JP2021189136A - Earthquake motion evaluation model generation method, earthquake motion evaluation model generation device, earthquake motion evaluation method, and earthquake motion evaluation device - Google Patents

Abstract

To provide an earthquake motion evaluation model generation method capable of generating an earthquake motion evaluation model for evaluating and predicting an earthquake motion considered to occur in the future with high accuracy while utilizing increase in data volume and improvement of information processing capacity of a computer to the maximum.SOLUTION: An earthquake motion evaluation model generation method includes: an acquisition step for acquiring a plurality of pieces of learning data 12 composed of a feature quantity and an objective variable from a database 10 associating earthquake motion various characteristic parameters including at least azimuthal characteristics of an earthquake motion with an earthquake motion observation record and storing them when the earthquake motion is observed by using the earthquake motion various characteristic parameters as the feature quantity and an earthquake motion index obtained from the earthquake motion observation record as the objective variable with respect to each combination between a plurality of earthquakes and at least one observation point at which earthquake motions caused by the plurality of earthquakes are observed; and a generation step for generating an earthquake motion evaluation model 13 by learning correlation between the feature quantity and the objective variable by machine learning on the basis of the plurality of pieces of learning data 12.SELECTED DRAWING: Figure 3

Description

The present invention relates to a seismic motion evaluation model generation method, a seismic motion evaluation model generation device, a seismic motion evaluation method, and a seismic motion evaluation device.

Technology for evaluating and predicting the characteristics of earthquake motions that are expected to occur in the future based on various data obtained from past earthquakes (slip-breaking phenomenon of source faults) and earthquake motions (ground motion caused by earthquakes) Because of its usefulness, it is widely used in society, from characteristic analysis and interpretation of earthquake motion to behavior prediction and structural design of buildings and structures, and even earthquake disaster prevention.

Conventionally, some data selected by individual experts with their own judgment is analyzed by a professional method, and then the maximum amplitude (maximum acceleration, maximum speed, maximum displacement) of seismic motion due to future earthquakes. Etc.) and response spectrum (maximum response amplitude of one mass point system when seismic motion is input to one mass point system having a certain natural period and decay constant), etc. Evaluation formulas, seismic motion prediction formulas, etc.) have been developed and utilized (see, for example, Patent Document 1 and Non-Patent Document 1).

Japanese Unexamined Patent Publication No. 2008-39446

Nobuyuki Morikawa and Hiroyuki Fujiwara, A New Ground Motion Prediction Equation for Japan Applicable up to M9 Mega-Earthquake, Journal of Disaster Research, Vol.8, No.5, 2013, p.878-888

Prediction formulas such as practical distance attenuation formulas developed and utilized by previous studies depend on the seismic observation record data used in the studies of individual experts, and are large-scale earthquakes and the largest. Since seismic motion is a rare natural phenomenon, there is an imbalance in the quality and quantity of existing findings reflected in those prediction formulas. In addition, the increase in the amount of data and the improvement of the information processing capacity of computers due to daily accumulation and analysis have the potential to greatly contribute to the improvement of the problem, but each time new data is obtained, it is individual. It takes a huge amount of time and effort for experts to reexamine and re-evaluate the prediction formula, so there are not a few limits.

The present invention has been made in consideration of such circumstances, and while making the best use of the increase in the amount of data and the improvement of the information processing capacity of the computer, the characteristics of the seismic motion due to the earthquake that is considered to occur in the future are enhanced. It is an object of the present invention to provide a seismic motion evaluation model generation method, a seismic motion evaluation model generation device, a seismic motion evaluation method, and a seismic motion evaluation device that can be evaluated and predicted with accuracy.

The present invention solves the above-mentioned problems, and the seismic motion evaluation model generation method according to the embodiment of the present invention is
It is a seismic motion evaluation model generation method that generates a seismic motion evaluation model by machine learning using a computer.
For each combination of multiple earthquakes and at least one observation point where seismic motion due to the multiple earthquakes was observed, the seismic motion characteristic parameters including at least the orientation characteristics of the seismic motion and the seismic motion observation record when the seismic motion was observed are recorded. Acquisition of a plurality of learning data composed of the feature amount and the objective variable from the associated and stored database, with the seismic motion characteristic parameters as feature quantities and the seismic motion index obtained from the seismic motion observation record as the objective variable. Process and
The seismic motion evaluation model as a learned model of the machine learning by learning the correlation between the feature amount and the objective variable by the machine learning based on the plurality of learning data acquired in the acquisition step. Includes a generation process and.

Further, the seismic motion evaluation model generator according to the embodiment of the present invention is
It is a computer and includes a control unit that executes each process included in the above-mentioned seismic motion evaluation model generation method.

Further, the seismic motion evaluation method according to the embodiment of the present invention is
It is a seismic motion evaluation method that evaluates the characteristics of the seismic motion based on the seismic motion evaluation model generated by the seismic motion evaluation model generation method using a computer.
A reception process that accepts the seismic motion characteristic parameters to be predicted, including at least the seismic motion azimuth characteristics.
The prediction target is based on the objective variable output from the seismic motion evaluation model by inputting the seismic motion characteristic parameters of the prediction target received in the reception step into the seismic motion evaluation model as the feature quantities. Includes a prediction step for predicting the seismic motion index corresponding to the seismic motion characteristic parameters.

Further, the seismic motion evaluation device according to the embodiment of the present invention is
It is a computer and includes a control unit that executes each process included in the above-mentioned seismic motion evaluation method.

According to the seismic motion evaluation model generator and the seismic motion evaluation device according to the embodiment of the present invention, a machine is obtained by learning the correlation between a feature amount and an objective variable by machine learning based on a plurality of learning data. Since the seismic motion evaluation model is generated as a trained model for learning, the characteristics of seismic motion due to an earthquake that is expected to occur in the future are evaluated and predicted with high accuracy while making the best use of the increase in the amount of data and the improvement of the information processing capacity of the computer. It is possible to provide a seismic motion evaluation model that can be used.

Further, according to the seismic motion evaluation method according to the embodiment of the present invention and the seismic motion evaluation device, the seismic motion evaluation model generated by the seismic motion evaluation model generation device and the seismic motion evaluation model generation method can be used for individual specialization. It is possible to evaluate and predict the characteristics of seismic motion caused by an earthquake that is expected to occur in the future with high accuracy without depending on the experience of the house.

It is a schematic block diagram which shows an example of the seismic motion evaluation system 1 which concerns on 1st Embodiment of this invention. It is a block diagram which shows an example of the seismic motion evaluation system 1 which concerns on 1st Embodiment of this invention. It is a function explanatory drawing which shows an example of the seismic motion evaluation model generation apparatus 3 and the seismic motion evaluation model generation method which concerns on 1st Embodiment of this invention. It is a data structure diagram which shows an example of a database 10. It is a scatter diagram which shows the correlation of each feature amount of the set of learning data 12 (learning data set A). It is a schematic diagram which shows the outline of the gradient boosting tree. It is a functional explanatory diagram which shows an example of the seismic motion evaluation apparatus 4 and the seismic motion evaluation method which concerns on 1st Embodiment of this invention. An example of the output result when the seismic motion index is expressed as the distance attenuation characteristic is shown. (A) is the maximum acceleration PGA according to the moment magnitude Mw, and (b) is the pseudo velocity response spectrum pSv (1.0s) according to the epicenter direction Λ. ). The degree of influence of each feature (earthquake motion characteristic parameter) on the objective variable (earthquake motion index) is shown. (A) is Kanto model A5, (b) is Kanto model A6, (c) is Kanto model B5, and (d) is. It is a figure which shows each of the Kanto model B6. It is a scatter diagram which shows the correlation between the observation value obtained from the seismic motion observation record, and the evaluation value obtained by the seismic motion evaluation model 13 for each objective variable (earthquake motion index) in the Kanto model A6. The "evaluation value / observation value" ratio of each objective variable (earthquake motion index) is shown in a histogram, (a) is a diagram showing Kanto model A5, and (b) is a diagram showing Kanto model A6. It is a figure which shows the average and variation of the "evaluation value / observation value" ratio of each objective variable (earthquake motion index) about Kanto model A5, A6, B5, B6. It is a function explanatory drawing which shows an example of the seismic motion evaluation model generation apparatus 3 and the seismic motion evaluation model generation method which concerns on 2nd Embodiment of this invention. It is a scatter diagram which shows the correlation of each feature amount of the set of learning data 12, (a) is a learning data set S, and (b) is a learning data set T. It is a functional explanatory diagram which shows an example of the seismic motion evaluation apparatus 4 and the seismic motion evaluation method which concerns on 2nd Embodiment of this invention. An example of the output result when the seismic motion index is expressed as the azimuth characteristic is shown, (a) is a figure which shows the point model S4, and (b) is a figure which shows the point model T4 respectively. It is a figure which shows the degree of influence of each feature quantity (earthquake motion characteristic parameter) with respect to objective variable (earthquake motion index), (a) is a point model S4, and (b) is a point model T4. For each objective variable (earthquake motion index), the correlation between the observed value obtained from the seismic motion observation record and the evaluation value obtained by the seismic motion evaluation model 13 is shown. It is a scatter diagram which shows each. The "evaluation value / observation value" ratio of each objective variable (earthquake motion index) is shown in a histogram, (a) is a diagram showing a point model S4, and (b) is a diagram showing a point model T4. It is a figure which shows the average and variation of the "evaluation value / observation value" ratio of each objective variable (earthquake motion index) about a point model S4, T4. It is a schematic diagram which shows the epicenter characteristic, the propagation characteristic, and the site characteristic of the seismic motion at a plurality of adjacent observation points. It is a figure which shows the 1st observation point group C in which a plurality of observation points c1 to c3 are grouped, and the 2nd observation point group D in which a plurality of observation points d1 to d3 are grouped. It is a function explanatory drawing which shows an example of the seismic motion evaluation model generation apparatus 3 and the seismic motion evaluation model generation method which concerns on 3rd Embodiment of this invention. It is a functional explanatory diagram which shows an example of the seismic motion evaluation apparatus 4 and the seismic motion evaluation method which concerns on 3rd Embodiment of this invention. It is a block diagram which shows an example of the seismic motion evaluation system 1 which concerns on 4th Embodiment of this invention. It is a function explanatory drawing which shows an example of the seismic motion evaluation model generation apparatus 3 and the seismic motion evaluation model generation method which concerns on 4th Embodiment of this invention. It is a scatter diagram which shows the interpolation relation of the 1st learning data 12A and the 2nd learning data 12B. It is a figure which shows the interdependence by the cluster analysis of each feature quantity. Various data formats of the epicenter direction Λ are shown, (a) the definition of each data format is shown, and (b) the degree of influence of each feature amount (earthquake motion characteristic parameter) on the objective variable (earthquake motion index) in each data format. It is a figure which shows.

Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

(First Embodiment)
FIG. 1 is a schematic configuration diagram showing an example of a seismic motion evaluation system 1 according to the first embodiment of the present invention. FIG. 2 is a block diagram showing an example of the seismic motion evaluation system 1 according to the first embodiment of the present invention.

The seismic motion evaluation system 1 collects seismic motion observation records in which seismic motion caused by an earthquake is observed at a plurality of observation points, and provides the observation results to the outside. An analysis data providing device 2B that analyzes the seismic source and scale and provides the analysis results to the outside, and an underground structure data providing device 2C that provides the underground structure parameters related to the underground structure used for evaluation and prediction of seismic motion to the outside, and data provision. Seismic motion based on the seismic motion evaluation model generator 3 that learns the data provided by the devices 2A to 2C by machine learning and generates the seismic motion evaluation model 13 and the seismic motion evaluation model 13 generated by the seismic motion evaluation model generator 3. It is provided with a seismic motion evaluation device 4 for evaluating and predicting the above, and a network 5 for connecting the devices.

The network 5 communicates various data and signals by wireless communication or wired communication, and any communication standard is used. In the present embodiment, the seismic motion evaluation model generation device 3 and the seismic motion evaluation device 4 will be described as separate devices, but may be configured as one device (earthquake motion processing device).

The observation data providing device 2A provides time history waveform data of three components in the north-south direction, the east-west direction, and the vertical direction measured by seismometers (not shown) installed at a plurality of observation points when an earthquake occurs. Each is collected as a seismic motion observation record, and observation data including the seismic motion observation record and additional information such as an observation point position indicating the position of the observation point is provided to the outside. In this embodiment, the observation data is, for example, data provided by the strong motion observation network K-NET of the National Research Institute for Earth Science and Disaster Prevention (hereinafter referred to as "National Research Institute for Earth Science and Disaster Prevention"), and is particularly the first in the Kanto region. The data observed at 138 observation points (see Fig. 1) installed in the six prefectures of Tokyo (Tokyo, Kanagawa, Chiba, Saitama, Ibaraki, Tochigi, Gunma) will be used.

The analysis data providing device 2B analyzes the epicenter and scale of the earthquake based on the seismic motion observation record, and as the analysis result, for example, moment magnitude, Meteorological Agency magnitude, epicenter position, epicenter depth, earthquake type, fault type, and Provide analysis data including seismic source mechanism solutions to the outside. In the present embodiment, the analysis data providing device 2B will be described as using, for example, the data provided by the broadband seismic observation network F-NET of the National Research Institute for Earth Science and Disaster Prevention and the Japan Meteorological Agency.

The underground structure data providing device 2C externally obtains underground structure data including, for example, seismic basement surface depth, engineering basement surface depth, layer thickness, density, seismic wave propagation velocity, Q value, damping constant, etc. as underground structure parameters. To provide to. In the present embodiment, the underground structure data providing device 2C will be described as using, for example, the data provided by the earthquake hazard station J-SHIS of the National Research Institute for Earth Science and Disaster Prevention.

The data providing devices 2A to 2C may provide the provided data regarding the earthquake to the seismic motion evaluation device 4 in real time when an earthquake occurs, or when a data request is received from the seismic motion evaluation device 4. , The provided data regarding the request (provided data regarding a plurality of earthquakes that meet a predetermined condition among the earthquakes that have occurred in the past) may be provided to the seismic motion evaluation device 4. Further, in the present embodiment, the data providing devices 2A to 2C will be described as being three separate devices, but the present invention is not limited to this, and the data providing devices 2A to 2C may be configured as one device or other data providing devices. May be further added.

(About the configuration of the seismic motion evaluation model generator 3 and the process by each part)
The seismic motion evaluation model generation device 3 is based on the provided data (observation data, analysis data, underground structure data) provided by the data providing devices 2A to 2C, for example, a gradient boosting tree, a random forest, a neural network, and a convolutional neural network. By executing a machine learning algorithm such as a network (CNN), a seismic motion evaluation model 13 is generated as a trained model of machine learning.

The seismic motion evaluation model generation device 3 is composed of a general-purpose or dedicated computer, and as shown in FIG. 2, a control composed of a storage unit 30 composed of an HDD, a memory, etc., and a processor such as a CPU, GPU, etc. The unit 31 includes a communication unit 32 which is a communication interface with the network 5, an input unit 33 composed of a keyboard, a mouse, and the like, and a display unit 34 composed of a display, a touch panel, and the like.

The storage unit 30 controls the operations of the database 10 in which the provided data provided by the data providing devices 2A to 2C is registered and updated, the seismic motion evaluation model 13 which is a trained model, and the seismic motion evaluation model generation device 3. The seismic motion evaluation model generation program 300 that realizes the seismic motion evaluation model generation method is stored.

The control unit 31 functions as a DB management unit 310, an acquisition unit 311 and a generation unit 312 by executing the seismic motion evaluation model generation program 300. The details of the functions of each part will be described later.

FIG. 3 is a functional explanatory diagram showing an example of the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method according to the first embodiment of the present invention.

(About the database management process by the DB management unit 310 and the database 10)
The DB management unit 310 manages the database 10 based on the provided data (observation data, analysis data, underground structure data) provided by the data providing devices 2A to 2C. Specifically, the DB management unit 310 said every time an earthquake occurs, the seismic motion due to the earthquake is observed at at least one observation point, and the data providing devices 2A to 2C provide the provided data regarding the earthquake. For each combination of the earthquake and at least one observation point where the seismic motion due to the earthquake is observed, the seismic motion characteristic parameters and the seismic motion observation record at the time when the seismic motion is observed are registered in the database 10.

FIG. 4 is a data structure diagram showing an example of the database 10. The database 10 contains seismic motion characteristic parameters and seismic motion observation records when the seismic motion was observed for each combination of a plurality of earthquakes that occurred in the past and at least one observation point where the seismic motion caused by the multiple earthquakes was observed. A plurality of seismic motion data 11 in a state associated with the above are registered and stored.

The seismic motion characteristic parameters are various parameters that describe the various characteristics of the seismic motion, and the seismic motion characteristics include, for example, the source characteristic and the propagation characteristic of the seismic motion, and the site characteristic, the directional characteristic, and the seismic motion observation characteristic. Including further. In the present embodiment, the seismic motion characteristic parameters are those in which the provided data (observation data, analysis data, underground structure data) provided by the data providing devices 2A to 2C are classified and recorded according to the above-mentioned seismic motion characteristics. Is. The epicenter characteristics, propagation characteristics, site characteristics, directional characteristics, and seismic motion observation characteristics included in the seismic motion characteristic parameters will be described below.

The epicenter characteristics are, for example, magnitude (moment magnitude Mw, Meteorological Agency magnitude, etc.), seismic center position (latitude lat_eq, longitude long_eq), epicenter depth H, earthquake type Type (inland crustal earthquake, plate boundary earthquake, slab earthquake), Fault type Mech (normal fault / reverse fault / strike-slip fault), seismic source mechanism solution (strike 1, tilt angle dip1, slip angle lake1), and conjugate solution of seismic source mechanism solution (strike2, tilt angle dip2, slip angle racke2) Etc. at least one. The epicenter characteristics according to this embodiment are the moment magnitude Mw, the epicenter position (latitude lat_eq, longitude lon_eq), the epicenter depth H, the earthquake type Type, the fault type Mech, the epicenter mechanism solution (Strike1, dip1, rake1), and the epicenter. It is a conjugate solution of the mechanism solution (Strike2, dip2, lake2).

The propagation characteristic is, for example, at least one of the epicentral distance X, the shortest fault distance, and the epicenter distance. The propagation characteristic according to the present embodiment is the epicentral distance X, which is calculated as the distance between the observation point position and the epicenter position.

The site characteristic is, for example, at least one of the observation point position (latitude lat_site, longitude long_site), seismic base plane depth, engineering base plane depth, layer thickness, density, seismic wave propagation velocity, Q value, and attenuation constant. Is. The site characteristics according to this embodiment are the observation point position (latitude lat_site, longitude long_site), the top layer S wave velocity VS1, the surface layer 10 m average S wave velocity AVS10, the surface layer 30 m average S wave velocity AVS30, microtopography classification JCODE, S. Wave velocity 700m / s layer upper surface depth D7, S wave velocity 1400m / s layer upper surface depth D17, S wave velocity 2100m / s layer upper surface depth D24, and seismic basement surface depth D28 (= S wave velocity 2700m / s layer upper surface depth). If the surface layer 30m average S wave velocity AVS30 cannot be obtained from the underground structure data of the observation point, the surface layer 30m average S wave velocity AVS30 of the 250m mesh of the Seismic Hazard Station J-SHIS of the Disaster Prevention Research Institute is used instead. When the 10 m average S wave velocity AVS10 cannot be obtained from the underground structure data of the observation point, the surface layer 30 m average S wave velocity AVS30 is used instead. In addition, the seismic base surface depth D28 is the bottom depth of the 28th layer of the deep ground model of the mesh including the target observation point position published at the Seismic Hazard Station J-SHIS of the National Research Institute for Earth Science and Disaster Prevention (seismic base in the same model). The P wave velocity is 5000 m / s and the S wave velocity is 2700 m / s, which is equal to the depth of the upper surface of the 29th layer).

The azimuth characteristic is, for example, the epicenter azimuth Λ indicating the azimuth in which the epicenter is located with respect to the observation point. Therefore, the epicenter direction Λ is calculated as the direction in which the epicenter position exists with reference to the observation point position. At that time, the epicenter direction Λ is determined clockwise with true north as 0 °, and the amount is discontinuous with true north as the boundary. Therefore, the directional characteristics according to the present embodiment are sin Λ and cos Λ representing the epicenter direction Λ. Use a pair.

The seismic motion observation characteristic is, for example, a seismic motion direction component Comp indicating that the time history waveform data recorded as the seismic motion observation record is in the north-south direction, the east-west direction, or the vertical direction.

The seismic motion observation record is, for example, a time history waveform of acceleration, velocity, and displacement of three components in the north-south direction, the east-west direction, and the vertical direction. Various seismic motion indexes can be obtained by evaluating and analyzing the seismic motion observation record by a predetermined evaluation method. The seismic motion observation record according to this embodiment is the observation point 138 points installed in the Kanto region's six prefectures (Tokyo, Kanagawa, Chiba, Saitama, Ibaraki, Tochigi, Gunma) in the strong motion observation network K-NET (Fig. 1). It is a time history waveform for two horizontal components (north-south direction and east-west direction) observed in (see).

The seismic motion index includes at least one of the amplitude characteristic, the periodic characteristic, and the temporal characteristic of the seismic motion.

The amplitude characteristic is at least one of the maximum acceleration PGA, maximum velocity, and maximum displacement of the seismic motion obtained from the seismic motion observation record (acceleration, velocity, and time history waveform of displacement). The amplitude characteristic according to this embodiment is the maximum acceleration PGA.

The periodic characteristic is a response value for at least one period in a response spectrum or a Fourier spectrum obtained from seismic motion observation records (acceleration, velocity, and time history waveform of displacement). The response spectrum is, for example, an acceleration response spectrum with respect to a predetermined attenuation constant (for example, 5%), a pseudo velocity response spectrum pSv, a velocity response spectrum, a displacement response spectrum, and the like. The Fourier spectrum is, for example, an acceleration Fourier spectrum, a velocity Fourier spectrum, a displacement Fourier spectrum, or the like. The periodic characteristics according to the present embodiment are pseudo-velocity response spectra pSv (0.1s) and pSv (0) having a decay constant of 5% in each period of 0.1 second, 0.5 second, 1 second, 3 seconds, and 5 seconds. .5s), pSv (1.0s), pSv (3.0s), pSv (5.0s).

The temporal characteristic is, for example, the response duration for at least one period in the response duration spectrum obtained from the seismic motion observation record (acceleration, velocity, and displacement time history waveform). The response duration spectrum is, for example, an acceleration response duration spectrum, a velocity response duration spectrum TSv, a displacement response duration spectrum, or the like with respect to a predetermined attenuation constant (for example, 5%). The temporal characteristics according to the present embodiment are the velocity response duration spectrum TSv (0.1s), TSv ( 0.5s), TSv (1.0s), TSv (3.0s), and TSv (5.0s). The parameters that define the start and end of the response duration are p1 = 0.03 and p2 = 0.95.

(About the acquisition process by the acquisition unit 311 and the learning data 12)
As shown in FIG. 3, the acquisition unit 311 uses the seismic motion characteristic parameters as feature variables from the plurality of seismic motion data 11 registered in the database 10, and the seismic motion index obtained from the seismic motion observation record as the objective variable. Acquire a plurality of learning data 12 composed of objective variables. The learning data 12 is data used as learning data (training data), verification data, and test data in supervised learning. Further, a set of learning data 12 composed of a plurality of learning data 12 is called a learning data set.

At that time, the acquisition unit 311 selects the seismic motion data 11 based on the earthquake that satisfies at least one of the predetermined seismic condition and the predetermined seismic motion condition as the learning data 12. It is preferable that the seismic conditions and the seismic motion conditions are set so that the seismic motion having a large amplitude to some extent is distributed in a well-balanced manner in the metropolitan area, for example.

Predetermined seismic conditions include the first to sixth seismic conditions shown below. For example, the first earthquake condition is that the earthquake occurred during the period from 1996 to January 15, 2019, and the maximum seismic intensity is 4 or more and the seismic intensity is 2 or more in Chiyoda-ku, Tokyo. The second seismic condition is that the deviation of the epicenter position and the deviation of the earthquake occurrence area are below the predetermined threshold, after excluding the earthquakes in the Kanto region where the number of points where the seismic intensity is recorded is very small (about one digit). And. Regarding the earthquake scale (Major magnitude Mj of the Meteorological Agency), it should be balanced when separated into three stages of 3.0 <Mj ≤ 5.0, 5.0 <Mj ≤ 5.5, 5.5 <Mj. The third earthquake condition. When multiple earthquakes occur in the same earthquake occurrence area, the fourth earthquake condition is to prevent the deviation of the epicenter depth from occurring. The fifth seismic condition is that the moment magnitude Mw is obtained by the wideband seismic observation network F-net. The sixth earthquake condition is to exclude earthquakes with extremely wide fault planes (for example, the 2011 off the Pacific coast of Tohoku Earthquake (Mj9.0)). A total of 74 earthquakes fall under the above-mentioned first to sixth earthquake conditions.

The predetermined seismic motion conditions include the first to fifth seismic motion conditions shown below. For example, the first seismic motion condition is that the maximum acceleration of the three-component combined value when the three-component acceleration is combined is 1 cm / s ^{2 or more.} The second seismic motion condition is to exclude those with small amplitude and large noise (small S / N ratio) in the seismic motion observation record. The third seismic motion condition is to exclude those whose recording time is insufficient (head shortage, tail shortage, both shortage) in the seismic motion observation record. The fourth seismic motion condition is to exclude those in which it is difficult to analyze the records due to the mixture of different earthquakes and different events in the seismic motion observation records. Regarding the response spectrum and the Fourier spectrum in the seismic motion index, the fifth seismic motion condition is to exclude those with large noise mainly in the band having a period of 0.05 to 10 seconds.

The learning unit 311 selected by the acquisition unit 311 by selecting the seismic motion data 11 satisfying the first seismic motion condition for the seismic motion caused by a total of 74 earthquakes satisfying the first to sixth seismic conditions. The set of data 12 (hereinafter referred to as “learning data set A”) consists of seismic motion data 11 having a total of 14104 (7052 records × horizontal two components). Further, the acquisition unit 311 selects the seismic motion data 11 satisfying the first to fifth seismic motion conditions for the seismic motion caused by a total of 74 earthquakes satisfying the first to sixth earthquake motion conditions, thereby causing the acquisition unit 311. The set of training data 12 selected by (hereinafter referred to as “learning data set B”) consists of seismic motion data 11 having a total of 11488 (5744 records × horizontal two components).

In the following, the feature quantities constituting the learning data 12 according to the present embodiment are the moment magnitude Mw, the epicenter depth H, the epicentral distance X, and the epicenter direction Λ (of the 25 types of seismic motion characteristic parameters shown in FIG. 4). It will be described as six types of seismic ground motion characteristic parameters: (pair of sinΛ and cosΛ), surface layer 30 m average S wave velocity AVS30, and seismic basement surface depth (S wave velocity 2700 m / s layer top surface depth) D28.

Further, the objective variables constituting the learning data 12 according to the present embodiment are 11 types of seismic motion indexes shown in FIG. 4, that is, maximum acceleration PGA, pseudo-velocity response spectrum pSv (0.1s), pSv (0.5s). ), PSv (1.0s), pSv (3.0s), pSv (5.0s), velocity response duration spectrum TSv (0.1s), TSv (0.5s), TSv (1.0s), TSv (3.0s), TSv (5.0s) will be described.

FIG. 5 is a scatter diagram showing the correlation of each feature amount of the set of learning data 12 (learning data set A). The characteristics of the learning data set A are that there are few records of small-scale earthquakes in the distance, there are no records of large-scale earthquakes in the short distance, most of them are distributed shallower than the epicenter depth of 100 km, and the Kanto region. The epicenter direction Λ for each observation point is dispersed in all directions, but there are many cases in the northeast direction (earthquake center direction Λ ≈ 45 °).

(About the generation process by the generation unit 312 and the seismic motion evaluation model 13)
As shown in FIG. 3, the generation unit 312 learns the correlation between the feature amount and the objective variable by machine learning based on the plurality of learning data 12 acquired by the acquisition unit 311. The seismic motion evaluation model 13 is generated and stored in the storage unit 30. In this embodiment, a case where a gradient boosting decision tree is used as a machine learning algorithm in machine learning will be described.

FIG. 6 is a schematic diagram showing an outline of a gradient boosting tree. The gradient boosting tree is a learning device that combines gradient boosting and a decision tree. Gradient boosting is a method of constructing a strong learner (high-performance machine learning model) by combining multiple weak learners (low-performance machine learning model). Decision trees are a method of generating machine learning models that can be classified and regressed by performing conditional branching using the branching structure of trees. A gradient boosting tree that combines these two methods is a method of combining a plurality of weak learners generated by a decision tree by gradient boosting.

In the present embodiment, among the 11 types of seismic motion indexes, it is considered that the number of data for the maximum acceleration PGA and the pseudo-velocity response spectrum pSv decreases sharply as the amplitude increases. Therefore, in the present embodiment, in order to reduce the bias generated in the distribution of the objective variable, common logarithms (log ₁₀ PGA and log ₁₀ pSv) are used as the objective variable data for the maximum acceleration PGA and the pseudo-velocity response spectrum pSv, respectively. It was to be. Further, as a loss function in the gradient boosting tree, the least squares method (normal distribution) is applied to the maximum acceleration PGA and the pseudo velocity response spectrum pSv, and the Poisson distribution is applied to the velocity response duration spectrum TSv. ..

Further, in the present embodiment, when the generation unit 312 generates the seismic motion evaluation model 13 using the gradient boosting tree, the seismic motion evaluation model 13 is generated for each objective variable (11 types of seismic motion indexes). That is, the first seismic motion evaluation model 13A in which the correlation between the feature amount (6 types of seismic motion characteristic parameters) and the objective variable (maximum acceleration PGA = 1st type seismic motion index) is learned, and the feature amount (6 types). A third seismic motion such as the second seismic motion evaluation model 13B in which the correlation between the seismic motion characteristic parameters) and the objective variable (pseudo-velocity response spectrum pSv (0.1s) = second type seismic motion index) is learned. The evaluation models 13C to the eleventh seismic motion evaluation model 13K were generated respectively, and a total of 11 seismic motion evaluation models 13A to 13K were generated. Hereinafter, it is assumed that the seismic motion evaluation model 13 includes a total of 11 seismic motion evaluation models 13A to 13K.

(About the configuration of the seismic motion evaluation device 4 and the process by each part)
The seismic motion evaluation device 4 evaluates and predicts seismic motion based on the seismic motion evaluation model 13 generated by the seismic motion evaluation model generation device 3, and outputs the result to an output medium such as a display medium or a paper medium.

Like the seismic motion evaluation model generator 3, the seismic motion evaluation device 4 is composed of a general-purpose or dedicated computer, and as shown in FIG. 2, a storage unit 40 composed of an HDD, a memory, and the like, a CPU, and a GPU. A control unit 41 composed of a processor such as the above, a communication unit 42 which is a communication interface with the network 5, an input unit 43 composed of a keyboard, a mouse, etc., and a display unit 44 composed of a display, a touch panel, etc. To prepare for.

The storage unit 40 stores a seismic motion evaluation model 13 generated as a trained model by the seismic motion evaluation model generator 3 and a seismic motion evaluation program 400 that controls the operation of the seismic motion evaluation device 4 to realize a seismic motion evaluation method. There is.

By executing the seismic motion evaluation program 400, the control unit 41 functions as a reception unit 410, a prediction unit 411, and an output processing unit 412. The details of the functions of each part will be described later.

FIG. 7 is a functional explanatory diagram showing an example of the seismic motion evaluation device 4 and the seismic motion evaluation method according to the first embodiment of the present invention.

(Regarding the reception process by the reception department 410)
The reception unit 410 receives various seismic motion characteristic parameters to be predicted. Specifically, the reception unit 410 has, for example, the moment magnitude Mw, the epicenter position, and the epicenter depth H of the earthquake assumed as the prediction target by the user of the seismic motion evaluation device 4 (hereinafter referred to as “assumed earthquake”). Is received via the input unit 33, and the predicted point position indicating the position of the predicted point for which it is desired to predict how much seismic motion due to the assumed earthquake will occur is also received via the input unit 33. The predicted point position may be any position or may be the same as the observation point position. Further, the predicted point positions may be plural, and may be, for example, each grid point at a predetermined grid spacing (for example, 5 km spacing).

Then, the reception unit 410 calculates the epicentral distance X and the epicenter direction Λ based on the epicenter position of the assumed earthquake and the predicted point position of the seismic motion, and also uses the underground structure data provided by the underground structure data providing device 2C. Based on this, the surface layer 30 m average S wave velocity AVS30 and the seismic basement surface depth (S wave velocity 2700 m / s layer top surface depth) D28 with respect to the predicted point position of the seismic motion are acquired. As a result, the reception unit 410 can use the seismic motion characteristic parameters to be predicted (moment magnitude Mw of the assumed earthquake, epicenter depth H of the assumed earthquake, source distance X between the source of the assumed earthquake and the predicted point position, and predicted point position). The epicenter direction Λ indicating the direction in which the epicenter position of the assumed earthquake exists, the surface layer 30 m average S wave velocity AVS30 at the predicted point position, and the seismic base plane depth D28) at the predicted point position are accepted.

(About the prediction process by the prediction unit 411)
The prediction unit 411 uses the seismic motion characteristic parameters of the prediction target received by the reception unit 410 as feature quantities, and the seismic motion evaluation model 13 (six types of seismic motion characteristic parameters as feature quantities, and 11 types of seismic motion indexes as objective variables. By inputting to the trained model in which the correlation between the two is learned), the seismic motion index corresponding to the seismic motion characteristic parameters to be predicted is predicted based on the objective variable output from the seismic motion evaluation model 13. At that time, when the reception unit 410 accepts a plurality of assumed earthquakes or a plurality of predicted point positions to receive a plurality of seismic motion characteristic parameters as prediction targets, the prediction unit 411 receives a plurality of seismic motions. By inputting each of the various characteristic parameters into the seismic motion evaluation model 13, the seismic motion index corresponding to each of the plurality of seismic motion characteristic parameters is predicted.

In the present embodiment, since the seismic motion evaluation model 13 includes a total of 11 seismic motion evaluation models 13A to 13K generated for each objective variable (11 types of seismic motion indexes), the prediction unit 411 is used to predict various seismic motions. The characteristic parameters are input to a total of 11 seismic motion evaluation models 13A to 13K as feature quantities, and the prediction target is for each objective variable (11 types of seismic motion indicators) output from the total of 11 seismic motion evaluation models 13A to 13K. Predict the seismic motion index corresponding to the seismic motion characteristic parameters of. That is, the maximum acceleration PGA is predicted using the first seismic motion evaluation model 13A in which the correlation between the feature quantity (six types of seismic motion characteristic parameters) and the maximum acceleration PGA (= the first type of seismic motion index) is learned. Using the second seismic motion evaluation model 13B, which learned the correlation between the feature quantity (6 types of seismic motion characteristic parameters) and the pseudo-velocity response spectrum pSv (0.1s) (= the second type of seismic motion index). By predicting each objective variable from the third seismic motion evaluation model 13C to the eleventh seismic motion evaluation model 13K, such as predicting the pseudo-velocity response spectrum pSv (0.1s), a total of 11 pieces. Eleven kinds of seismic motion indexes were predicted using the seismic motion evaluation models 13A to 13K.

(About the output processing process by the output processing unit 412)
The output processing unit 412 outputs the seismic motion index predicted by the prediction unit 411 to a visible output medium. For example, when the output medium is a display medium such as the display unit 44, the output processing unit 412 generates display data (output data) for display on the display medium and displays and outputs the display data on the display medium. When the output medium is a paper medium, the output processing unit 412 generates print data (output data) for printing on the paper medium and prints and outputs the print data on the paper medium. The output processing unit 412 may output the output data to, for example, a seismic motion prediction map creation system, a hazard map creation system, or the like, or may output the output data to a disaster prevention system such as a public facility, a building, or a factory. You may do so.

For example, the output processing unit 412 has a plurality of seismic motions such that the seismic motion characteristic parameters of the prediction target received by the reception unit 410 have, for example, each grid point as the predicted point position for one assumed earthquake. When it is a characteristic parameter, for example, a contour diagram or a contour diagram so that the values of a plurality of seismic motion indexes at each predicted point position predicted by the prediction unit 411 based on the plurality of seismic motion characteristic parameters are superimposed on the map. , Output to the output medium as a color-coded mesh diagram.

Further, when the seismic motion characteristic parameters of the prediction target received by the reception unit 410 are a plurality of seismic motion characteristic parameters including a plurality of different directional characteristics centered on the predicted point position, the output processing unit 412 has the plurality. The values of multiple seismic motion indexes in each direction predicted by the prediction unit 411 based on the various seismic motion characteristic parameters of the above are expressed as the distances from the reference point in the output medium, and the multiple direction characteristics are centered on the reference point. By assigning to each direction, a plurality of seismic motion indexes are output to the output medium.

Further, when the seismic motion characteristic parameters of the prediction target received by the reception unit 410 are a plurality of seismic motion characteristic parameters including different epicenter distances, the output processing unit 412 is based on the plurality of seismic motion characteristic parameters. By expressing the values of the plurality of seismic motion indexes predicted by the prediction unit 411 as the distance attenuation characteristics, as shown in FIG. 8 (details will be described later), the plurality of seismic motion indexes are output to the output medium.

FIG. 8 shows an example of the output result when the seismic motion index is expressed as the distance attenuation characteristic, (a) is the maximum acceleration PGA according to the moment magnitude Mw, and (b) is the pseudo velocity response spectrum pSv according to the epicenter direction Λ. It is a figure which shows (1.0 s).

The output result shown in FIG. 8A is output when the assumed earthquake is an earthquake whose epicenter is in the direction of Ibaraki and the predicted point position is "SIT006 (Chichibu)" (see FIG. 1). Yes, it is assumed that the rock site will shake due to an interplate earthquake. In addition, the average of the existing distance damping formulas based on Non-Patent Document 1 is also output for reference, but in general, the seismic motion index is based on the seismic motion evaluation model 13 as in the past distance damping formulas. Can be said to have been predicted.

The output results shown in FIG. 8B show that the assumed earthquakes are interplate earthquakes in the northeast, south, and southwest directions (Λ = 45 °, 160 °, 230 °) and northwest directions (Λ = 320 °) when viewed from the city center. ) Is output in the case of an inland earthquake. Comparing these four assumed earthquakes, even if the moment magnitude Mw and the epicentral distance X are the same, there is a difference in the magnitude of the pseudo-velocity response spectrum pSv (1.0s), and the south direction is the interplate earthquake. The smallest, the magnitude relationship between the northeastern direction and the southwestern direction is switched when the epicentral distance X is about 50 km. In addition, in the northwestern inland earthquake, the slope of the distance attenuation is relatively small, and when the epicentral distance X is less than 100 km, the magnitude of the pseudo-velocity response spectrum pSv (1.0 s) is remarkable compared to other assumed earthquakes. It is getting smaller. Such differences are the result of the differences in the regional characteristics and propagation characteristics of the epicenter characteristics being reflected in the seismic motion evaluation model 13, and by considering the epicenter direction Λ, the detailed regional characteristics are reflected in the seismic motion evaluation model 13. It is considered to be the result of this.

(Characteristics of seismic motion evaluation model 13)
Next, the characteristics of the seismic motion evaluation model 13 will be described while comparing a plurality of seismic motion evaluation models 13 generated by using different learning data 12 by the seismic motion evaluation model generation device 3. At that time, for the plurality of seismic motion data 11 associated with the seismic motion characteristic parameters and the seismic motion observation record, the seismic motion index obtained from the seismic motion observation record is set as the "observed value", and the seismic motion characteristic parameters are input to the seismic motion evaluation model 13. The characteristics of the seismic motion evaluation model 13 will be described using the “observed value” and the “evaluation value” with the seismic motion index obtained by the above as the “evaluation value”.

Here, it is assumed that the "Kanto model A" is generated by machine learning using the learning data set A, and the "Kanto model A6" considering all the above six types as feature quantities and the seismic center direction Λ (sinΛ, cosΛ). "Kanto model A5" considering only 5 types excluding the above was generated and compared. In addition, it is assumed that the "Kanto model B" is generated by machine learning using the learning data set B, and the "Kanto model B6" considering all the above six types as feature quantities and the seismic center direction Λ (sinΛ, cosΛ) are used. "Kanto model B5" considering only 5 types except 5 was generated and compared.

(Regarding the degree of influence of each feature on the objective variable)
FIG. 9 shows the degree of influence of each feature amount (earthquake motion characteristic parameters) on the objective variable (earthquake motion index), (a) is Kanto model A5, (b) is Kanto model A6, and (c) is Kanto model B5. (D) is a diagram showing each of the Kanto model B6.

For the degree of influence of each feature amount shown in FIG. 9, one feature amount is selected from the training data 12 (training data set A or training data set B) used for machine learning, and only the data string is shuffled. It shows the degree of deterioration of the evaluation accuracy when the objective variable is evaluated by replacing it with a random data string and leaving the data strings of other features as they are. Therefore, when the evaluation accuracy is significantly deteriorated, the feature amount is important, and conversely, when the evaluation accuracy does not change, the feature amount is considered to have a small influence on the evaluation.

Comparing the Kanto model A5 shown in FIG. 9A and the Kanto model A6 shown in FIG. 9B, the influence of the epicentral distance X is dominant in the short period and the influence of the moment magnitude Mw is dominant in the long period. The degree of influence of the moment magnitude Mw increased with the period as a whole except for the velocity response duration spectrum TSv (0.1s), and the degree of influence of the epicentral distance X decreased with the period. The degree of influence of the epicenter depth H is slightly larger on the short cycle side, and the degree of influence of the surface layer 30 m average S wave velocity AVS30 is relatively larger in the velocity response duration spectrum TSv (1.0 s), and the depth of the seismic base surface. The degree of influence of D28 became large when the period of the velocity response duration spectrum TSv was 1 second or more. The degree of influence of the epicenter direction Λ is comparable to the epicenter depth H, the seismic base surface depth D28, and the surface layer 30 m average S wave velocity AVS30 when considering sinΛ and cosΛ together, and the period is relatively 1 second or less. It became large. Overall, it seems that the characteristics differ between the short-period side and the long-period side with a period of around 1 second as the boundary.

Comparing the Kanto model B5 shown in FIG. 9 (c) with the Kanto model B6 shown in FIG. 9 (d), the effect on the velocity response duration spectrum TSv became easier to understand than the result of the prototype Kanto model A. The degree of influence of the moment magnitude Mw increased with the period, and the degree of influence of the epicentral distance X and the source depth H decreased with the period. The degree of influence of the surface layer 30 m average S wave velocity AVS30 is relatively large in a period of 1 second, the degree of influence of the seismic base surface depth D28 is large in a period of 1 second or more, and the degree of influence of the epicenter direction Λ is relatively large. Became larger with a period of 1 second or less.

(About the prediction accuracy of the objective variable)
FIG. 10 is a scatter diagram showing the correlation between the observed value obtained from the seismic motion observation record and the evaluation value obtained by the seismic motion evaluation model 13 for each objective variable (earthquake motion index) in the Kanto model A6.

In FIG. 10, the maximum acceleration PGA, the pseudo velocity response spectra pSv (1.0s), pSv (3.0s), the velocity response duration spectra TSv (1.0s), and TSv (3.0s) are compared as objective variables. did. In both cases, the horizontal axis is the observed value and the vertical axis is the evaluation value. Of the training data 12, the training data used for machine learning is 64% of the total data, the verification data used for model verification is 36% of the total data, and the evaluation value is equal to the observed value. The case is shown by a solid line, and the case where the observed value is doubled or halved is shown by a broken line. The observed values were evaluated and modeled very well as a whole, and most of the evaluated values were within half of the observed values. This evaluation accuracy is considered to be high even in light of the conventional knowledge about seismic motion.

11A and 11B are histograms showing the “evaluation value / observation value” ratio of each objective variable (earthquake motion index), where (a) shows the Kanto model A5 and (b) shows the Kanto model A6. It was found that the value of the "evaluation value / observed value" ratio had a well-organized distribution centered on 1, and the variation of the velocity response duration spectrum TSv was smaller than that of the maximum acceleration PGA and the pseudo velocity response spectrum pSv.

・
FIG. 12 is a diagram showing the average μ and the variation σ of the “evaluation value / observation value” ratio of each objective variable (earthquake motion index) for the Kanto models A5, A6, B5, and B6.

The average of the "evaluation / observation" ratio is approximately 1, and the common logarithm standard deviation of the "evaluation / observation" ratio is 0.17 to 0. For the maximum acceleration PGA and the pseudo-velocity response spectrum pSv. 20. The velocity response duration spectrum TSv was 0.10 to 0.12. This evaluation accuracy is high even in the light of conventional knowledge about seismic motion. When the epicenter direction Λ is taken into consideration as a feature (Kanto model A5 → A6), the average and variation of each objective variable are slightly improved, and it is not a little important to consider the difference in seismic motion characteristics depending on the epicenter direction Λ. I was also able to confirm that. Further, when the training data set B was used instead of the training data set A (Kanto model A6 → B6), the average and variation of the velocity response duration spectrum TSv were slightly improved.

As described above, according to the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method according to the present embodiment, it is considered that they will occur in the future while making the best use of the increase in the amount of data and the improvement of the information processing capacity of the computer. It is possible to provide a seismic motion evaluation model 13 capable of evaluating and predicting the characteristics of seismic motion due to an earthquake with high accuracy.

Further, according to the seismic motion evaluation device 4 and the seismic motion evaluation method according to the present embodiment, by using the seismic motion evaluation model generation device 3 and the seismic motion evaluation model 13 generated by the seismic motion evaluation model generation method, individual experts can use the seismic motion evaluation device 4. It is possible to evaluate and predict the characteristics of seismic motion caused by an earthquake that is expected to occur in the future with high accuracy without depending on experience.

Furthermore, when the seismic motion characteristic parameters include at least the seismic motion azimuth characteristics, the correlation between the seismic motion azimuth characteristics and the seismic motion index is learned by machine learning, so that the underground structure around the site differs for each direction. Since the influence on the seismic motion is reflected in the seismic motion evaluation model 13, it is possible to provide the seismic motion evaluation model 13 capable of evaluating and predicting the characteristics of the seismic motion with high accuracy.

In addition, when the seismic motion index includes at least the temporal characteristics of the seismic motion, the correlation between the seismic motion characteristic parameters and the temporal characteristics of the seismic motion is learned by machine learning, and the temporal characteristics of the seismic motion due to the earthquake that is considered to occur in the future. It is possible to provide a seismic motion evaluation model 13 capable of evaluating and predicting the temporal characteristics of seismic motion with high accuracy as one of the characteristics of seismic motion without defining a prediction formula for evaluating and predicting.

(Second embodiment)
In the seismic motion evaluation system 1 according to the first embodiment, the seismic motion evaluation model generator 3 uses the observation data of 138 observation points installed in one metropolitan area and six prefectures in the Kanto region to obtain one seismic motion evaluation model 13 (for example, for example. Kanto models A5, A6, B5, B6) are generated, and the seismic motion evaluation device 4 evaluates and predicts seismic motion in the Kanto region using one seismic motion evaluation model 13. On the other hand, in the seismic motion evaluation system 1 according to the second embodiment, the seismic motion evaluation model generation device 3 generates the seismic motion evaluation model 13 for each observation point, and the seismic motion evaluation device 4 generates the seismic motion evaluation model for each observation point. It differs in that it evaluates and predicts the seismic motion at each observation point using No. 13. Since other basic configurations and operations are the same as those of the first embodiment, the differences between the two will be mainly described below.

(About seismic motion evaluation model generator 3)
In this embodiment, the seismic motion evaluation model generator 3 is an example of a point where the basement is shallow and the surface layer is hard from the observation points of the strong motion observation network K-NET of the National Research Institute for Earth Science and Disaster Prevention "SIT006 (Chichibu)" (FIG. 1). (See) and "TKY028 (Etchujima)" (see FIG. 1), which is an example of a point where the basement is deep and the surface layer is soft, will be described assuming that the seismic motion evaluation model 13 is generated. Here, the seismic motion evaluation model 13 in SIT006 (Chichibu) is referred to as "point model S", and the seismic motion evaluation model 13 in TKY028 (Etsunakajima) is referred to as "point model T". Let them be "learning data set S" and "learning data set T", respectively.

FIG. 13 is a functional explanatory diagram showing an example of the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method according to the second embodiment of the present invention.

The acquisition unit 311 is an earthquake that occurred during the period from 1996 to May 31, 2019 under predetermined seismic conditions from the plurality of seismic motion data 11 registered in the database 10, and is SIT006 (Chichibu) or TKY028 ( Based on the earthquake that satisfies the observation of seismic motion in (Etsunakajima), the maximum acceleration of the three-component composite value is 1 cm / s ² or more (first seismic motion condition) as the seismic motion that satisfies the predetermined seismic motion condition. And, the seismic motion data 11 based on the seismic motion satisfying that the moment magnitude Mw is obtained by the broadband seismic observation network F-net (fifth seismic motion condition) is used as the training data 12 (learning data sets S and T). select. As a result, the learning data set S is composed of a total of 1468 (734 records × horizontal two components) seismic motion data 11, and the learning data set T is composed of a total of 1314 (657 earthquakes × 2 components) seismic motion data 11.

At that time, since the point models S and T are each point-specific seismic motion evaluation model 13, the feature quantities constituting the learning data 12 are the surface layer common to each point among the above six types of seismic motion characteristic parameters. There are four types of seismic ground motion characteristic parameters: moment magnitude Mw, epicenter depth H, epicentral distance X, and epicenter direction Λ, excluding 30 m average S wave velocity AVS30 and seismic base plane depth D28. The objective variable constituting the learning data 12 is the above 11 types of seismic motion indexes, as in the first embodiment.

FIG. 14 is a scatter diagram showing the correlation of each feature amount of the set of training data 12, where (a) shows the learning data set S and (b) the learning data set T. The training data sets S and T have less data on short-range earthquakes than the training data set A shown in FIG. 5, but qualitatively have the same tendency as the training data set A except for that point. be.

As shown in FIG. 13, the generation unit 312 has feature quantities (four types of seismic motion characteristic parameters) based on a plurality of learning data 12 (learning data sets S and T) acquired by the acquisition unit 311. By learning the correlation between the objective variables (11 types of seismic motion indexes) by machine learning, seismic motion evaluation models 13 (point models S4 and T4) are generated for each observation point.

(About seismic motion evaluation device 4)
FIG. 15 is a functional explanatory diagram showing an example of the seismic motion evaluation device 4 and the seismic motion evaluation method according to the second embodiment of the present invention.

The reception unit 410 determines the seismic motion characteristic parameters (moment magnitude Mw of the assumed earthquake, the epicenter depth H of the assumed earthquake, the epicenter distance X between the epicenter of the assumed earthquake and the predicted point position, and the predicted point position). As a reference, the epicenter direction Λ) indicating the direction in which the epicenter position of the assumed earthquake exists is accepted. In this embodiment, the predicted point position is SIT006 (Chichibu) or TKY028 (Etchujima).

The prediction unit 411 uses the seismic motion characteristic parameters of the prediction target received by the reception unit 410 as feature quantities, and inputs the seismic motion evaluation model 13 (point models S4, T4) according to the predicted point position to evaluate the seismic motion. Based on the objective variable output from the model 13, the seismic motion index corresponding to the seismic motion characteristic parameters to be predicted is predicted. At that time, the prediction unit 411 uses the point model S4 when the predicted point position is SIT006 (Chichibu), and uses the point model T4 when the predicted point position is TKY028 (Etchujima).

The output processing unit 412 outputs the seismic motion index predicted by the prediction unit 411 to a visible output medium. For example, when the seismic motion characteristic parameters of the prediction target received by the reception unit 410 are a plurality of seismic motion characteristic parameters including a plurality of different directional characteristics centered on the predicted point position, the output processing unit 412 has the plurality. The values of multiple seismic motion indexes in each direction predicted by the prediction unit 411 based on the various seismic motion characteristic parameters of the above are expressed as the distances from the reference point in the output medium, and the multiple direction characteristics are centered on the reference point. By assigning to each direction, as shown in FIG. 16 (details will be described later), a plurality of seismic motion indexes are output to the output medium.

FIG. 16 shows an example of the output result when the seismic motion index (pseudo velocity response spectrum pSv and velocity response duration spectrum TSv) is expressed as directional characteristics, (a) is a point model S4, and (b) is a point model T4. It is a figure which shows each.

The output results shown in FIG. 16 include multiple seismic motion characteristic parameters obtained by changing the epicenter direction Λ at 5 ° intervals under the conditions of moment magnitude Mw = 6, epicenter depth H = 10 km, and epicentral distance X = 180 km. It is the output of the value of the pseudo-velocity response spectrum pSv at 5 ° intervals predicted based on the above. In FIG. 16A, the value of the pseudo-velocity response spectrum pSv in each direction is expressed as the distance from the center after being standardized by the maximum value with SIT006 (Chichibu) as the reference (center). In FIG. 16B, the value of the pseudo-velocity response spectrum pSv in each direction is expressed as the distance from the center after being standardized by the maximum value with TKO0028 (Etchujima) as the reference (center).

(Characteristics of seismic motion evaluation model 13 for each point)
Next, regarding the characteristics of the seismic motion evaluation model 13 for each point, the point model S4 generated by machine learning using the learning data set S and the point model T4 generated by machine learning using the learning data set T. I will explain while comparing with.

FIG. 17 is a diagram showing the degree of influence of each feature amount (earthquake motion characteristic parameters) on the objective variable (earthquake motion index), where (a) shows the point model S4 and (b) shows the point model T4.

In the point model S4 shown in FIG. 17A, the degree of influence on the maximum acceleration PGA and the short-period pseudo-velocity response spectrum pSv is the same as the moment magnitude Mw and the epicentral distance X, and the epicentral distance X is slightly higher. However, except for that, the influence of the moment magnitude Mw is the largest. The degree of influence of the epicenter direction Λ is greater on the short cycle side. The degree of influence of the epicenter direction Λ is larger than the pseudo velocity response spectrum pSv with respect to the velocity response duration spectrum TSv, exceeds the epicenter depth H, is equal to or exceeds the epicentral distance X, and is the Kanto models A5 and A6 shown in FIG. The result is much higher than that of. The degree of influence of the epicenter depth H is small.

In the point model T4 shown in FIG. 17B, the influence of the epicentral distance X is dominant in the short cycle, and the influence of the moment magnitude Mw is dominant in other cases. The degree of influence of the moment magnitude Mw increased with the period as a whole, and the degree of influence of the epicentral distance X decreased with the period as a whole. The degree of influence of the epicenter direction Λ is larger toward the short cycle side, and in particular, the velocity response duration spectrum TSv is larger than the maximum acceleration PGA and the pseudo velocity response spectrum pSv, and exceeds the epicentral distance X except for the cycle of 0.1 seconds, and has a short cycle. On the side, it also exceeded the moment magnitude Mw. The degree of influence of the epicenter depth H is small. In the deep-base point model T4, the influence of the moment magnitude Mw on the short-period velocity response duration spectrum TSv is relatively small as compared with the shallow-base point model S4.

FIG. 18 shows the correlation between the observed value obtained from the seismic motion observation record and the evaluation value obtained by the seismic motion evaluation model 13 for each objective variable (earthquake motion index). It is a scatter diagram which shows each point model T4. The observed values were evaluated and modeled very well as a whole, and most of the evaluated values were within half of the observed values.

FIG. 19 is a diagram showing the “evaluation value / observed value” ratio of each objective variable (earthquake motion index) as a histogram, (a) showing the point model S4, and (b) showing the point model T4. It was found that the value of the "evaluation value / observed value" ratio had a well-organized distribution centered on 1, and the variation of the velocity response duration spectrum TSv was smaller than that of the maximum acceleration PGA and the pseudo velocity response spectrum pSv.

FIG. 20 is a diagram showing the average and variation of the “evaluation value / observed value” ratio of each objective variable (earthquake motion index) for the point models S4 and T4. The average of the "evaluation / observation" ratio is approximately 1, and the common logarithm standard deviation of the "evaluation / observation" ratio is approximately 0.2 or more in the maximum acceleration PGA and the pseudo-velocity response spectrum pSv. The velocity response duration spectrum TSv was a little over 0.1.

As described above, according to the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method according to the present embodiment, the correlation between the seismic motion characteristic parameters and the seismic motion index is learned for each observation point by machine learning. Since the uncertain factors that the difference in the underground structure at each observation point gives to the seismic motion are excluded, it is possible to provide the seismic motion evaluation model 13 capable of evaluating and predicting the characteristics of the seismic motion with high accuracy.

Further, according to the seismic motion evaluation device 4 and the seismic motion evaluation method according to the present embodiment, the seismic motion evaluation model 13 for each observation point generated by the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method is used individually. It is possible to evaluate and predict the characteristics of seismic motion due to earthquakes that are expected to occur in the future with high accuracy, without relying on the experience of experts.

(Third embodiment)
In the seismic motion evaluation system 1 according to the second embodiment, the seismic motion evaluation model generation device 3 generates a seismic motion evaluation model 13 (for example, point models S4 and T4) for each observation point, and the seismic motion evaluation device 4 generates an observation point. The seismic motion at each observation point is evaluated and predicted using the seismic motion evaluation model 13 for each. On the other hand, in the seismic motion evaluation system 1 according to the third embodiment, the seismic motion evaluation model generation device 3 generates the seismic motion evaluation model 13 for each observation point group in which a plurality of observation points are grouped according to a predetermined classification standard. However, the seismic motion evaluation device 4 differs in that the seismic motion evaluation device 4 evaluates and predicts the seismic motion in each station group using the seismic motion evaluation model 13 for each station group. Since other basic configurations and operations are the same as those of the first and second embodiments, the differences between the two will be mainly described below.

FIG. 21 is a schematic diagram showing the source characteristics, propagation characteristics, and site characteristics of seismic motion at a plurality of adjacent observation points.

As in the case of the first to third observation points shown in FIG. 21, when a plurality of observation points are close to each other, when the observation points are in the immediate vicinity, or when the underground structure around the observation points is suddenly changed. Except for this, the source characteristics and propagation characteristics of seismic motion can be regarded as almost the same among multiple observation points in close proximity, and it is considered that the seismic motion input to the base directly under each observation point is almost the same. .. Regarding the site characteristics, it is considered possible to judge the similarity between a plurality of adjacent observation points by comparing the seismic motion index obtained from the seismic motion observation record and the seismic motion observation record.

Therefore, examples of the predetermined classification criteria include the first to third classification criteria shown below. For example, the first classification criterion classifies the plurality of observation points into the same observation point group when the distance between the plurality of observation points is smaller than a predetermined reference value. The second classification criterion is to classify the plurality of observation points into the same observation point group when the similarity when comparing the underground structural parameters at each of the plurality of observation points is higher than a predetermined reference value. Is. The third classification criterion is that each of the seismic motion observation records (time history waveform data) observed at a plurality of observation points or each of the seismic motion indicators obtained from each of the seismic motion observation records is, for example, GOF (Goodness of Fit). ) Is compared to classify the plurality of observation points into the same observation point group when they are more similar than a predetermined reference value.

FIG. 22 is a diagram showing a first observation point group C in which a plurality of observation points c1 to c3 are grouped and a second observation point group D in which a plurality of observation points d1 to d3 are grouped. be. By grouping a plurality of observation points c1 to c3 and d1 to d3 according to predetermined classification criteria, as shown in FIG. 22, the first observation point group C and the second observation point group D are combined. It is formed. Here, the seismic motion evaluation model 13 in the first observation point group C is referred to as "group model C", and the seismic motion evaluation model 13 in the second observation point group D is referred to as "group model D". Data 12 is referred to as "learning data set C" and "learning data set D".

FIG. 23 is a functional explanatory diagram showing an example of the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method according to the third embodiment of the present invention.

The acquisition unit 311 obtains the seismic motion data 11 based on the earthquake in which the seismic motion was observed at each observation point c1 to c3 forming the first observation point group from the database 10, and the training data 12 (learning data set C). Select as. Further, the acquisition unit 311 obtains the seismic motion data 11 based on the earthquake in which the seismic motion was observed at each observation point d1 to d3 forming the second observation point group from the database 10, and the training data 12 (learning data set). Select as D).

The generation unit 312 has feature quantities (4 types of seismic motion characteristic parameters) and objective variables (11 types) based on a plurality of learning data 12 (learning data sets C and D) acquired by the acquisition unit 311. By learning the correlation of the seismic motion index) by machine learning, seismic motion evaluation models 13 (group models C4 and D4) are generated for each observation point group.

FIG. 24 is a functional explanatory diagram showing an example of the seismic motion evaluation device 4 and the seismic motion evaluation method according to the third embodiment of the present invention.

The prediction unit 411 uses the seismic motion characteristic parameters of the prediction target received by the reception unit 410 as feature quantities, and inputs the seismic motion evaluation model 13 (group models C4, D4) according to the predicted point position to evaluate the seismic motion. Based on the objective variable output from the model 13, the seismic motion index corresponding to the seismic motion characteristic parameters to be predicted is predicted.

At that time, in the prediction unit 411, the prediction point position included in the seismic motion characteristic parameters to be predicted is inside the line connecting the observation points c1 to c3 and the observation points c1 to c3 (for example, the prediction point shown in FIG. 22). In the case of e1), the group model C4 is used, and the predicted point position is inside the line connecting the observation points d1 to d3 and the observation points d1 to d3 (for example, the prediction point e2 shown in FIG. 22). Uses the group model D4. When the predicted point position is between the first observation point group C and the second observation point group D as shown by the prediction point e3 in FIG. 22, the prediction point is predicted by using the group model C4. The seismic motion index may be averaged and the seismic motion index predicted by using the group model D4 may be averaged or proportionally divided according to the degree of similarity of the distance and the underground structure parameter.

As described above, according to the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method according to the present embodiment, the correlation between the seismic motion characteristic parameters and the seismic motion index is learned for each observation point group by machine learning. Therefore, it is possible to provide a seismic motion evaluation model that can evaluate and predict the characteristics of seismic motion with high accuracy because the variation of the learning data is reduced while facilitating the securing of learning data (teacher data). can.

Further, according to the seismic motion evaluation device 4 and the seismic motion evaluation method according to the present embodiment, the seismic motion evaluation model for each observation point group generated by the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method is used individually. It is possible to evaluate and predict the characteristics of seismic motion due to earthquakes that are expected to occur in the future with high accuracy, without relying on the experience of experts.

Furthermore, the observation point groups are classified and grouped according to predetermined classification criteria (distance between observation points, similarity of underground structure parameters, etc.), thereby excluding uncertain factors that the difference in underground structure gives to seismic motion. Therefore, it is possible to provide a seismic motion evaluation model capable of evaluating and predicting the characteristics of seismic motion with high accuracy.

(Fourth Embodiment)
In the seismic motion evaluation system 1 according to the first embodiment, the seismic motion evaluation model generation device 3 acquires learning data 12 from a database 10 based on earthquakes that have occurred in the past, and generates a seismic motion evaluation model 13. On the other hand, in the seismic motion evaluation system 1 according to the fourth embodiment, the seismic motion evaluation model generator 3 has the first database 10A based on the actual earthquakes that occurred in the past and the first database 10A based on the virtual earthquake calculated by the simulation. The difference is that the training data 12 is acquired from the database 10B of 2 and the seismic motion evaluation model 13 is generated. Since other basic configurations and operations are the same as those of the first embodiment, the differences between the two will be mainly described below.

FIG. 25 is a block diagram showing an example of the seismic motion evaluation system 1 according to the fourth embodiment of the present invention.

The seismic motion simulation device 6 calculates a virtual earthquake and a virtual seismic motion due to a virtual earthquake by executing a simulation according to a predetermined simulation method, and provides the simulation conditions and simulation results at that time to the outside as simulation data. At that time, any method can be adopted as the predetermined simulation method, and a plurality of methods may be adopted.

When the seismic motion simulation device 6 receives a data request from the seismic motion evaluation device 4, the simulation data related to the request (when a simulation condition is received from the seismic motion evaluation device 4, the simulation is performed based on the simulation condition. (Simulation data including the simulation result at the time of execution) may be provided to the seismic motion evaluation device 4. Further, the seismic motion simulation device 6 may provide simulation data by executing a simulation in its own device, or may provide simulation data when the simulation is executed by another device.

In the storage unit 30 of the seismic motion evaluation model generation device 3, in addition to the seismic motion evaluation model 13 and the seismic motion evaluation model generation program 300, a first database in which the provided data provided by the data providing devices 2A to 2C is registered and updated. 10A (corresponding to the database 10 according to the first embodiment) and the second database 10B in which the simulation data provided by the seismic motion simulation device 6 is registered / updated are stored.

By executing the seismic motion evaluation model generation program 300, the control unit 31 generates the first DB management unit 310A, the second DB management unit 310B, the first acquisition unit 311A, the second acquisition unit 311B, and the generation unit 31. It functions as a unit 312. The first DB management unit 310A and the first acquisition unit 311A correspond to the DB management unit 310 and the acquisition unit 311 according to the first embodiment, respectively.

FIG. 26 is a functional explanatory diagram showing an example of the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method according to the fourth embodiment of the present invention.

The first DB management unit 310A (first DB management process) records seismic motion characteristic parameters and seismic motion observation records when actual seismic motion is observed, based on the data provided by the data providing devices 2A to 2C. It is associated and registered in the first database 10A. Therefore, in the first database 10A, the seismic motion characteristics when the actual seismic motion is observed for each combination of the plurality of actual earthquakes and at least one observation point where the actual seismic motion due to the plurality of actual earthquakes is observed. A plurality of seismic motion data 11 in a state in which parameters and seismic motion observation records are associated are registered and stored.

The first acquisition unit 311A (first acquisition step) uses seismic motion characteristic parameters as characteristic quantities from a plurality of seismic motion data 11 registered in the first database 10A, and aims at a seismic motion index obtained from seismic motion observation records. As variables, a plurality of first training data 12A composed of feature quantities and objective variables are acquired.

The second DB management unit 310B (second DB management step) obtains various seismic motion characteristic parameters and seismic motion calculation results when the virtual seismic motion is calculated by simulation based on the simulation data provided by the seismic motion simulation device 6. It is associated and registered in the second database 10B. Therefore, in the second database 10B, the virtual seismic motion is simulated for each combination of a plurality of virtual earthquakes and at least one virtual observation point in which the virtual seismic motion caused by the plurality of virtual earthquakes is calculated as the seismic motion calculation result of the simulation. A plurality of seismic motion data 11 in a state in which the seismic motion characteristic parameters and the seismic motion calculation result when calculated by the above are associated with each other are registered and stored.

The second acquisition unit 311B (second acquisition step) aims at the seismic motion index obtained from the seismic motion calculation result by using the seismic motion characteristic parameters as feature quantities from the plurality of seismic motion data 11 registered in the second database 10B. As variables, a plurality of second learning data 12B composed of feature quantities and objective variables are acquired.

At that time, the second acquisition unit 311B obtains the second learning data 12B from the second database 10B so as to interpolate the distribution of the seismic motion characteristic parameters of the actual seismic motion by the distribution of the seismic motion characteristic parameters of the virtual seismic motion. It is preferable to acquire more than one.

FIG. 27 is a scatter diagram showing the interpolation relationship between the first learning data 12A and the second learning data 12B. The scatter plot in FIG. 27 shows the correlation between the moment magnitude Mw (horizontal axis) and the epicentral distance X (vertical axis) as the distribution of various seismic motion characteristic parameters of the actual seismic motion in the first learning data 12A (shade portion by gray scale). Is shown.

The second acquisition unit 311B is surrounded by a data shortage portion (for example, a broken line in FIG. 27) in which the number of data or the ratio thereof is smaller than a predetermined threshold in the first learning data 12A acquired from the first database 10A. For the second learning, it is searched whether or not the seismic motion characteristic parameters of the inner part) are registered in the second database 10B, and is composed of the seismic motion characteristic parameters of the virtual seismic motion corresponding to the data shortage part. Data 12B is acquired from the second database 10B. At that time, if the second learning data 12B corresponding to the data shortage portion is not registered in the second database 10B, the second acquisition unit 311B sets simulation conditions based on the data shortage portion. By requesting the seismic motion simulation device 6, the second learning data 12B corresponding to the data shortage portion may be acquired from the seismic motion simulation device 6. In the example shown in FIG. 27, the distribution of the seismic motion characteristic parameters has been described as being based on the moment magnitude Mw and the epicentral distance X, but the distribution is not limited to this, and among the characteristics included in the seismic motion characteristic parameters. , The distribution may have at least two characteristics as parameters.

Further, the seismic motion data 11 based on the plurality of simulation methods may be registered in the second database 10B, and in that case, the second acquisition unit 311B contains the seismic motion data 11 based on the plurality of simulation methods. A plurality of second learning data 12B may be acquired from the registered second database 10B. At that time, the second acquisition unit 311B may acquire the second learning data 12B so as to interpolate the distribution of the seismic motion characteristic parameters between the simulation methods, for example, or the second acquisition unit 311B may acquire the second learning data 12B. You may compare the performance and select the seismic motion data 11 based on the simulation method suitable for the seismic motion characteristic parameters as the second learning data 12B.

As shown in FIG. 26, the generation unit 312 (generation step) includes a plurality of first learning data 12A acquired by the first acquisition unit 311A and a plurality of data acquired by the second acquisition unit 311B. The seismic motion evaluation model 13 is generated as a trained model by learning the correlation between the feature amount and the objective variable by machine learning based on the second learning data 12B of the above.

The seismic motion evaluation device 4 evaluates and predicts the seismic motion based on the seismic motion evaluation model 13 generated by the seismic motion evaluation model generation device 3, and outputs the result to the output medium. Since the seismic motion evaluation device 4 is configured in the same manner as in the first embodiment, detailed description thereof will be omitted.

As described above, according to the seismic motion evaluation model generation device 3 and the seismic motion evaluation model generation method according to the present embodiment, the first learning data 12A based on the actual earthquakes that occurred in the past and the virtual earthquake calculated by the simulation. The correlation between the seismic motion characteristic parameters and the seismic motion index is learned by machine learning using the second learning data 12B based on the above. As a result, the bias of the learning data is reduced while facilitating the securing of the learning data (teacher data), so that the seismic motion evaluation model 13 capable of evaluating and predicting the characteristics of the seismic motion with high accuracy is provided. be able to.

Further, according to the seismic motion evaluation device 4 and the seismic motion evaluation method according to the present embodiment, by using the seismic motion evaluation model generation device 3 and the seismic motion evaluation model 13 generated by the seismic motion evaluation model generation method, individual experts can use the seismic motion evaluation device 4. It is possible to evaluate and predict the characteristics of seismic motion caused by an earthquake that is expected to occur in the future with high accuracy without depending on experience.

Further, the second learning data 12B is selected so as to complement the seismic motion characteristic parameters that have not occurred as an actual earthquake in the past and cannot be prepared by the first learning data 12A. Since the bias of the distribution of the learning data is reduced, it is possible to provide the seismic motion evaluation model 13 capable of evaluating and predicting the characteristics of the seismic motion with high accuracy.

(Other embodiments)
Although the embodiments of the present invention have been described above, the present invention is not limited to the above embodiments and can be appropriately modified without departing from the technical idea of the present invention.

For example, in the second embodiment, the seismic motion evaluation model generator 3 creates two point models S4 and T4, and in the third embodiment, the seismic motion evaluation model generator 3 creates two group models C4, It has been described as creating D4. On the other hand, the seismic motion evaluation model generation device 3 may generate the seismic motion evaluation model 13 for each of three or more observation points or for each of three or more observation point groups. Further, the fourth embodiment may be combined with the second embodiment or the third embodiment, and the seismic motion evaluation model generator 3 is simulated with the first database 10A based on the actual earthquakes that have occurred in the past. The learning data 12 may be acquired from the second database 10B based on the virtual earthquake calculated by the above method, and the seismic motion evaluation model 13 may be generated for each observation point or each observation point group.

When a new earthquake occurs and the seismic motion observation record is observed, the seismic motion evaluation model generator 3 according to each of the above embodiments is updated in the database 10 by the seismic motion data 11 based on the new earthquake, and after the update. The seismic motion evaluation model 13 may be automatically regenerated based on the database 10. Then, the seismic motion evaluation device 4 according to each of the above embodiments may automatically re-evaluate and re-predict the characteristics of the seismic motion based on the regenerated seismic motion evaluation model 13.

(About selection of features)
In each of the above embodiments, the feature quantities constituting the learning data 12 have been described as assuming that 6 types or 4 types of seismic motion characteristic parameters are selected from the 25 types of seismic motion characteristic parameters shown in FIG. , Arbitrary seismic motion characteristic parameters may be selected and combined as feature quantities from 25 types of seismic motion characteristic parameters, or other seismic motion characteristic parameters other than 25 types of seismic motion characteristic parameters may be further combined as feature quantities. May be good. At that time, a cluster analysis of each feature amount (see FIG. 28 described later) may be performed, and various seismic motion characteristic parameters to be feature amounts may be selected based on the analysis result. For example, when classified into a plurality of clusters, seismic motion characteristic parameters representing each cluster may be selected as feature quantities, or when there are a plurality of seismic motion characteristic parameters having high interdependence. In the above, representative seismic motion characteristic parameters may be selected as feature quantities.

FIG. 28 is a diagram showing the interdependence of each feature amount by cluster analysis. As a result of performing cluster analysis of each feature in the training data set A, 25 types of seismic motion characteristic parameters were classified into three clusters as shown in FIG. 28. That is, as zoned diagonally in FIG. 28, there are three regions in which the mutual information amount is relatively large.

The first is located in the upper right of FIG. 28, and is the epicenter position (latitude lat_eq, longitude long_eq), moment magnitude Mw, epicenter depth H, epicenter mechanism solution (Strike1, dip1, racke1), and conjugate solution of the epicenter mechanism solution (Strike1, dip1, racke1). It is a "earthquake source cluster" consisting of Strike2, dip2, racke2), earthquake type Type, fault type Tech, and epicentral distance X. The second is the observation point position (latitude lat_site, longitude long_site) located in the central part, surface layer 10m average S wave velocity AVS10, surface layer 30m average S wave velocity AVS30, top layer S wave velocity VS1, S wave velocity 700m. "Site cluster" consisting of / s layer top surface depth D7, S wave velocity 1400 m / s layer top surface depth D17, S wave velocity 2100 m / s layer top surface depth D24, seismic base surface depth D28, and microtopography classification JCODE. ". The third is the "epicenter direction cluster" located in the lower left, which consists of the epicenter directions (sinΛ, cosΛ). The direction component Comp of the seismic motion located in the upper right corner (in this embodiment, the north-south direction component or the east-west direction component) does not belong to any of the clusters and is independent.

Source clusters and site clusters were classified into source characteristics and site characteristics, respectively. The epicentral distance X, which is a propagation characteristic, is aggregated in the epicenter cluster, but the mutual information between the epicentral distance X and other features of the epicenter cluster is as small as about 0.2 or less, and the connection between the two is weak. It turned out.

(About the epicenter direction Λ)
In each of the above embodiments, the epicenter direction Λ, which is one of the features, is described as applying a pair of sin Λ and cos Λ (hereinafter referred to as “case C10”), but the epicenter direction Λ is another data format. May be applied.

FIG. 29 shows various data formats of the epicenter direction Λ, (a) shows the definition of each data format, and (b) shows each feature quantity (earthquake motion characteristic parameter) with respect to the objective variable (earthquake motion index) in each data format. It is a figure which shows the degree of influence of). Note that FIG. 29B is a result when the point model T4 generated by machine learning using the learning data set T is used.

As the data format of the feature quantity of the epicenter direction Λ, the cases C11 and C12 apply the continuous quantity using the trigonometric function as in the case C10, the case C11 applies only the sin Λ, and the case C12 applies. , CosΛ only. Further, in case C01, ΛS defining true north as the epicenter direction 0 ° was applied as it was, and in case C02, ΛN defining true south as the epicenter direction 0 ° was applied as it was. In cases C10-12, C01, and C02, the epicenter direction Λ is given as a continuous quantity for all directions, but for C01 for the northern earthquake and C02 for the southern earthquake, the epicenter is near 0 °, respectively. The azimuth values are discontinuous.

Further, as the data format of the feature amount of the seismic center direction Λ, the continuous amount corresponding to the seismic center direction Λ is not given as in the cases C10-12, C01, and C02, but as in the cases C21 to C24 and C33, all of them are used. Of a plurality of orientation divisions (for example, 12 orientations) in which the orientation is divided into a predetermined number of divisions, the orientation division corresponding to the seismic center orientation Λ may be given as a discrete quantity. Case C24 applies four sets of numbers corresponding to the number of occurrences of four directions N, S, E, and W in the character string of case C22, and case C33 changes case C23 to 20 directions.

As shown in FIG. 29 (b), the degree of influence of each feature amount in each case has the same tendency in each case, but in the point model T4, the degree of influence of the feature amount with respect to the epicenter direction Λ is the epicenter direction Λ. Is relatively small in the five cases C21 to C24 and C33 given by the discrete quantity using the orientation division, and relatively high in the cases C11 to C12, C01 and C02 where the epicenter orientation Λ is given by the continuous quantity. Do you get it.

(About the program)
In the above embodiment, the seismic motion evaluation model generation program 300 and the seismic motion evaluation program 400 have been described as being stored in the storage units 30 and 40, respectively, but the CD-ROM is a file in an installable format or an executable format. It may be recorded and provided on a computer-readable recording medium such as a DVD or a USB memory. Further, the seismic motion evaluation model generation program 300 and the seismic motion evaluation program 400 may be provided by storing them on a computer connected to a network such as the Internet and downloading them via the network.

1 ... Seismic motion evaluation system,
2A ... Observation data providing device, 2B ... Analysis data providing device,
2C ... Underground structure data providing device,
3 ... Seismic motion evaluation model generator, 4 ... Seismic motion evaluation device,
5 ... network, 6 ... seismic motion simulation device,
10 ... database, 10A ... first database, 10B ... second database,
11 ... Seismic motion data,
12 ... Learning data, 12A ... First learning data, 12B ... Second learning data,
13, 13A ~ 13K ... Seismic motion evaluation model,
30 ... storage unit, 31 ... control unit, 32 ... communication unit, 33 ... input unit, 34 ... display unit,
40 ... storage unit, 41 ... control unit, 42 ... communication unit, 43 ... input unit, 44 ... display unit,
300 ... Seismic motion evaluation model generation program,
310 ... DB management unit, 310A ... first DB management unit, 310B ... second DB management unit,
311 ... Acquisition unit, 311A ... First acquisition unit, 311B ... Second acquisition unit, 312 ... Generation unit,
400 ... Seismic motion evaluation program,
410 ... Reception unit, 411 ... Prediction unit, 412 ... Output processing unit

Claims (7)

It is a seismic motion evaluation model generation method that generates a seismic motion evaluation model by machine learning using a computer.
For each combination of multiple earthquakes and at least one observation point where seismic motion due to the multiple earthquakes was observed, various seismic motion characteristic parameters including at least the orientation characteristics of the seismic motion and seismic motion observation records when the seismic motion was observed. From the database that is stored in association with the above, a plurality of learning data composed of the feature amount and the objective variable are acquired with the seismic motion characteristic parameters as feature quantities and the seismic motion index obtained from the seismic motion observation record as the objective variable. Acquisition process and
The seismic motion evaluation model as a learned model of the machine learning by learning the correlation between the feature amount and the objective variable by the machine learning based on the plurality of learning data acquired in the acquisition step. To produce a production process, including,
Seismic motion evaluation model generation method. The azimuth characteristic is
The epicenter direction indicating the direction in which the epicenter is located with respect to the observation point.
The method for generating a seismic motion evaluation model according to claim 1. The epicenter direction is
It is given by a continuous quantity using trigonometric functions or a discrete quantity using a segmented orientation in which all directions are divided by a predetermined number of segments.
The method for generating a seismic motion evaluation model according to claim 2. A seismic motion evaluation method for evaluating the characteristics of the seismic motion based on the seismic motion evaluation model generated by the seismic motion evaluation model generation method according to any one of claims 1 to 3, using a computer.
A reception process that accepts the seismic motion characteristic parameters to be predicted, including at least the seismic motion azimuth characteristics.
The prediction target is based on the objective variable output from the seismic motion evaluation model by inputting the seismic motion characteristic parameters of the prediction target received in the reception step into the seismic motion evaluation model as the feature quantities. A prediction step for predicting the seismic motion index corresponding to the seismic motion characteristic parameters, and the like.
Seismic motion evaluation method. It further includes an output processing step of outputting the seismic motion index predicted in the prediction step to a visible output medium.
The output processing step is
When the seismic motion characteristic parameters of the prediction target received in the reception step are a plurality of the seismic motion characteristic parameters including a plurality of different orientation characteristics centered on the prediction point, the plurality of the seismic motion characteristics. The values of the plurality of seismic motion indexes predicted in the prediction step based on the parameters are expressed as the distances from the reference point in the output medium, and the plurality of the orientation characteristics are expressed in each orientation centering on the reference point. By assigning to, a plurality of the seismic motion indexes are output to the output medium.
The seismic motion evaluation method according to claim 4. It ’s a computer,
A control unit for executing each step included in the seismic motion evaluation model generation method according to any one of claims 1 to 3.
Seismic motion evaluation model generator. It ’s a computer,
A control unit for executing each step included in the seismic motion evaluation method according to claim 4 or 5.
Seismic motion evaluation device.

Images