimating the degree of damage to a building.

The output information estimated by the estimation unit 133 preferably includes information on the amount of cumulative plastic deformation of the building. This makes it easier to improve the accuracy of estimating the degree of damage to a building.

The output information estimated by the estimation unit 133 preferably includes information on the disaster rank. This makes it easier to improve the accuracy of estimating the degree of damage to a building.

In addition to acceleration response spectrum analysis, response spectrum analysis includes velocity response spectrum analysis, displacement response spectrum analysis, etc. Any of these can be adopted as response spectrum analysis, and response spectrum analysis is not particularly limited.

(8) Aspects As is clear from the above embodiments and modifications, the present disclosure includes the following aspects.

The first aspect is a method for estimating the degree of damage to a building, and includes an acquisition step and an estimation step. The acquisition step is a step of acquiring input information including maximum acceleration of earthquake motion and information on spectrum values in response spectrum analysis of time-series waveforms of earthquake motion. The estimation step is a step of estimating output information based on input information using a learned model built in advance. The learned model uses input information as input data, and is constructed by machine learning using output information regarding the degree of damage caused by earthquake motion of a building as training data. According to this aspect, information on the time-series waveform of the earthquake motion is not required as input information, and information on the maximum acceleration of the earthquake motion and the spectrum value in the response spectrum analysis of the time-series waveform of the earthquake motion is required by using the trained model. Only. Therefore, there is no need to perform time history response analysis as in the past, and the time required for estimating the degree of disaster can be easily shortened. Furthermore, there is no need to install any or almost no earthquake sensors in the building, making it easier to reduce costs.

The second aspect is a method for estimating the degree of damage to a building based on the first aspect. In the second aspect, the input information includes information on the predetermined acceleration duration. The predetermined acceleration duration is the time during which the acceleration of the seismic motion exceeds the predetermined acceleration obtained by multiplying the maximum acceleration of the seismic motion by a positive value less than 1, which occupies the predetermined duration of the seismic motion. According to this aspect, it becomes easier to improve the accuracy of estimating the degree of damage to a building.

The third aspect is a method for estimating the degree of damage to a building based on the first or second aspect. In the third aspect, the input information includes information on measured seismic intensity. According to this aspect, it becomes easier to improve the accuracy of estimating the degree of damage to a building.

A fourth aspect is a method for estimating the degree of damage to a building based on any one of the first to third aspects. In the fourth aspect, the output information includes information on the maximum amount of deformation of the building. According to this aspect, it becomes easier to improve the accuracy of estimating the degree of damage to a building.

A fifth aspect is a method for estimating the degree of damage to a building based on the fourth aspect. In a fifth aspect, the building comprises multiple tiers. The output information includes information on maximum deformation amounts for multiple layers. According to this aspect, it becomes easier to improve the accuracy of estimating the degree of damage to a building.

A sixth aspect is a method for estimating the degree of damage to a building based on any one of the first to fifth aspects. In a sixth aspect, the building comprises multiple tiers. The output information includes information on the maximum amount of interlayer deformation. According to this aspect, it becomes easier to improve the accuracy of estimating the degree of damage to a building.

A seventh aspect is a method for estimating the degree of damage to a building based on any one of the first to sixth aspects. In the seventh aspect, the output information includes information on the amount of cumulative plastic deformation of the building. According to this aspect, it becomes easier to improve the accuracy of estimating the degree of damage to a building.

An eighth aspect is a method for estimating the degree of damage to a building based on any one of the first to seventh aspects. In the eighth aspect, the output information includes information on disaster rank. According to this aspect, it becomes easier to improve the accuracy of estimating the degree of damage to a building.

A ninth aspect is a method for estimating the degree of damage to a building based on any one of the first to eighth aspects. In the ninth aspect, the learned model uses the neural network 4. According to this aspect, it becomes easier to improve the accuracy of estimating the degree of damage to a building.

A tenth aspect is a building damage degree estimation system 1, which includes an acquisition section 132 and an estimation section 133. The acquisition unit 132 acquires input information including information on maximum acceleration of earthquake motion and spectrum values in response spectrum analysis of time-series waveform of earthquake motion. The estimation unit 133 estimates output information based on input information using a learned model built in advance. The learned model uses input information as input data, and is constructed by machine learning using output information regarding the degree of damage caused by earthquake motion of a building as training data. According to this aspect, information on the time-series waveform of the earthquake motion is not required as input information, and information on the maximum acceleration of the earthquake motion and the spectrum value in the response spectrum analysis of the time-series waveform of the earthquake motion is required by using the trained model. Only. Therefore, there is no need to perform time history response analysis as in the past, and the time required for estimating the degree of disaster can be easily shortened. Furthermore, there is no need to install any or almost no earthquake sensors in the building, making it easier to reduce costs.

An eleventh aspect is a method for learning the degree of damage to a building, which includes an acquisition step and a learning step. The acquisition step is a step of acquiring input information including maximum acceleration of earthquake motion and information on spectrum values in response spectrum analysis of time-series waveforms of earthquake motion. The learning step is a step of constructing, by machine learning, a trained model that uses input information as input data and uses output information regarding the degree of damage caused by earthquake motion of buildings as training data. According to this aspect, information on the time-series waveform of the earthquake motion is not required as input information, and information on the maximum acceleration of the earthquake motion and the spectrum value in the response spectrum analysis of the time-series waveform of the earthquake motion is required by using the trained model. Only. Therefore, there is no need to perform time history response analysis as in the past, and the time required for estimating the degree of disaster can be easily shortened. Furthermore, there is no need to install any or almost no earthquake sensors in the building, making it easier to reduce costs.

A twelfth aspect is a building disaster degree learning system 2, which includes an acquisition section 232 and a learning section 233. The acquisition unit 232 acquires input information including information on maximum acceleration of earthquake motion and spectrum values in response spectrum analysis of time-series waveform of earthquake motion. The learning unit 233 constructs, by machine learning, a trained model that uses input information as input data and uses output information regarding the degree of damage caused by earthquake motion of buildings as training data. According to this aspect, information on the time-series waveform of seismic motion is not required as input information, and information on the maximum acceleration of seismic motion and the spectrum value in response spectrum analysis of the time-series waveform of seismic motion is required by using a trained model. Only. Therefore, there is no need to perform time history response analysis as in the past, and the time required to estimate the degree of disaster can be easily shortened. Furthermore, there is no need to install any or almost no earthquake sensors in the building, making it easier to reduce costs.

A thirteenth aspect is a program that executes the building damage estimation method according to any one of the first to ninth aspects or the building disaster learning method according to the eleventh aspect on one or more processors. let According to this aspect, information on the time-series waveform of the earthquake motion is not required as input information, and information on the maximum acceleration of the earthquake motion and the spectrum value in the response spectrum analysis of the time-series waveform of the earthquake motion is required by using the trained model. Only. Therefore, there is no need to perform time history response analysis as in the past, and the time required for estimating the degree of disaster can be easily shortened. Furthermore, there is no need to install any or almost no earthquake sensors in the building, making it easier to reduce costs.

1 Disaster damage estimation system 132 Acquisition unit 133 Estimation unit 2 Disaster damage learning system 232 Acquisition unit 233 Learning unit 3 Building 4 Neural network

Claims (13)

an acquisition step of acquiring input information including maximum acceleration of earthquake motion and information on a periodic band of spectral values in a range of 0.0 (sec) or more and less than 10.0 (sec) in response spectrum analysis of the time-series waveform of the earthquake motion; ,
A trained model that uses the input information as input data, and that is constructed in advance by machine learning using output information regarding the degree of damage caused by the earthquake motion of a building as training data, based on the input information. an estimation step of estimating the output information ,
The information on the range of the periodic band of 0.0 (sec) or more and less than 10.0 (sec) of the spectrum value is the maximum value of the periodic band of 0.0 (sec) or more and less than 0.1 (sec), the periodic band A maximum value of 0.1 (sec) or more and less than 0.2 (sec), ... Contains each information of a maximum value of a period band of 9.9 (sec) or more and less than 10.0 (sec),
A method for estimating the degree of damage to buildings. The input information includes information on a predetermined acceleration duration,
The predetermined acceleration duration is the time during which the acceleration of the seismic motion exceeds a predetermined acceleration obtained by multiplying the maximum acceleration of the seismic motion by a positive value less than 1, which occupies the predetermined duration of the seismic motion.
The method for estimating the degree of damage to a building according to claim 1. The input information includes information on measured seismic intensity;
The method for estimating the degree of damage to a building according to claim 1 or 2. The output information includes information on the maximum amount of deformation of the building.
The method for estimating the degree of damage to a building according to any one of claims 1 to 3. The building comprises a plurality of layers,
the output information includes information on the maximum deformation amount for the plurality of layers;
The method for estimating the degree of damage to a building according to claim 4. The building comprises a plurality of layers,
The output information includes information on the maximum interlayer deformation amount.
The method for estimating the degree of damage to a building according to any one of claims 1 to 5. The output information includes information on the amount of cumulative plastic deformation of the building.
The method for estimating the degree of damage to a building according to any one of claims 1 to 6. The output information includes information on disaster rank.
The method for estimating the degree of damage to a building according to any one of claims 1 to 7. The trained model uses a neural network,
The method for estimating the degree of damage to a building according to any one of claims 1 to 8. an acquisition unit that acquires input information including maximum acceleration of earthquake motion and information on a periodic band of spectrum values in a range of 0.0 (sec) or more and less than 10.0 (sec) in response spectrum analysis of the time-series waveform of the earthquake motion; ,
A trained model that uses the input information as input data, and that is constructed in advance by machine learning using output information regarding the degree of damage caused by the earthquake motion of the building as training data, based on the input information. an estimation unit that estimates the output information ,
The information on the range of the periodic band of 0.0 (sec) or more and less than 10.0 (sec) of the spectrum value is the maximum value of the periodic band of 0.0 (sec) or more and less than 0.1 (sec), the periodic band A maximum value of 0.1 (sec) or more and less than 0.2 (sec), ... Contains each information of a maximum value of a periodic band of 9.9 (sec) or more and less than 10.0 (sec),
Building damage estimation system. an acquisition step of acquiring input information including maximum acceleration of earthquake motion and information on a periodic band of spectral values in a range of 0.0 (sec) or more and less than 10.0 (sec) in response spectrum analysis of the time-series waveform of the earthquake motion; ,
a learning step of constructing, by machine learning, a trained model that uses the input information as input data, and that uses output information regarding the degree of damage caused by the earthquake motion of the building as training data ;
The information on the range of the periodic band of 0.0 (sec) or more and less than 10.0 (sec) of the spectrum value is the maximum value of the periodic band of 0.0 (sec) or more and less than 0.1 (sec), the periodic band A maximum value of 0.1 (sec) or more and less than 0.2 (sec), ... Contains each information of a maximum value of a period band of 9.9 (sec) or more and less than 10.0 (sec),
A method for learning the degree of damage to buildings. an acquisition unit that acquires input information including maximum acceleration of earthquake motion and information on a periodic band of spectrum values in a range of 0.0 (sec) or more and less than 10.0 (sec) in response spectrum analysis of the time-series waveform of the earthquake motion; ,
a learning unit that constructs, by machine learning, a trained model that uses the input information as input data, and that uses output information regarding the degree of damage caused by the earthquake motion of a building as training data ;
The information on the range of the periodic band of 0.0 (sec) or more and less than 10.0 (sec) of the spectrum value is the maximum value of the periodic band of 0.0 (sec) or more and less than 0.1 (sec), the periodic band A maximum value of 0.1 (sec) or more and less than 0.2 (sec), ... Contains each information of a maximum value of a periodic band of 9.9 (sec) or more and less than 10.0 (sec),
Building damage level learning system. one or more processors,
Executing the building disaster degree estimation method according to any one of claims 1 to 9 or the building disaster degree learning method according to claim 11,
program.

Images