JP7336945B2 - Earthquake damage prediction system - Google Patents

Description

The present invention relates to an earthquake damage prediction system that predicts damage to target buildings due to earthquakes.

Conventionally, an earthquake damage prediction system for predicting damage to target buildings due to earthquakes has been proposed. For example, in the technique disclosed in Patent Document 1, the estimated amount of ground vibration acting on an existing building due to an earthquake and the presence or absence of liquefaction in the ground of the existing building are estimated, and the damage status of the building frame is calculated as a parameter. An earthquake damage prediction system has been proposed.

JP-A-11-84017

Here, in the earthquake damage prediction system of Patent Document 1, it is true that the damage state of the foundation or frame of the building is estimated from the amount of ground vibration obtained by the earthquake using structural analysis etc. for the frame of the building. Is possible. However, for example, damage to non-structural parts such as roofs, outer walls, ceilings, or partition walls cannot be easily estimated by structural analysis. It may differ greatly from the damage to non-structural parts during an actual earthquake.

SUMMARY OF THE INVENTION An object of the present invention is to provide an earthquake damage prediction system capable of more accurately predicting damage to non-structural parts of a target building due to an earthquake.

In view of the above problems, an earthquake damage prediction system according to the present invention is an earthquake damage prediction system for predicting damage to a target building due to an earthquake, which comprises: magnitude of an earthquake at an epicenter, depth of the epicenter, Distance to the building, earthquake information including numerical values corresponding to the earthquake type, numerical values corresponding to the structural type of the existing building, numerical values corresponding to the frame type of the existing building, and values adopted for the existing building Structural information of the existing building including numerical values corresponding to vibration isolation, vibration damping, or earthquake resistance, the number of floors of the existing building, and the total floor area of the existing building, and the existing building including numerical values corresponding to the use of the existing building Usage information of the building, an earthquake damage value quantifying the presence or absence of damage due to the earthquake, the level of the damage, or the probability of the damage to the non-structural parts of the existing building excluding the foundation and skeleton of the existing building; is used as teacher data, the calculation of the earthquake damage value of the non-structural part corresponding to the target building is machine-learned, and from the earthquake information, the structural information, and the usage information corresponding to the target building, the target A computing device is provided for calculating the earthquake damage value of the non-structural part corresponding to the building.

According to the present invention, the computing device uses earthquake information, structural information of the existing building, and usage information of the existing building as training data, and further quantifies the actual damage occurring in the non-structural part of the existing building. The values are also training data, and the calculation of the earthquake damage value is machine-learned. Therefore, it is difficult to predict only the actual structural analysis, and the quantified earthquake damage value (that is, the presence or absence of damage due to the earthquake, the level of the damage, or the probability of the damage) of the non-structural part of the target building can be accurately obtained. It is possible to easily predict the damage to non-structural parts of the target building when an earthquake occurs. An “existing building” is a building that was actually shaken by an earthquake, and the earthquake information of the earthquake that shook the existing building, the structural information of the existing building, the usage information of the existing building, the earthquake damage value of the existing building, etc. It is a building that can be entered. A "target building" is an arbitrary building to which earthquake information predicted for the target building, structural information corresponding to the target building, usage information corresponding to the target building, and the like can be input.

More preferably, the non-structural part is a roof, an outer wall, a ceiling, or a partition wall. Such non-structural parts may crack or fall off due to an earthquake. Damage to these non-structural parts due to earthquakes is usually difficult to predict. It is most suitable for predicting the damage of non-structural parts due to earthquakes.

In a more preferred embodiment, the nonstructural part is a roof, and the teacher data further includes roof information including a numerical value corresponding to the type of the roof of the existing building and a weight per unit area of the roof of the existing building. , the damage is cracking or falling of the roof, or the impact of these within the building, and the arithmetic unit calculates from the earthquake information, the structural information, the usage information, and the roof information corresponding to the target building , calculating the earthquake damage value of the roof corresponding to the target building.

According to the knowledge of the inventors, when identifying the cracks and falling of the roof during an earthquake, or the effects of these in the building, it is particularly important to use numerical values corresponding to the type of the roof of the existing building and the roof of the existing building. It was found that the effect of the unit area weight of Therefore, by using at least these data as teacher data and using a computing device that performs machine learning, it is possible to more accurately calculate the earthquake damage value of the roof corresponding to the target building.

In another preferred aspect, the non-structural portion is an exterior wall material, the teacher data further includes exterior wall material information including a numerical value corresponding to the material of the exterior wall material of the existing building, and the damage is: Cracks and falling of the exterior wall material, or the influence of these in the building, and the arithmetic device calculates the target building from the earthquake information, the structural information, the usage information, and the exterior wall material information corresponding to the target building. The earthquake damage value of the exterior wall material corresponding to the building is calculated.

According to the knowledge of the inventors, when identifying cracks or falling off of exterior wall materials during an earthquake, or the impact of these on the inside of a building, the installation structure of exterior wall materials is specified within a certain range. It was found that the numerical value corresponding to the material of the outer wall material of the building has a large effect. Therefore, by using at least this data as training data and using a machine-learning arithmetic unit, it is possible to more accurately calculate the earthquake damage value of the external wall material corresponding to the target building.

In a more preferred embodiment, the non-structural part is a ceiling, and the training data includes a numerical value corresponding to the structure of the ceiling of the existing building, a unit area weight of the ceiling of the existing building, and a weight per unit area of the ceiling of the existing building. Ceiling information including the number of floors of the existing existing building is further included, the damage is cracking or falling of the ceiling, or the impact of these within the building, and the computing device receives the earthquake information corresponding to the target building. , the earthquake damage value of the ceiling corresponding to the target building is calculated from the structural information, the usage information, and the ceiling information.

According to the knowledge of the inventors, when identifying the cracks and falling of the ceiling at the time of an earthquake, or the effects of these in the building, the structure of the ceiling, the weight per unit area of the ceiling, and the size of the building in which the ceiling exists. It was found that the rank had a large effect. Therefore, by using at least these data as teacher data and using an arithmetic unit that performs machine learning, it is possible to more accurately calculate the earthquake damage value of the ceiling corresponding to the target building.

In a more preferred mode, the non-structural portion is a partition wall, the teacher data further includes partition wall information including a numerical value corresponding to the structure of the partition wall of the existing building, and the damage is the partition wall. Cracks and falling of walls, or the effects of these in the building, and the computing device determines the target building from the earthquake information, the structural information, the usage information, and the partition wall information corresponding to the target building. Calculate the seismic damage value of the corresponding partition wall.

According to the findings of the inventors, it has been found that the structure of the partition wall has a great influence when identifying cracks or falling of the partition wall at the occurrence of an earthquake, or the effects in the building caused by these. Therefore, by using at least these data as teacher data and using a computing device that performs machine learning, it is possible to more accurately calculate the earthquake damage value of the partition wall corresponding to the target building.

In another preferred embodiment in which the non-structural portion is a partition wall, the non-structural portion is a partition wall, and the partition wall comprises a lightweight steel frame base provided with runners and studs and a base plate is attached, and the training data includes the width of the runner of the existing building, the length of the runner, the plate thickness of the runner, the unit area weight of the base plate, and the mounting structure of the base plate Corresponding numerical values are further included as partition wall information, the damage is a crack or fall of the partition wall, or an impact in the building caused by these, and the computing device receives the earthquake information and the structure corresponding to the target building. The earthquake damage value of the partition wall corresponding to the target building is calculated from the information, the usage information, and the partition wall information.

According to the findings of the inventors, when a partition wall has a lightweight steel frame foundation with runners and studs, when identifying cracks, falling off, or the effects of these in the building during an earthquake, , the dimensions of the runner, the weight per unit area of the base plate, and the attachment structure of the base plate have a large effect. Therefore, by using at least these data as teacher data and using a computing device that performs machine learning, it is possible to more accurately calculate the earthquake damage value of the partition wall corresponding to the target building.

In another preferred embodiment when the non-structural part is a partition wall, the non-structural part is a partition wall, and the partition wall is an ALC board attached to the frame of the building via a mounting bracket. The training data includes the thickness of the ALC board of the existing building, the width of the ALC board, the number of mounting brackets for mounting the ALC board, and the presence or absence of studs or wind-resistant beams that support the ALC board. as partition wall information, wherein the damage is a crack or fall of the partition wall, or an impact in the building caused by these, and the arithmetic unit stores the earthquake information corresponding to the target building, the The earthquake damage value of the partition wall corresponding to the target building is calculated from the structural information, the usage information, and the partition wall information.

According to the knowledge of the inventors, when the partition wall has a structure in which the ALC board is attached to the building skeleton, when an earthquake occurs, the partition wall cracks, falls, or when identifying the impact inside the building due to these In addition, it was found that the thickness of the ALC board, the width of the ALC board, the number of mounting brackets for mounting the ALC board, and the presence or absence of studs or wind-resistant beams supporting the ALC board have a large effect. Therefore, by using at least these data as teacher data and using a computing device that performs machine learning, it is possible to more accurately calculate the earthquake damage value of the partition wall corresponding to the target building.

In another preferred embodiment when the non-structural part is a partition wall, the non-structural part is a partition wall, the partition wall is a masonry partition wall, and the teacher data includes: The partition wall information further includes a numerical value corresponding to the presence or absence of reinforcing bars in the partition wall of the existing building, and the damage is a crack or falling of the partition wall, or an effect of these in the building, and the computing device. calculates the earthquake damage value of the partition wall corresponding to the target building from the earthquake information, the structural information, the use information, and the partition wall information corresponding to the target building.

According to the knowledge of the inventors, when the partition wall is a masonry partition wall, when an earthquake occurs, the partition wall cracks or falls off, or when identifying the impact of these in the building, the inside of the partition wall It was found that the presence or absence of reinforcing bars has a large effect. Therefore, by using at least these data as teacher data and using a computing device that performs machine learning, it is possible to more accurately calculate the earthquake damage value of the partition wall corresponding to the target building.

In a more preferred embodiment, the teacher data includes a surface ground amplification factor of the ground of the existing building, a depth from the ground surface of the construction site obtained by a boring test at the construction site of the existing building, and soil properties corresponding to the depth. and the ground information including the N value in the standard penetration test according to the depth, and the basic information including the numerical value corresponding to the basic structure of the existing building, further including the arithmetic device calculates the earthquake damage value of the non-structural part corresponding to the target building from the earthquake information, the structural information, the usage information, the ground information, and the basic information corresponding to the target building.

According to this aspect, since the arithmetic unit is learned by further considering the ground information and the basic information, the foundation structure of the building, the ground and its influence are further taken into consideration, and the target building can be handled with higher accuracy. It is possible to calculate the seismic damage value of non-structural parts.

According to the present invention, it is possible to more accurately predict damage to non-structural parts of a target building due to an earthquake.

BRIEF DESCRIPTION OF THE DRAWINGS It is a schematic conceptual diagram for demonstrating the earthquake damage prediction system which concerns on 1st Embodiment of this invention. FIG. 2 is a table showing teacher data for machine learning by the arithmetic device of the earthquake damage prediction system shown in FIG. 1 ; FIG. 2 is a table showing teacher data for additional machine learning by the arithmetic device of the earthquake damage prediction system shown in FIG. 1 ; FIG. 2 is a schematic conceptual diagram during learning of a neural network of the arithmetic unit shown in FIG. 1; FIG. 2 is a schematic conceptual diagram of utilization of a neural network of the arithmetic unit shown in FIG. 1; FIG. 9 is a table showing training data for machine learning by the arithmetic device of the earthquake damage prediction system according to the second embodiment of the present invention. FIG. 5 is a schematic conceptual diagram of a neural network of an arithmetic device according to a second embodiment; FIG. 11 is a table showing teacher data for machine learning by the arithmetic unit of the earthquake damage prediction system according to the third embodiment of the present invention. FIG. 11 is a schematic conceptual diagram of a neural network of an arithmetic device according to a third embodiment; FIG. 11 is a table showing teacher data for machine learning by the arithmetic device of the earthquake damage prediction system according to the fourth embodiment of the present invention. (a) A schematic perspective view of a partition wall in which a base plate is attached to a lightweight steel frame base provided with runners and studs, and (b) a partition in which an ALC board is attached to a building frame via mounting brackets. FIG. 4 is a schematic perspective view of a wall; FIG. 11 is a schematic conceptual diagram of a neural network of an arithmetic device according to a fourth embodiment; FIG. 11 is a schematic conceptual diagram of a neural network of an arithmetic device according to a first modified example of the fourth embodiment; FIG. 12 is a schematic conceptual diagram of a neural network of an arithmetic device according to a second modified example of the fourth embodiment; FIG. 11 is a schematic conceptual diagram of a neural network of an arithmetic device according to a third modified example of the fourth embodiment;

An earthquake damage prediction system 1 according to first to fourth embodiments of the present invention will be described below with reference to the drawings.

[First embodiment]
1. Device Configuration of Earthquake Damage Prediction System 1 As shown in FIG. This is to predict the damage to the target building due to an earthquake. The earthquake damage prediction system 1 will be described below.

The main server 10 of the earthquake damage prediction system 1 includes a computing device 10A that generates a program for predicting damage to a target building (calculating earthquake damage values of nonstructural parts), and a keyboard that inputs data to the computing device 10A. and an input device 2 . In addition, the non-structural part here means a part excluding the frame (specifically, columns and beams) of the building or the foundation.

The arithmetic unit 10A includes a storage unit 10a such as a ROM or RAM in which information about existing buildings damaged by an earthquake is stored, and an arithmetic unit 10b such as a CPU for generating a machine-learned program. I have.

The portable terminal 20 is capable of communicating with the main server 10 via a network, and is installed with a program for calculating the earthquake damage value of the roof machine-learned by the arithmetic unit 10b of the main server 10 . The mobile terminal 20 includes an arithmetic unit 20A for calculating the seismic damage value of the roof, which is a non-structural part, and a display/input unit 20B such as a touch panel for inputting and displaying data.

The mobile terminal 20 includes, as an arithmetic unit 20A, a storage unit 20a made up of a ROM or a RAM or the like for storing the program for calculating the seismic damage value of the roof from the main server 10 and the data input by the display/input unit 20B; and a calculation unit 20b including a CPU or the like that executes an earthquake damage value calculation program.

In this embodiment, the earthquake damage prediction system 1 is provided with a main server 10 and a mobile terminal 20 separately. not to be

Moreover, the earthquake damage prediction system 1 may be configured as one system instead of being configured with the main server 10 and the mobile terminal 20 . Specifically, the storage unit 10a of the main server 10 stores data stored in the storage unit 20a of the mobile terminal 20, and the calculation unit 10b of the main server 10 performs calculation by the calculation unit 20b of the mobile terminal 20. Therefore, the mobile terminal 20 may be omitted. In addition, the portable terminal 20 does not perform the above-described storage and calculation, and the main server 10 performs all the storage and calculation, and the portable terminal 20 stores only the input information for the target building and the calculation result of the earthquake damage value of the roof. may be output.

2. Arithmetic device 10A (machine learning)
In this embodiment, machine learning by the arithmetic device 10A of the main server 10 will be described below. In the present embodiment, the arithmetic unit 10A provides an earthquake damage value that quantifies the presence or absence of damage due to an earthquake, the level of damage, or the probability of damage to the roof as a non-structural part of the existing building excluding the foundation and skeleton of the existing building. Build a model that has learned to compute. It should be noted that, in another embodiment to be described later, the object of this non-structural portion is different, and the information of the structural portion of the teacher data and its earthquake damage value are different.

In the present embodiment, the arithmetic device 10A provides (1) earthquake information, (2) structure information of the existing building, (3) usage information of the existing building, (4) roof information of the existing building, and (5) information about the existing building. This is a part for machine learning using earthquake damage values of non-structural parts (roofs) as teacher data, and the result of learning in this arithmetic unit 10A is constructed as a model. Note that the generated program is transmitted to the mobile terminal 20 at the time of utilization. Each piece of information serving as teacher data will be described below with reference to FIGS. 2 and 3. FIG.

3. Teacher data 3-1. Earthquake Information The earthquake information includes "magnitude of the earthquake at the epicenter", "depth of the epicenter", "distance from the epicenter to existing buildings", and "numerical value corresponding to the type of earthquake".

"Earthquake magnitude at epicenter" is the magnitude of an earthquake that occurred around an existing building that is the target of training data for machine learning. It is indicated by the scale (magnitude) of the earthquake. "Focus depth" is the depth of the epicenter of the earthquake. The "distance from the epicenter to the existing building" is, for example, the distance between the earthquake source fault and the existing building, using the fault shortest distance. For example, it may be a substantial distance between the epicenter and an existing building. The "value that quantifies the type of earthquake" is, for example, a value that is associated with three types of interplate earthquakes, oceanic intraplate earthquakes, and inland intracrustal earthquakes. It should be noted that these classified earthquake types may be further subdivided and associated with numerical values.

In this embodiment, the above-described earthquake information is used as teacher data. For example, the maximum ground motion acceleration and maximum ground motion velocity acting on the existing building are calculated using the distance attenuation formula, and these values are used as teacher data. Alternatively, if an existing building is provided with an acceleration sensor, these measured values may be used as teacher data.

3-2. Structural Information of Existing Buildings Structural information of existing buildings includes “numerical values corresponding to the structural form of the existing building”, “numerical values corresponding to the frame form of the existing building”, "value corresponding to earthquake resistance", "number of floors of existing buildings", and "total floor area of said existing buildings".

The "numerical value corresponding to the structural type of the existing building" is a numerical value associated with each structural type classified by the structural type of the existing building, such as S, RC, SRC, and wooden. "Numerical value corresponding to the frame format of the existing building" is the numerical value corresponding to Ramen and Breath. "Values corresponding to the seismic isolation, damping, or earthquake resistance adopted in the existing building" are the presence or absence of seismic isolation devices for seismic isolation, the presence or absence of damping devices such as dampers for vibration control, and the presence or absence of vibration control devices such as dampers for seismic resistance. , the presence or absence of braces, and the like. These numerical values may be assigned in descending order of roof damage susceptibility.

In addition, as shown in FIG. 3, the structural information of the existing building may further include "the value of the structural calculation result of the existing building" as teacher data, as shown in FIG. As the structural calculation result of the existing building, for example, the existing building's horizontal bearing capacity margin and the existing building's relative deformation angle (short term, final stage) are used as teacher data.

3-3. Use information of existing buildings The use information of existing buildings includes “numerical values corresponding to the use of existing buildings”. Numerical values corresponding to the uses of existing buildings vary depending on their uses, such as hospitals, schools, factories, hotels, condominiums, office buildings, detached houses, collective housing, etc. It is a value quantified and set according to these uses. For example, these numerical values may be assigned in descending order of roof damage susceptibility.

3-4. Roof Information of Existing Building The roof information of existing building includes "numerical value corresponding to the type of roof of existing building" and "weight per unit area of roof of existing building". "Number corresponding to the roof type of the existing building" is classified into, for example, folded plate roof, deck type dry insulation roof, deck concrete roof, roof with extruded cement plate, etc., and the corresponding numerical value is. If the existing building is a residence, the "numerical value corresponding to the type of roof of the existing building" may be a numerical value corresponding to the structure of the roof, roof tiles, or the like. For example, these numerical values may be assigned in descending order of roof damage susceptibility. "The weight per unit area of the roof of the existing building" is the weight of the roof per unit area of the roof surface when the roof of the existing building is viewed from above, and it may be the weight per unit area including at least the roof material. It is not particularly limited as long as the weight of the roof is specified according to the same standard. Therefore, the "weight per unit area of the roof of the existing building" may be, for example, the weight of the roof material alone, or the average weight per unit area of the roof structure from the roof surface to the ceiling of the top floor. .

3-5. Information on Roof Damage to Existing Buildings "Information on roof damage to existing buildings" quantifies the damage to the roofs of existing buildings caused by the above-mentioned earthquake. Specifically, "roof damage information on existing buildings" is an earthquake damage value that quantifies the presence or absence of damage due to an earthquake, the level of damage, or the probability of damage to the roof of an existing building, excluding the foundation and frame of the existing building. is. For example, "presence or absence of damage due to an earthquake" was quantified with "0" and "1" to indicate whether the roof of an existing building was cracked, fell off, or whether the crack or fall affected the inside of the building. is the earthquake damage value. The effects in the building due to cracks or falling off include rain leaks and the like. "Earthquake damage level" is an earthquake damage value that quantifies the degree of damage to existing buildings due to roof cracks, roof falls, or the impact of these on the roofs of existing buildings. is severely damaged. The "probability of damage due to an earthquake" is quantified as the probability of damage according to the degree of damage to the existing building due to cracks in the roof, falling off of the roof, or these. For example, a higher probability may be set as the degree of influence on the existing building due to cracks, falling roofs, or any of these increases. Damage may be obtained as a probability by logistic regression analysis or the like using a numerical value of the degree of influence of the building.

3-6. Ground Information of Existing Building Further, although not shown in the models of FIGS. 4 and 5, as shown in FIG. 3, the training data may include ground information of the existing building, if necessary. The ground information of the existing building includes "surface ground amplification factor of the ground of the existing building", "depth from the ground surface of the construction site obtained from the boring test of the construction site of the existing building", and "soil quality according to the depth". A numerically set value”, and an “N value in a standard penetration test according to depth”.

"Surface ground amplification factor of existing building ground" is a quantification of the magnitude of shaking of the stratum deposited near the ground surface of an existing building (surface ground) during an earthquake, and indicates the vulnerability of the ground to earthquakes. Numeric value. The "depth from the ground surface of the construction point" to be input is, for example, the depth that makes up each soil type, such as the bottom depth of the layer that makes up each soil type, and the information on the layer thickness of that soil type can be specified. The value of "depth" associated with the value of the soil parameter and the N value is not particularly limited as long as the "depth" can be input.

Generally, in the boring test, the type of soil is investigated according to the depth from the ground surface. Here, the "value set by quantifying the soil quality" is a parameter quantified and set for each soil quality. For example, the soil quality can include clay, silt, fine sand, coarse sand, fine gravel, gravel, coarse gravel, cobble, or boulder, and is quantified for each soil quality. For example, a numerical value corresponding to the particle size distribution according to the type of soil may be set. For example, as the particle size distribution according to the type of soil increases, the bearing capacity to support the load from the building increases (ground subsidence is less likely). .

"N value in the standard penetration test" refers to the N value measured in the standard penetration test (test in accordance with JIS A 1219), the converted N value based on the Swedish sounding test (test in accordance with JIS A 1212), etc. The larger the N value, the greater the bearing capacity to support the load from the building.

3-7. Basic Information of Existing Building Further, although not shown in the models of FIGS. 4 and 5, as shown in FIG. 3, the training data may include basic information of the existing building, if necessary. The basic information is parameters set by quantifying the type of foundation selected for the existing building. Specific examples include a direct foundation structure with piles, a pile draft foundation structure with piles, a pile foundation structure with friction piles, and a pile foundation structure with support piles supported by a support layer. It is set by quantifying the basic form (form of the basic structure) of For example, numerical values are assigned in descending order of roof damage.

Furthermore, in the case of a pile foundation structure, depending on the length of the friction pile, the type of the friction pile, etc., the numerical value of the size may be set according to the order of susceptibility to ground subsidence. Furthermore, in addition to the foundation structure, for example, a foundation structure improved with cement milk may also be included, and the above-mentioned numerical values are set for this structure as well.

4. Neural network Using the information shown in Fig. 2 as training data and, if necessary, using the information shown in Fig. 3 as training data, the earthquake information corresponding to the target building, structural information, and usage information are used to determine the roof earthquake corresponding to the target building. Learning to calculate the damage value by machine learning.

In the present embodiment, the arithmetic device 10A is configured so that the output values calculated with respect to the input values of the existing buildings A, B, C, . . . The calculation of the earthquake damage value is machine-learned so as to converge to the earthquake damage value (actual earthquake damage value) set based on the actual roof damage.

This learning may be performed, for example, by repeatedly correcting a correction coefficient by which each variable is multiplied for a predetermined mathematical formula made up of variables corresponding to the number shown in FIGS. 2 and 3 . In the present embodiment, as shown in FIGS. 4 and 5, the arithmetic device 10A includes a deep neural network ((DNN): hereinafter referred to as "neural network"), and the neural network 11 is configured for the earthquake shown in FIG. The earthquake damage value of the roof of the target building is calculated as an output value from the target building information corresponding to the information, the structure information of the existing building, the usage information of the existing building, and the roof information of the existing building.

In this embodiment, the neural network 11 has an input layer 11A. The input layer 11A receives earthquake information, structural information, usage information, roof information, and earthquake damage values of the roofs of the existing buildings A, B, C, . . . . Note that the ground information and basic information of the existing building may be used as input values for the training data. The input layer 11A is composed of a plurality of input layer neuron elements 11a corresponding to the number of input values of earthquake information, structural information, usage information, and roof information. The number of input layer neuron elements 11a in the input layer 11A can be increased or decreased according to the number of input condition data, and a model of the neural network 11 is constructed according to the number of input data due to this increase or decrease.

The neural network 11 has an output layer 11E. The output layer 11E is composed of one output layer neuron element 11e that outputs the earthquake damage value of the roof. The neural network 11 has three intermediate layers 11B, 11C, 11D. Three intermediate layers 11B, 11C, 11D are provided between the input layer 11A and the output layer 11E. Each intermediate layer 11B, 11C, 11D includes a plurality of intermediate layer neuron elements 11b, 11c, 11d directly or indirectly coupled to these elements.

In this embodiment, the intermediate layer is composed of three layers, but for example, the intermediate layer may be composed of one, two, four or more layers, It is not limited to layers. Further, the number of intermediate layer neuron elements 11b, 11c, and 11d constituting each of the intermediate layers 11B, 11C, and 11D corresponds to the number of input layer neuron elements 11a. The number of neuron elements 11a may be changed, and each intermediate layer may include the number of intermediate layer neuron elements different from the number of input data.

The intermediate layers 11B, 11C, and 11D perform predetermined calculations using the neuron parameter values of the neuron elements in the same layer on the input layer 11A side, and output the calculation results to the neuron elements on the output layer 11E side. It is. Specifically, the intermediate layer neuron elements 11b, 11c, and 11d and the output layer neuron element 11e receive the neuron parameter values input from the input layer neuron element 11a or the intermediate layer neuron elements 11b, 11c, and 11d on the input layer 11A side. is used to perform a predetermined calculation.

Specifically, each of the intermediate layer neuron elements 11b, 11c, 11d and the output layer neuron element 11e, which execute operations, has a predetermined activation function, and the input data (parameter values) are Calculate the value of the neuron parameter by substituting it into the activation function. For example, each intermediate layer neuron element 11b of the intermediate layer 11B receives the input value of each input layer neuron element 11a as a neuron parameter value, and the activation function calculates the neuron parameter value for each input layer neuron element 11a. be done.

At this time, the values of the neuron parameters output from the intermediate layer neuron elements 11b, 11c, and 11d of the intermediate layers 11B, 11C, and 11D correspond to the values of the neuron parameters calculated by the activation function in the elements. , is multiplied by a weighting factor, and is output to the output layer neuron element 11e or the intermediate layer neuron elements 11c and 11d on the output layer 11E side.

In this embodiment, the earthquake damage values of the roofs of the existing buildings A, B, C, . The weighting coefficients by which the neuron parameter values of the intermediate layer neuron elements 11b, 11c, and 11d are multiplied are repeatedly corrected until the value (the input value as teacher data) converges within a preset range. In this manner, in this embodiment, the calculation (method) of the earthquake damage value of the roof and the output value can be learned in advance by machine learning. In this way, machine learning is executed in the computing device 10A using the earthquake information, the structural information of the existing building, the usage information of the existing building, and the roof information of the existing building at the plurality of construction points P1, P2, P3, . . . Then, a model for calculating the seismic damage value of the roof of the target building is constructed from the information of any target building.

In the present embodiment, the weighting coefficients multiplied by the neuron parameter values calculated by the activation functions in the intermediate layer neuron elements 11b, 11c, and 11d are assigned to the intermediate layer neuron elements 11b, 11c, and 11d. In addition to this, the output layer neuron element 11e may also be provided with a weighting factor by which the value of the neuron parameter calculated by the activation function is multiplied.

The program of the model constructed in this way is installed in the arithmetic device 20A, and the arithmetic device 20A, as shown in FIG. From the roof, the earthquake damage value of the roof corresponding to the target building is calculated.

According to this embodiment, the arithmetic unit 20A uses earthquake information, existing building structure information, and existing building usage information as training data, and furthermore, calculates damage that has occurred on the roof, which is an actual non-structural part of the existing building. The quantified earthquake damage value is also used as training data, and a program that performs machine learning to calculate the earthquake damage value is provided. Therefore, the quantified earthquake damage value of the roof damage of the target building (that is, the presence or absence of roof damage due to an earthquake, the level of roof damage, or the probability of roof damage), which is difficult to predict by actual structural analysis alone, is calculated. Accurately calculated, it is possible to easily predict the damage to the roof of the target building when an earthquake occurs.

In this embodiment, roof information is input as teacher data, but roof information need not be used as teacher data if, for example, rough prediction of roof damage is desired.

[Second embodiment]
The difference between the second embodiment and the first embodiment is that the non-structural part is the outer wall material, and the outer wall material information of the existing building is used as training data, and the calculation of the earthquake damage value of the outer wall material of the target building is performed by a machine. This is a point I learned. Therefore, only the differences from the first embodiment will be referred to in FIGS. 6 and 7, and detailed description thereof will be omitted.

As shown in FIG. 6, in the second embodiment, out of the training data of the first embodiment, instead of the roof information of the existing building, exterior wall material information including "numerical values corresponding to the material of the exterior wall material of the existing building" , and instead of the damage information on the roof of the existing building, "damage information on the external wall material of the existing building" is used.

"Numerical value corresponding to the material of the exterior wall material of the existing building" includes ceramic siding, extruded cement board, square corrugated steel plate, heat insulating sandwich panel, ALC board, glass curtain wall, etc. Furthermore, ceramic siding and extrusion molding In the case of a cement board and an ALC board, it is a numerical value associated with the presence or absence of tile attachment. For example, these numerical values may be assigned in descending order of the number of applications in which the exterior wall material is likely to be damaged.

"Damage information on exterior wall materials of existing buildings" is earthquake damage values that quantify the presence or absence of earthquake damage, the level of damage, or the probability of damage to the exterior wall materials of existing buildings, excluding the foundation and frame of the existing building. . Here, damage to exterior wall materials means cracks, falling off, or the effects of these within the building. The effects in the building due to cracks or falling off include rain leaks and the like. The setting of numerical values for "presence or absence of damage due to earthquake", "level of damage due to earthquake", and "probability of damage due to earthquake" is the same as for damage to the roof material described above.

Then, in the arithmetic device 10A, as shown in FIG. 6, by machine learning, the earthquake information, structural information, usage information, outer wall information, and earthquake damage value of the outer wall material of the existing building are used as training data for an arbitrary target building. Build a calculation model for the earthquake damage value of exterior wall materials. The arithmetic device 20A calculates the earthquake damage value of the exterior wall material corresponding to the target building from the earthquake information, the structural information, the usage information, and the exterior wall material information corresponding to any target building at the time of utilization using the constructed model. do. The configuration, learning, and utilization of the neural network 11 are the same as in the first embodiment, so detailed description thereof will be omitted.

By using the arithmetic unit 10A that performs machine learning using the information described above as training data, the same effect as the effect of the first embodiment can be obtained, and the earthquake damage value of the outer wall corresponding to an arbitrary target building can be obtained from less input information. can be calculated more accurately. In this embodiment, external wall material information is input as training data, but external wall material information may not be used as training data if, for example, rough prediction of roof damage is desired. The information shown in FIG. 3 shown in the first embodiment may be further used as teacher data.

[Third embodiment]
The third embodiment differs from the first embodiment in that the non-structural part is the ceiling, and the ceiling information of the existing building is used as training data, and the calculation of the earthquake damage value of the ceiling information of the target building is machine-learned. It is a point. Therefore, only the differences from the first embodiment will be referred to in FIGS. 8 and 9, and detailed description thereof will be omitted.

As shown in FIG. 8, in the third embodiment, instead of the roof information of the existing building in the training data of the first embodiment, "numerical value corresponding to the structure of the ceiling of the existing building", "ceiling of the existing building Ceiling information including "weight per unit area" and "floor number of the existing building where the ceiling exists" is used. Furthermore, instead of the damage information on the roof of the existing building, "damage information on the ceiling of the existing building" is used. In addition, the clearance between the inner and outer walls and the ceiling may be used as teacher data.

The "numerical value corresponding to the ceiling structure of the existing building" is a numerical value associated with a suspended ceiling, a direct ceiling, a system ceiling, an optical ceiling, a curtain ceiling, or the like. For example, these numerical values may be assigned in descending order of ceiling damage susceptibility.

"Weight per unit area of the ceiling of an existing building" is the weight of the ceiling per unit area of the ceiling of an existing building. It is not particularly limited as long as it is. Thus, the "weight per unit area of the ceiling of an existing building" may be, for example, the weight of the ceiling slab or ceiling panel alone, or in the case of a suspended ceiling, the unit area including the ceiling panel, ceiling joists and joists. It may be an average weight per unit. "Number of floors of existing building with ceiling" is the number of floors of existing building with damaged ceiling.

The ``damage information on the ceiling of an existing building'' is an earthquake damage value that quantifies the presence or absence of damage due to an earthquake, the level of damage, or the probability of damage to the ceiling of an existing building, excluding the foundation and frame of the existing building. Here, damage to the ceiling means cracks, falls, or the effects of these within the building. Impacts within the building due to cracks or falls include rain leaks, unavailability of rooms or passageways, and the like. The setting of numerical values for "presence or absence of damage due to earthquake", "level of damage due to earthquake", and "probability of damage due to earthquake" is the same as for damage to the roof material described above.

Then, in the arithmetic unit 10A, as shown in FIG. 9, by machine learning, the earthquake information, the structural information, the usage information, the ceiling information, and the earthquake damage value of the ceiling of the existing building are used as teacher data, and the ceiling for an arbitrary target building is calculated. construct a calculation model for the earthquake damage value of The arithmetic unit 20A calculates the earthquake damage value of the ceiling corresponding to the target building from the earthquake information, the structural information, the usage information, and the ceiling information corresponding to any target building when using the constructed model. The configuration, learning, and utilization of the neural network 11 are the same as in the first embodiment, so detailed description thereof will be omitted.

By using the arithmetic unit 10A that performs machine learning using the information described above as teacher data, the same effect as the effect of the first embodiment can be obtained, and the seismic damage value of the ceiling corresponding to an arbitrary target building can be obtained from less input information. can be calculated more accurately. In this embodiment, ceiling information is input as teacher data, but ceiling information may not be used as teacher data if, for example, rough prediction of roof damage is desired. The information shown in FIG. 3 shown in the first embodiment may be further used as teacher data.

[Fourth embodiment]
The difference between the third embodiment and the first embodiment is that the non-structural part is a partition wall, and the partition wall information of the existing building is used as training data to calculate the earthquake damage value of the partition wall information of the target building. This is the point of machine learning. Therefore, only the differences from the first embodiment will be referred to in FIGS. 10 to 15, and detailed description thereof will be omitted.

As shown in FIG. 10, in the fourth embodiment, in the training data of the first embodiment, instead of the roof information of the existing building, partition wall information including "numerical values corresponding to the structure of the partition wall of the existing building" is used. , and instead of the damage information on the roof of the existing building, the "damage information on the partition wall of the existing building" is used.

The "numerical value corresponding to the structure of the partition wall of the existing building" is the structure in which the base plate 63 is attached to the lightweight steel frame base 69 provided with the runner 61 and the stud 62 shown in FIG. ) in which the ALC board 66 is attached to the building frame 65 via the mounting bracket 68, or the masonry structure using bricks, concrete blocks, or the like. The "numerical value corresponding to the structure of the partition wall of the existing building" is the floor number of the existing building where the damaged ceiling exists.

"Damage information on partition walls of existing buildings" is an earthquake damage value that quantifies the presence or absence of damage due to an earthquake, the level of damage, or the probability of damage to the partition walls of existing buildings, excluding the foundation and frame of the existing building. . Here, the damage to the partition wall means cracking, falling, or the influence in the building due to these, and the influence in the building means whether the room or the passage can be used or not. The setting of numerical values for "presence or absence of damage due to earthquake", "level of damage due to earthquake", and "probability of damage due to earthquake" is the same as for damage to the roof material described above.

Then, as shown in FIG. 12, in the arithmetic unit 10A, by machine learning, earthquake information, structural information, usage information, partition wall information, and partition wall earthquake damage value of the existing building are used as training data for an arbitrary target building. A model for calculating seismic damage values of partition walls is constructed. When utilizing the constructed model, the computing device 20A calculates the earthquake damage value of the partition wall corresponding to the target building from the earthquake information, structural information, usage information, and partition wall information corresponding to any target building. do. In particular, the structure of the partition wall is greatly affected by the earthquake damage value because the structure of the partition wall is greatly different. A seismic damage value can be calculated. In the present embodiment, partition wall information is input as teacher data, but partition wall information may not be used as teacher data if rough damage to partition walls is to be predicted, for example. The information shown in FIG. 5 shown in the first embodiment may be further used as teacher data.

Here, in the case of the structure of the partition wall 60A shown in FIG. 11(a), as in the first modification of the fourth embodiment, the partition wall information of the existing building, which serves as teacher data, includes the "runner" of the existing building. 61 width B1", "length L of the runner 61 of the existing building", "thickness of the runner 61", "unit area weight of the base plate 63", and "numerical value corresponding to the installation structure of the base plate 63". You may use (refer partition wall information 1 of FIG. 10). In this case, a model of the neural network 11 shown in FIG. 13 is constructed.

“Width B1 of runner 61” is the maximum width of runner 61 in a direction orthogonal to the longitudinal direction. “Length L of runner 61 of existing building” is the length along the longitudinal direction of runner 61 . "Thickness of runner 61" is the thickness of runner 61 bent in a U-shape. The “weight per unit area of base plate 63 ” is due to the material of base plate 63 . In FIG. 11(b), two plate materials are stuck together, but in this case, the "weight per unit area of base plate 63" is the sum of the weight per unit area of each plate material. The number of base plates 63 may be one, or three or more. The "numerical value corresponding to the mounting structure of the base plate 63" is a numerical value associated with the joint structure between the base plate 63 and the runner 61, and is a numerical value corresponding to the type of structure such as a joint structure via screws. . For example, these numerical values may be assigned in descending order of the applications in which the partition wall 60A is likely to be damaged.

In the case of the structure of the partition wall 60B shown in FIG. 11(b), as in the second modification of the fourth embodiment, the partition wall information of the existing building, which serves as teacher data, includes "ALC board of the existing building. 66 thickness", "Width B of ALC board 66", "Number of mounting brackets 68 for mounting ALC board 66", and "With or without studs 67 or wind beams (not shown) supporting ALC board 66". "Numeric value" may be used (see partition wall information 2 in FIG. 10). In this case, a model of the neural network 11 shown in FIG. 14 is constructed.

The “thickness of the ALC board 66 of the existing building” is the thickness of the ALC board 66 itself. “Width B of ALC board 66” is the size of ALC board 66 in the horizontal direction. The "number of fittings 68 for attaching the ALC board 66" is, for example, the number of fittings 68 attached to the frame 65 or the like. Note that the mounting pitch of the metal fittings 68 of the existing building may be used as teaching data. Furthermore, the "numerical value corresponding to the presence or absence of the studs 67 or the wind-resistant beams (not shown) supporting the ALC board 66" may be assigned, for example, "0" for none and "1" for presence. Along with "numerical values corresponding to the presence or absence of studs 67 or wind-resistant beams (not shown) that support the ALC board 66", "numerical values for the rigidity and strength of the studs 67 or wind-resistant beams (not shown) that support the ALC board 66" "value obtained by converting" may be used as teacher data.

Furthermore, when the structure of the partition wall is a masonry structure, as in the third modification of the fourth embodiment, the partition wall information of the existing building, which serves as teacher data, includes "reinforcing reinforcing bars in the partition wall." A numerical value corresponding to the presence/absence” may be used (see partition wall information 3 in FIG. 10). In this case, a model of the neural network 11 shown in FIG. 15 is constructed.

The "numerical value corresponding to the presence/absence of reinforcing bars in the partition wall" may be assigned to the presence/absence of reinforcing bars in the partition wall on which bricks or concrete blocks are piled, for example, "0" for none and "1" for yes.

In the first to third modifications, the arithmetic device 10A, as shown in FIGS. Using earthquake damage values of walls as training data, a calculation model of earthquake damage values of partition walls for an arbitrary target building is constructed. When utilizing the constructed model, the computing device 20A calculates the earthquake damage value of the partition wall corresponding to the target building from the earthquake information, structural information, usage information, and ceiling information corresponding to any target building. . The configuration, learning, and utilization of the neural network 11 are the same as in the first embodiment, so detailed description thereof will be omitted.

By using the arithmetic device 10A that performs machine learning using the above-described information as teacher data, the same effect as the effect of the first embodiment can be obtained, and earthquake damage to the partition wall corresponding to any target building can be achieved from less input information. values can be calculated more accurately. In the present embodiment, partition wall information is input as teacher data, but partition wall information may not be used as teacher data if rough damage to partition walls is to be predicted, for example. The information shown in FIG. 5 shown in the first embodiment may be further used as teacher data.

Although the embodiments of the present invention have been described in detail above, the present invention is not limited to the above-described embodiments, and various designs can be made without departing from the spirit of the invention described in the claims. Changes can be made.

1: Earthquake damage prediction system, 10: Main server, 10A: Computing device, 20: Portable terminal, 20A: Computing device, 11: Neural network, 61: Runner

Claims (4)

An earthquake damage prediction system for predicting damage to a target building due to an earthquake,
earthquake information including the magnitude of the earthquake at the epicenter, the depth of the epicenter, the distance from the epicenter to the existing building, and numerical values corresponding to the type of the earthquake;
A numerical value corresponding to the structural form of the existing building, a numerical value corresponding to the frame form of the existing building, a numerical value corresponding to the seismic isolation, vibration damping, or earthquake resistance adopted in the existing building, the number of floors of the existing building, and the Structural information of the existing building including the total floor area of the existing building;
Usage information of the existing building including a numerical value corresponding to the usage of the existing building;
Presence or absence of damage due to the earthquake, the level of the damage, or an earthquake damage value that quantifies the probability of the damage to the non-structural parts of the existing building excluding the foundation and skeleton of the existing building, as teaching data, Machine learning is performed to calculate the earthquake damage value of the non-structural part corresponding to the target building,
an arithmetic device for calculating the earthquake damage value of the non-structural part corresponding to the target building from the earthquake information, the structural information, and the usage information corresponding to the target building ;
The non-structural part is a partition wall, and the partition wall has a structure in which a base plate is attached to a lightweight steel frame base provided with runners and studs,
The teacher data further includes partition wall information including numerical values corresponding to the structure of the partition wall of the existing building, and the partition wall information includes the width and length of the runner of the existing building. , a numerical value corresponding to the plate thickness of the runner, the unit area weight of the base plate, and the mounting structure of the base plate,
The damage is cracking or falling of the partition wall, or the impact of these in the building,
The computing device calculates the earthquake damage value of the partition wall corresponding to the target building from the earthquake information, the structural information, the usage information, and the partition wall information corresponding to the target building. earthquake damage prediction system. An earthquake damage prediction system for predicting damage to a target building due to an earthquake,
earthquake information including the magnitude of the earthquake at the epicenter, the depth of the epicenter, the distance from the epicenter to the existing building, and numerical values corresponding to the type of the earthquake;
A numerical value corresponding to the structural form of the existing building, a numerical value corresponding to the frame form of the existing building, a numerical value corresponding to the seismic isolation, vibration damping, or earthquake resistance adopted in the existing building, the number of floors of the existing building, and the Structural information of the existing building including the total floor area of the existing building;
Usage information of the existing building including a numerical value corresponding to the usage of the existing building;
Presence or absence of damage due to the earthquake, the level of the damage, or an earthquake damage value that quantifies the probability of the damage to the non-structural parts of the existing building excluding the foundation and skeleton of the existing building, as teaching data, Machine learning is performed to calculate the earthquake damage value of the non-structural part corresponding to the target building,
an arithmetic device for calculating the earthquake damage value of the non-structural part corresponding to the target building from the earthquake information, the structural information, and the usage information corresponding to the target building ;
The non-structural part is a partition wall, and the partition wall has a structure in which an ALC board is attached to a building skeleton via a mounting bracket,
The teacher data further includes partition wall information including a numerical value corresponding to the structure of the partition wall of the existing building, and the partition wall information includes the thickness of the ALC board of the existing building and the ALC board. The width of the ALC board, the number of mounting brackets for mounting the ALC board, and the presence or absence of studs or wind resistant beams that support the ALC board,
The damage is cracking or falling of the partition wall, or the impact of these in the building,
The computing device calculates the earthquake damage value of the partition wall corresponding to the target building from the earthquake information, the structural information, the usage information, and the partition wall information corresponding to the target building. earthquake damage prediction system. An earthquake damage prediction system for predicting damage to a target building due to an earthquake,
earthquake information including the magnitude of the earthquake at the epicenter, the depth of the epicenter, the distance from the epicenter to the existing building, and numerical values corresponding to the type of the earthquake;
A numerical value corresponding to the structural form of the existing building, a numerical value corresponding to the frame form of the existing building, a numerical value corresponding to the seismic isolation, vibration damping, or earthquake resistance adopted in the existing building, the number of floors of the existing building, and the Structural information of the existing building including the total floor area of the existing building;
Usage information of the existing building including a numerical value corresponding to the usage of the existing building;
Presence or absence of damage due to the earthquake, the level of the damage, or an earthquake damage value that quantifies the probability of the damage to the non-structural parts of the existing building excluding the foundation and skeleton of the existing building, as teaching data, Machine learning is performed to calculate the earthquake damage value of the non-structural part corresponding to the target building,
an arithmetic device for calculating the earthquake damage value of the non-structural part corresponding to the target building from the earthquake information, the structural information, and the usage information corresponding to the target building ;
the non-structural part is a partition wall, the partition wall is a masonry partition wall,
The teacher data further includes partition wall information including a numerical value corresponding to the structure of the partition wall of the existing building, and the partition wall information includes whether or not there is reinforcing steel in the partition wall of the existing building. is the corresponding numerical value,
The damage is cracking or falling of the partition wall, or the impact of these in the building,
The computing device calculates the earthquake damage value of the partition wall corresponding to the target building from the earthquake information, the structural information, the usage information, and the partition wall information corresponding to the target building. earthquake damage prediction system. In the training data, the surface ground amplification factor of the ground of the existing building, the depth from the ground surface of the construction site obtained by a boring test at the construction site of the existing building, and the soil quality corresponding to the depth are set numerically. ground information, including the value, and the N value in the standard penetration test according to the depth;
further comprising basic information including numerical values corresponding to the basic structure of the existing building,
The computing device calculates the earthquake damage value of the nonstructural part corresponding to the target building from the earthquake information, the structural information, the use information, the ground information, and the basic information corresponding to the target building. The earthquake damage prediction system according to any one of claims 1 to 3, characterized in that:

Images