JP7512139B2 - Earthquake motion evaluation model providing method and earthquake motion evaluation model providing device - Google Patents

Description

The present invention relates to a method for providing an earthquake motion evaluation model and a device for providing an earthquake motion evaluation model.

The technology to evaluate and predict the characteristics of earthquake motions from earthquakes that are expected to occur in the future based on various data obtained from past earthquakes (slip ruptures of earthquake source faults) and seismic motions (ground shaking caused by earthquakes) is so useful that it is widely used in society for a variety of purposes, from analyzing and interpreting the characteristics of seismic motions to predicting the behavior and structural design of buildings and structures, and even for earthquake disaster prevention.

Traditionally, some data selected by individual experts using their own judgment has been analyzed using specialized techniques to develop and utilize earthquake motion evaluation formulas (sometimes called distance attenuation formulas or earthquake motion prediction formulas) to evaluate and predict the maximum amplitude (maximum acceleration, maximum velocity, maximum displacement, etc.) of earthquake motion due to future earthquakes and response spectra (maximum response amplitude of a single-mass system when earthquake motion is input to the single-mass system with a certain natural period and damping constant) (see, for example, Patent Document 1 and Non-Patent Document 1).

JP 2008-39446 A

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

In recent years, environmental changes such as the fragmentation and sophistication of specialized fields and the decrease in the number of experts due to the declining birthrate have limited the time and effort that individual experts can directly work on proposing and improving such earthquake motion evaluation formulas. In addition, practical earthquake motion evaluation formulas that have been developed and used through previous studies depend on the data from earthquake observation records used in the studies of individual experts, and because large earthquakes and the largest earthquake motions are rare natural phenomena, there is an imbalance in the quality and quantity of previous knowledge reflected in these prediction formulas.

The increasing amount of data that is accumulated and analyzed every day, the improved information processing capabilities of computers, and advances in simulation technology have the potential to greatly contribute to improving these issues, but in order for this to happen, individual experts must reexamine and reevaluate the seismic motion evaluation formulas every time new observational data or simulation results are obtained due to the occurrence of a new earthquake. However, there is a limit to how much time and effort individual experts can devote to such work each time, and in reality this work is progressing slowly.

The present invention has been made in consideration of these circumstances, and aims to provide a method and device for providing an earthquake motion evaluation model that allows new observation data and simulation results obtained each time an earthquake occurs to be imported into a database at any time, making it possible to effectively use big data in evaluating and predicting earthquake motion.

The present invention solves the above-mentioned problems, and a method for providing a seismic motion evaluation model according to one embodiment of the present invention includes the steps of:
A method for providing an earthquake motion evaluation model to an external party, the method comprising the steps of: providing an updated version of an earthquake motion evaluation model based on machine learning using earthquake motion data, which is training data that is registered in a database in association with earthquake motion characteristic parameters and earthquake motion observation records;
a database updating step of collecting new seismic motion data and updating the database with the new seismic motion data obtained by performing a predetermined preprocessing on the new seismic motion data;
a machine learning process for acquiring new learning data from the updated database, performing the machine learning, and generating an update plan for the seismic motion evaluation model by machine learning the correlation between the seismic motion characteristic parameters and the seismic motion index value obtained from the seismic motion observation record;
The method includes a model updating step of performing a predetermined verification process on the update plan, and if the result shows that the update plan is valid, updating the latest version with the update plan.

In addition, the earthquake motion evaluation model providing device according to one embodiment of the present invention comprises:
The computer includes a control unit that executes each step included in the method for providing a seismic motion evaluation model.

According to the earthquake motion evaluation model providing device and earthquake motion evaluation device of one embodiment of the present invention, new observation data and simulation results obtained each time an earthquake occurs are subjected to a predetermined preprocessing and then adopted to update the database as needed, while new learning data is obtained from the database to generate update proposals for the earthquake motion evaluation model as needed, and if the validity of the update proposal is found through a predetermined verification process, the latest version of the earthquake motion evaluation model is updated with the update proposal and provided to the outside. Therefore, the earthquake motion evaluation model is automatically generated, verified, and improved as new data is obtained and provided to the outside, allowing big data to be effectively used for evaluating and predicting earthquake motion.

1 is a schematic configuration diagram showing an example of a seismic motion evaluation system 1 according to an embodiment of the present invention. 1 is a block diagram showing an example of a seismic motion evaluation system 1 according to an embodiment of the present invention. 1 is a functional explanatory diagram showing an example of a seismic motion evaluation model providing device 3 and a seismic motion evaluation model providing method according to an embodiment of the present invention. FIG. FIG. 2 is a data configuration diagram showing an example of a database 10. 1 shows an example of analysis of a time history waveform, where (a) is a Fourier amplitude spectrum diagram, (b) is a time history waveform diagram including a different earthquake or a different event, and (c) is a time history waveform diagram with a head shortage. FIG. 1 is a schematic diagram showing an overview of gradient boosting trees. FIG. 2 is a functional explanatory diagram showing an example of a seismic motion evaluation device 4 and a seismic motion evaluation method according to an embodiment of the present invention.

One embodiment of the present invention will be described below with reference to the attached drawings.

Figure 1 is a schematic diagram showing an example of a seismic motion evaluation system 1 according to an embodiment of the present invention. Figure 2 is a block diagram showing an example of a seismic motion evaluation system 1 according to an embodiment of the present invention.

The earthquake motion evaluation system 1 includes an observation data providing device 2A that collects earthquake motion observation records in which earthquake motion due to earthquakes is observed at multiple observation points and provides the observation results to the outside, an analysis data providing device 2B that analyzes the epicenter and magnitude of earthquakes based on the earthquake motion observation records and provides the analysis results to the outside, an underground structure data providing device 2C that provides to the outside underground structure parameters related to the underground structure used in the evaluation and prediction of earthquake motion, and an earthquake motion simulation data providing device 2D that executes an earthquake motion simulation according to a predetermined simulation method and provides to the outside the simulation conditions and simulation results at that time as simulation data.

The earthquake motion evaluation system 1 also includes an earthquake motion evaluation model providing device 3 that updates the database 10 from time to time based on new data provided by the data providing devices 2A to 2D and uses the updated database 10 to update and provide the latest version 131 of the earthquake motion evaluation model 13 based on machine learning from time to time, an earthquake motion evaluation device 4 that evaluates and predicts earthquake motion based on the earthquake motion evaluation model 13 (latest version 131) provided by the earthquake motion evaluation model providing device 3, and a network 5 that connects the devices.

The earthquake motion evaluation model 13 provided to the earthquake motion evaluation device 4 by the earthquake motion evaluation model providing device 3 can be widely used in society, for example, from analysis and interpretation of earthquake motion characteristics to behavior prediction and structural design of buildings and structures, and even earthquake disaster prevention.

When an earthquake occurs, the observation data providing device 2A collects three-component time history waveform data in the north-south, east-west, and up-down directions measured by seismometers (not shown) installed at multiple observation points as seismic motion observation records, and provides observation data including the seismic motion observation records and additional information such as the observation point locations indicating the locations of the observation points to the outside. In this embodiment, the observation data is, for example, data provided by the strong earthquake observation network K-NET of the National Research Institute for Earth Science and Disaster Resilience (hereinafter referred to as "NIED"), and in particular, data observed at 138 observation points (see Figure 1) installed in the Kanto region (Tokyo, Kanagawa, Chiba, Saitama, Ibaraki, Tochigi, and Gunma).

The analysis data providing device 2B analyzes the epicenter and scale of the earthquake based on the seismic motion observation records, and provides the analysis results to the outside, including, for example, moment magnitude, Japan Meteorological Agency magnitude, epicenter location, epicenter depth, earthquake type, fault type, and focal mechanism solution. In this embodiment, the analysis data providing device 2B is described as using data provided by, for example, the broadband earthquake 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 provides underground structure data including, for example, earthquake bedrock depth, engineering bedrock depth, layer thickness, density, seismic wave propagation velocity, Q value, attenuation constant, etc., as underground structure parameters to the outside. In this embodiment, the underground structure data providing device 2C is described as using data provided by, for example, the Earthquake Hazard Station J-SHIS of the National Research Institute for Earth Science and Disaster Prevention.

When an earthquake occurs, the observation data providing device 2A, the analysis data providing device 2B, and the underground structure data providing device 2C may provide data relating to the earthquake in real time to the earthquake motion assessment model providing device 3, or when a data request is received from the earthquake motion assessment model providing device 3, they may provide data relating to the request (which may be data relating to multiple earthquakes that have occurred in the past and meet certain conditions) to the earthquake motion assessment model providing device 3. In addition, in this embodiment, the data providing devices 2A to 2C are described as three separate devices, but this is not limited to this and they may be configured as one device, or other data providing devices may be added.

When an earthquake occurs or when it is assumed that a virtual earthquake has occurred, the earthquake motion simulation data providing device 2D analyzes the earthquake motion caused by the earthquake according to a predetermined simulation method, and provides the simulation conditions and simulation results at that time to the outside as simulation data. In this case, any method can be adopted as the predetermined simulation method, and multiple methods may be adopted. Furthermore, the earthquake motion simulation data providing device 2D may provide simulation data by executing an earthquake motion simulation on its own device, or may provide simulation data when an earthquake motion simulation is executed on another device.

The earthquake motion simulation data providing device 2D may provide simulation data relating to a new earthquake motion simulation to the earthquake motion evaluation model providing device 3 at any time when a new earthquake motion simulation is executed, or when a data request is received from the earthquake motion evaluation model providing device 3, it may provide simulation data relating to the request to the earthquake motion evaluation model providing device 3 (if simulation conditions are received from the earthquake motion evaluation model providing device 3, this may be simulation data including the simulation results when an earthquake motion simulation is executed based on the simulation conditions).

Network 5 communicates various data and signals via wireless or wired communication, and any communication standard may be used.

(Configuration of the earthquake motion evaluation model providing device 3 and processes performed by each part)
The earthquake motion evaluation model providing device 3 collects provided data (observation data, analysis data, underground structure data, simulation data) provided by the data providing devices 2A to 2D, and updates the database 10 as needed. The earthquake motion evaluation model providing device 3 uses the updated database 10 to execute machine learning algorithms such as gradient boosting trees, random forests, neural networks, and convolutional neural networks (CNN), thereby generating an update proposal 130 for the earthquake motion evaluation model 13 as a trained model of machine learning. The earthquake motion evaluation model providing device 3 updates a latest version 131 of the earthquake motion evaluation model 13 with the update proposal 130, and provides the updated latest version 131 of the earthquake motion evaluation model 13 to an external device (in this embodiment, the earthquake motion evaluation device 4).

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

The storage unit 30 stores a database 10 in which the provided data provided by the data providing devices 2A to 2D is registered and updated, a trained model, an earthquake motion evaluation model 13, and an earthquake motion evaluation model providing program 300 that controls the operation of the earthquake motion evaluation model providing device 3 to realize the earthquake motion evaluation model providing method. The database 10 may be stored in an external storage device instead of the storage unit 30, and in that case, the earthquake motion evaluation model providing device 3 can access the database 10 by communicating with the external storage device via the network 5.

The control unit 31 executes the earthquake motion assessment model provision program 300 to function as a database update unit 310, a machine learning unit 311, and a model update unit 312. The functions of each unit will be described in detail later.

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

(Regarding the database update process by the database update unit 310)
The database update unit 310 collects new seismic motion data 11A based on new data provided by the data providing devices 2A to 2D (observation data, analysis data, underground structure data, simulation data), and updates the database 10 with update seismic motion data 11B that has been subjected to a predetermined pre-processing on the new seismic motion data 11A.

Figure 4 is a data configuration diagram showing an example of database 10. In database 10, for each combination of multiple earthquakes and at least one observation point where seismic motion due to the multiple earthquakes was observed, multiple pieces of seismic motion data 11 are registered and stored, in which seismic motion characteristic parameters and seismic motion observation records at the time the seismic motion was observed are associated with the seismic motion. When the simulation data includes simulation conditions and simulation results when seismic motion due to an actual earthquake or a virtual earthquake is calculated by seismic motion simulation, the simulation conditions are treated as seismic motion characteristic parameters and the simulation results as seismic motion observation records or seismic motion indices, and the simulation data is registered in database 10 as seismic motion data 11.

The earthquake motion characteristic parameters are various parameters that describe the characteristics of earthquake motion, and the characteristics of earthquake motion include, for example, the source characteristics and propagation characteristics of earthquake motion, as well as site characteristics, orientation characteristics, and earthquake motion observation characteristics. In this embodiment, the earthquake motion characteristic parameters are data provided by the data providing devices 2A to 2D (observation data, analysis data, underground structure data) that are classified and recorded according to the above-mentioned characteristics of earthquake motion. The source characteristics, propagation characteristics, site characteristics, orientation characteristics, and earthquake motion observation characteristics included in the earthquake motion characteristic parameters are described below.

The source characteristics include at least one of the following: magnitude (moment magnitude Mw, Japan Meteorological Agency magnitude, etc.), epicenter location (latitude lat_eq, longitude lon_eq), source depth H, earthquake type Type (inland crustal earthquake, plate boundary earthquake, intraslab earthquake), fault type Mech (normal fault, reverse fault, strike-slip fault), source mechanism solution (strike Strike1, dip angle dip1, slip angle rake1), and the conjugate solution of the source mechanism solution (strike Strike2, dip angle dip2, slip angle rake2), etc. The source characteristics in this embodiment are moment magnitude Mw, epicenter location (latitude lat_eq, longitude lon_eq), source depth H, earthquake type Type, fault type Mech, source mechanism solution (Strike1, dip1, rake1), and conjugate solution of the source mechanism solution (Strike2, dip2, rake2).

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

The site characteristics are, for example, at least one of the following: observation point location (latitude lat_site, longitude lon_site), seismic bedrock depth, engineering bedrock depth, layer thickness, density, seismic wave propagation velocity, Q value, and attenuation constant. The site characteristics according to this embodiment are the observation point location (latitude lat_site, longitude lon_site), the S-wave velocity of the top layer VS1, the average S-wave velocity of the surface 10 m AVS10, the average S-wave velocity of the surface 30 m AVS30, the microtopography classification JCODE, the depth of the top surface of the S-wave velocity 700 m/s layer D7, the depth of the top surface of the S-wave velocity 1400 m/s layer D17, the depth of the top surface of the S-wave velocity 2100 m/s layer D24, and the depth of the seismic bedrock surface D28 (= the depth of the top surface of the S-wave velocity 2700 m/s layer). In addition, when the average S-wave velocity AVS30 of the surface layer 30 m of the 250 m mesh of the earthquake hazard station J-SHIS of the National Research Institute for Earth Science and Disaster Prevention cannot be obtained from the underground structure data of the observation point, the average S-wave velocity AVS30 of the surface layer 30 m is substituted, and when the average S-wave velocity AVS10 of the surface layer 10 m is not obtained from the underground structure data of the observation point, the average S-wave velocity AVS30 of the surface layer 30 m is substituted. In addition, the seismic bedrock depth D28 is the bottom surface depth of the 28th layer of the deep ground model of the mesh that includes the target observation point location published by the earthquake hazard station J-SHIS of the National Research Institute for Earth Science and Disaster Prevention (equal to the top surface depth of the 29th layer with a P-wave velocity of 5000 m/s and an S-wave velocity of 2700 m/s, which corresponds to the seismic bedrock in the same model).

The direction characteristic is, for example, the epicenter direction Λ, which indicates the direction 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 is located with respect to the observation point position. In this case, the epicenter direction Λ is defined clockwise with true north as 0°, and since it is a discontinuous quantity with true north as the boundary, the direction characteristic in this embodiment uses a pair of sinΛ and cosΛ that represents the epicenter direction Λ.

The earthquake motion observation characteristic is, for example, the directional component Comp of earthquake motion, which indicates that the data of the time history waveform recorded as the earthquake motion observation record is in either the north-south direction, the east-west direction, or the up-down direction.

The earthquake motion observation record is, for example, a time history waveform of the three components of acceleration, velocity, and displacement in the north-south, east-west, and up-down directions. The earthquake motion observation record is evaluated and analyzed using a specified evaluation method to obtain various earthquake motion indices. The earthquake motion observation record in this embodiment is a time history waveform of two horizontal components (north-south and east-west directions) observed at 138 observation points (see Figure 1) installed in the Kanto region (Tokyo, Kanagawa, Chiba, Saitama, Ibaraki, Tochigi, and Gunma) as part of the K-NET strong earthquake observation network.

The seismic motion index includes at least one of the amplitude characteristics, periodic characteristics, and time characteristics of the seismic motion.

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

The periodic characteristic is a response value for at least one period in a response spectrum or Fourier spectrum obtained from a seismic motion observation record (time history waveforms of acceleration, velocity, and displacement). The response spectrum is, for example, an acceleration response spectrum, a pseudo velocity response spectrum pSv, a velocity response spectrum, and a displacement response spectrum for a predetermined damping constant (for example, 5%). The Fourier spectrum is, for example, an acceleration Fourier spectrum, a velocity Fourier spectrum, and a displacement Fourier spectrum. The periodic characteristics according to this embodiment are five pseudo velocity response spectra pSv (0.1s), pSv (0.5s), pSv (1s), pSv (3s), and pSv (5s) with a damping constant of 5% at periods of 0.1 seconds, 0.5 seconds, 1 second, 3 seconds, and 5 seconds.

The time-dependent characteristic is, for example, the response duration for at least one period in the response duration spectrum obtained from the earthquake motion observation record (time history waveforms of acceleration, velocity, and displacement). The response duration spectrum is, for example, the acceleration response duration spectrum, the velocity response duration spectrum TSv, and the displacement response duration spectrum for a predetermined damping constant (for example, 5%). The time-dependent characteristic according to this embodiment is five velocity response duration spectra TSv (0.1s), TSv (0.5s), TSv (1s), TSv (3s), and TSv (5s) with a damping constant of 5% in periods of 0.1 seconds, 0.5 seconds, 1 second, 3 seconds, and 5 seconds. The parameters that define the start and end of the response duration are p1 = 0.03 and p2 = 0.95.

The update conditions under which the database update unit 310 updates the database 10 may be, for example, when a predetermined number or more of new earthquake motion data 11A are collected, when a predetermined number or more of earthquake motion data 11B for update are accumulated, at a predetermined period (for example, every hour, every day, every week, every month, etc.), or a combination of these conditions. In addition, the update conditions may be changed as appropriate based on instructions from the administrator of the earthquake motion evaluation model providing device 3, or the same update conditions may be set at all times.

(Regarding Predetermined Preprocessing of New Seismic Motion Data 11A)
The predetermined pre-processing by the database update unit 310 includes at least one of an exclusion process for analyzing the new seismic motion data 11A and excluding the seismic motion data 11A when it is inappropriate to register the data in the database 10, and a processing process for performing a correction process and a noise removal process on the new seismic motion data 11A. Therefore, when the database update unit 310 performs both the exclusion process and the processing process as the pre-processing, the new seismic motion data 11A that was not excluded in the exclusion process is used as the update seismic motion data 11B, and the new seismic motion data 11A after the correction process and the noise removal process are performed in the processing process is newly registered in the database 10.

In the exclusion process, for example, the time history waveform of the earthquake motion observation record included in the new earthquake motion data 11A is analyzed, and it is determined whether or not it is appropriate to register the earthquake motion data 11A in the database 10. If it is determined to be inappropriate, the earthquake motion data 11A is excluded from being registered in the database 10. Note that in the exclusion process, if the administrator determines that the new earthquake motion data 11A is inappropriate, the earthquake motion data 11A may be excluded based on an instruction operation by the administrator indicating that fact.

For example, the exclusion process includes at least one of a first exclusion process, a second exclusion process, and a third exclusion process. The first exclusion process is determined to be inappropriate when the signal-to-noise ratio (S/N ratio) in the time history waveform is smaller than a predetermined value. The second exclusion process is determined to be inappropriate when a different earthquake or different event is mixed in the time history waveform. The third exclusion process is determined to be inappropriate when the recording time of the seismic observation record in the time history waveform is insufficient (insufficient head, insufficient tail, or both).

Figure 5 shows an example of analysis of time history waveforms, where (a) is a Fourier amplitude spectrum diagram, (b) is a time history waveform diagram including a different earthquake or different event, and (c) is a time history waveform diagram with head shortage.

When an expert judges the suitability of the earthquake motion data 11A based on, for example, whether the amplitude is small and the noise is large in the earthquake motion observation record included in the earthquake motion data 11A, it is generally difficult to judge the presence or absence of noise and the signal-to-noise ratio using only numerical data. Therefore, the expert creates, for example, a Fourier amplitude spectrum diagram (image data) shown in FIG. 5(a) based on the time history numerical data, visually checks the Fourier amplitude spectrum diagram to identify noise, sets the period range of a frequency filter to remove noise, and finally judges the suitability of the earthquake motion data 11A. In addition, the expert visually checks, for example, the time history waveform diagram (image data) shown in FIG. 5(b) and (c) to judge whether another earthquake or another event is mixed in and whether the recording time is insufficient. Therefore, it is possible to adopt a method in which a machine learning model is generated in advance using the image data to which the judgment result when the expert visually judges the suitability of the earthquake motion data 11A is assigned as a correct answer label, and the suitability of the earthquake motion data 11A is judged in the exclusion process using the machine learning model.

As a preliminary step for the first exclusion process, a first time history waveform evaluation model 14A is generated by machine learning the correlation between image data based on a time history waveform to which an abnormal state related to the signal-to-noise ratio is assigned as a correct answer label (see FIG. 5(a)), and the correlation between the image data and the abnormal state is stored in the storage unit 30. Then, in the first exclusion process, image data based on the time history waveform of the seismic motion observation record included in the new seismic motion data 11A is input to the first time history waveform evaluation model 14A. Then, based on the abnormal state related to the signal-to-noise ratio output from the first time history waveform evaluation model 14A, the suitability of the new seismic motion data 11A is judged, and if it is judged to be inappropriate, the seismic motion data 11A is excluded.

As a preliminary step for the second exclusion process, a second time history waveform evaluation model 14B is generated by machine learning the correlation between image data based on a time history waveform to which an abnormal state related to another earthquake or another event is assigned as a correct answer label (see FIG. 5(b)), and the correlation between the image data and the abnormal state is determined, and the second time history waveform evaluation model 14B is stored in the storage unit 30. Then, in the second exclusion process, image data based on the time history waveform of the seismic motion observation record included in the new seismic motion data 11A is input to the second time history waveform evaluation model 14B. As a result, the suitability of the new seismic motion data 11A is determined based on the abnormal state related to another earthquake or another event output from the second time history waveform evaluation model 14B, and the seismic motion data 11A is excluded if it is determined to be inappropriate.

As a preliminary step for the third exclusion process, a third time history waveform evaluation model 14C is generated by machine learning the correlation between image data based on a time history waveform to which an abnormal state related to the recording time is assigned as a correct answer label (see FIG. 5(c)), and the correlation between the image data and the abnormal state is stored in the storage unit 30. Then, in the third exclusion process, image data based on the time history waveform of the seismic motion observation record included in the new seismic motion data 11A is input to the third time history waveform evaluation model 14C. As a result, the suitability of the new seismic motion data 11A is determined based on the abnormal state related to the recording time output from the third time history waveform evaluation model 14C, and the seismic motion data 11A is excluded if it is determined to be inappropriate.

The first to third time history waveform evaluation models 14A to 14C can use any machine learning algorithm, for example, a convolutional neural network (CNN). As for the image data, any image data including a time history waveform diagram, a Fourier amplitude spectrum diagram, etc. can be used as long as it is image data based on a time history waveform. In this case, the image data may have been subjected to processing such as scale standardization.

The processing process includes, as a correction process, performing, for example, baseline correction, etc., on the time history waveform of the earthquake motion observation record included in the new earthquake motion data 11A. In addition, the processing process includes, as a noise removal process, performing, for example, time history filtering, frequency filtering, etc., on the time history waveform of the earthquake motion observation record included in the new earthquake motion data 11A.

(Machine learning process by machine learning unit 311)
As shown in Fig. 3, the machine learning unit 311 acquires a plurality of new learning data 12 consisting of feature quantities and objective variables from a plurality of seismic motion data 11 (including update seismic motion data 11B) registered in the updated database 10, with seismic motion characteristic parameters as feature quantities and seismic motion indices obtained from seismic motion observation records as objective variables. The learning data 12 is data used as learning data (training data), verification data, and test data in supervised learning. A collection of the learning data 12 consisting of a plurality of learning data 12 is called a learning dataset.

When acquiring multiple pieces of new learning data 12 from the updated database 10, the machine learning unit 311 may acquire the new learning data 12 so as to intentionally include the updating earthquake motion data 11B, or may acquire the new learning data 12 randomly, regardless of whether the updating earthquake motion data 11B is included. In addition, the machine learning unit 311 may select earthquake motion data 11 based on an earthquake that satisfies at least one of a specified earthquake condition and a specified earthquake motion condition as the learning data 12. For example, it is preferable that the earthquake conditions and the earthquake motion conditions are set so that earthquake motions with a relatively large amplitude are distributed in a balanced manner.

In this embodiment, the feature quantities constituting the learning data 12 are six of the 25 types of earthquake motion characteristic parameters shown in FIG. 4: moment magnitude Mw, epicenter depth H, epicenter distance X, epicentral direction Λ (a pair of sinΛ and cosΛ), average S-wave velocity AVS30 in the surface layer 30 m, and depth of the seismic bedrock surface (depth to the top of the layer with an S-wave velocity of 2700 m/s) D28. The objective variable constituting the learning data 12 is one of the 11 types of earthquake motion indices shown in FIG. 4: maximum acceleration PGA.

As shown in FIG. 3, the machine learning unit 311 performs machine learning using multiple pieces of learning data 12 acquired from the updated database 10, generates an update proposal 130 for the seismic motion assessment model 13 as a trained model in which correlations between feature quantities and objective variables have been machine-learned, and temporarily stores the proposed update in the storage unit 30. In this embodiment, a case in which a gradient boosting decision tree is used as a machine learning algorithm in the machine learning will be described.

Figure 6 is an overview of gradient boosting trees. Gradient boosting trees are learning devices that combine gradient boosting and decision trees. Gradient boosting is a method for constructing a strong learner (high-performance machine learning model) by combining multiple weak learners (low-performance machine learning models). Decision trees are a method for generating machine learning models capable of classification and regression by performing conditional branching using the branching structure of a tree. Gradient boosting trees, which combine these two methods, are a method for combining multiple weak learners generated by decision trees using gradient boosting.

Of the 11 types of seismic motion indices shown in Fig. 4, the number of data for the maximum acceleration PGA and the pseudo velocity response spectrum pSv is considered to decrease rapidly as the amplitude increases. Therefore, in order to reduce bias in the distribution of the objective variables, it is preferable to use common logarithms (log ₁₀ PGA and log ₁₀ pSv) as the objective variable data for the maximum acceleration PGA and the pseudo velocity response spectrum pSv, respectively. In addition, as the loss function in the gradient boosting tree, it is preferable to apply the least squares method (normal distribution) to the maximum acceleration PGA and the pseudo velocity response spectrum pSv, and to apply the Poisson distribution to the velocity response duration spectrum TSv.

Furthermore, when the model update unit 312 uses the gradient boosting tree to generate the earthquake motion evaluation model 13 (update proposal 130), if the 11 types of earthquake motion indexes shown in FIG. 4 are adopted as the objective variables, the earthquake motion evaluation model 13 (update proposal 130) may be generated for each objective variable. That is, the first earthquake motion evaluation model 13A is generated by machine learning the correlation between the feature amount and the objective variable (maximum acceleration PGA = first type of earthquake motion index), and the second earthquake motion evaluation model 13B is generated by machine learning the correlation between the feature amount and the objective variable (pseudo velocity response spectrum pSv (0.1 s) = second type of earthquake motion index), and so on. In this way, a total of 11 earthquake motion evaluation models 13A to 13K may be generated, such as the third earthquake motion evaluation model 13C to the eleventh earthquake motion evaluation model 13K.

(Model updating process by the model updating unit 312)
The model update unit 312 performs a predetermined verification process on the update plan 130 , and if the result shows that the update plan 130 is valid, updates the latest version 131 with the update plan 130 .

(Regarding a Predetermined Verification Process for the Update Proposal 130)
The predetermined verification process by the model update unit 312 quantitatively or qualitatively verifies the validity of the update plan 130, and if the validity of the update plan 130 is recognized as a result (in the case of "OK" in FIG. 3), the latest version 131 is updated with the update plan 130 as described above. On the other hand, if the validity of the update plan 130 is not recognized (in the case of "NG" in FIG. 3), the latest version 131 is not updated with the update plan 130, and the update plan 130 is not provided to the outside. In that case, the update plan 130 may be deleted from the storage unit 30, or may remain stored in the storage unit 30 without being provided to the outside. In addition, in the verification process, if the administrator determines that the update plan 130 is not valid, the latest version 131 may not be updated with the update plan 130 based on an instruction operation by the administrator indicating that effect.

The quantitative verification process includes, for example, at least one of a first verification process, a second verification process, and a third verification process.

In the first verification process, the update plan 130 is verified by comparing the predicted value of the earthquake motion index based on the update plan 130 with the predicted value (reference value) of the earthquake motion index based on the latest version 131 in the present or past. In the first verification process, for example, the earthquake motion characteristic parameters included in the test data in the learning dataset are input as features to the update plan 130, and the objective variable output from the update plan 130 is obtained as the predicted value of the earthquake motion index based on the update plan 130. In addition, the same test data is input as features to the latest version 131 in the present or past, and the objective variable output from the earthquake motion evaluation model 13 is obtained as a reference value. Then, the two values are compared, and if they are equal to or less than a predetermined threshold, the validity of the update plan 130 is recognized. Note that, when multiple test data are used, the average value, standard deviation, etc. of both are compared, and if they are equal to or less than a predetermined threshold, the validity of the update plan 130 is recognized.

The second verification process verifies the update proposal 130 by comparing the predicted value of the earthquake motion index based on the update proposal 130 with the observed value (reference value) of the earthquake motion index obtained from the earthquake motion observation record included in the earthquake motion data 11. In the second verification process, as in the first verification process, the predicted value of the earthquake motion index based on the update proposal 130 is obtained using test data in the learning dataset. Also, the observed value of the earthquake motion index is obtained as a reference value from the earthquake motion observation record included in the same test data. Then, the two values are compared, and if they are below a predetermined threshold, the validity of the update proposal 130 is recognized. Note that when multiple test data are used, the average values and standard deviations of the two are compared, and if they are below a predetermined threshold, the validity of the update proposal 130 is recognized, as in the first verification process.

In the third verification process, the update plan 130 is verified by comparing the predicted value of the earthquake motion index based on the update plan 130 with the calculated value (reference value) of the earthquake motion index obtained from a predetermined earthquake motion evaluation formula. In the third verification process, as in the first verification process, the predicted value of the earthquake motion index based on the update plan 130 is obtained using test data in the learning dataset. In addition, the calculated value of the earthquake motion index is obtained as a reference value by substituting the earthquake motion characteristic parameters contained in the same test data into a predetermined earthquake motion evaluation formula. Then, the validity of the update plan 130 is recognized if the two values are compared and are equal to or less than a predetermined threshold. Note that, when multiple test data are used, the average values and standard deviations of the two are compared and the validity of the update plan 130 is recognized if they are equal to or less than a predetermined threshold, as in the first verification process.

The qualitative verification process verifies the update proposal 130, for example, based on whether the framework of the earthquake motion assessment model 13 generated as the update proposal 130 appears physically unnatural.

When the validity of the update proposal 130 of the earthquake motion evaluation model 13 is confirmed in the above-mentioned verification process, the model update unit 312 updates the latest version 131 of the earthquake motion evaluation model 13 with the update proposal 130. As a result, the latest version 131 updated with the update proposal 130 is made available to the outside and is made public. At that time, the latest version 131 may be automatically distributed to a distribution destination registered in advance (in this embodiment, the earthquake motion evaluation device 4) via the network 5, or may be made public in a downloadable state. In addition, not only the latest version 131 of the earthquake motion evaluation model 13 but also the updated database 10 may be made public. Note that even if the latest version 131 is updated with the update proposal 130, whether or not to make the latest version 131 public to the outside may be finally determined based on an instruction operation of the administrator.

(Configuration of the earthquake motion evaluation device 4 and processes performed by each part)
The earthquake motion evaluation device 4 evaluates and predicts earthquake motion based on the earthquake motion evaluation model 13 (latest version 131) provided by the earthquake motion evaluation model providing device 3, and outputs the results to an output medium such as a display medium or paper medium.

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

The memory unit 40 stores the earthquake motion evaluation model 13 provided as a trained model by the earthquake motion evaluation model providing device 3, and an earthquake motion evaluation program 400 that controls the operation of the earthquake motion evaluation device 4 to realize the earthquake motion evaluation method.

The control unit 41 executes the earthquake motion evaluation program 400 to function as a reception unit 410, a prediction unit 411, and an output processing unit 412. The functions of each unit will be described in detail later.

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

(Regarding the reception process by the reception unit 410)
The reception unit 410 receives various seismic motion characteristic parameters of the prediction target. Specifically, the reception unit 410 receives, for example, the moment magnitude Mw, epicenter position, and epicenter depth H of an earthquake (hereinafter referred to as an "anticipated earthquake") that the user of the seismic motion evaluation device 4 assumes as a prediction target via the input unit 33, and also receives, via the input unit 33, a prediction point position indicating the position of a prediction point at which it is desired to predict the extent of seismic motion caused by the anticipated earthquake. The prediction point position may be any position, or may be the same as the observation point position. In addition, there may be multiple prediction point positions, and for example, each lattice point at a predetermined lattice interval (for example, 5 km interval).

Then, the reception unit 410 calculates the epicenter distance X and epicenter direction Λ based on the epicenter position of the expected earthquake and the predicted point position of the earthquake motion, and acquires the average S-wave velocity AVS30 of the surface layer 30 m for the predicted point position of the earthquake motion and the depth of the earthquake bedrock surface (depth of the top surface of the S-wave velocity 2700 m/s layer) D28 based on the underground structure data provided by the underground structure data providing device 2C. As a result, the reception unit 410 receives the earthquake motion characteristic parameters of the prediction target (moment magnitude Mw of the expected earthquake, epicenter depth H of the expected earthquake, epicenter distance X between the epicenter of the expected earthquake and the predicted point position, epicenter direction Λ indicating the direction in which the epicenter position of the expected earthquake exists based on the predicted point position, average S-wave velocity AVS30 of the surface layer 30 m at the predicted point position, and earthquake bedrock surface depth D28 at the predicted point position).

(Regarding the prediction process by the prediction unit 411)
The prediction unit 411 inputs the earthquake motion characteristic parameters of the prediction target received by the reception unit 410 as feature quantities into an earthquake motion evaluation model 13 (in this embodiment, a trained model that has learned the correlation between six types of earthquake motion characteristic parameters as feature quantities and one type of earthquake motion index as an objective variable) and predicts an earthquake motion index (predicted value) corresponding to the earthquake motion characteristic parameters of the prediction target based on the objective variable output from the earthquake motion evaluation model 13. In this case, when the reception unit 410 receives a plurality of earthquake motion characteristic parameters as the prediction target by receiving a plurality of expected earthquakes or a plurality of prediction point positions, the prediction unit 411 inputs each of the plurality of earthquake motion characteristic parameters into the earthquake motion evaluation model 13 to predict the earthquake motion index (predicted value) corresponding to each of the plurality of earthquake motion characteristic parameters.

When the earthquake motion evaluation model 13 includes a total of 11 earthquake motion evaluation models 13A to 13K generated for each objective variable (11 types of earthquake motion indices), the prediction unit 411 inputs the earthquake motion characteristic parameters of the prediction target as features to each of the 11 earthquake motion evaluation models 13A to 13K, and predicts the earthquake motion index (predicted value) corresponding to the earthquake motion characteristic parameters of the prediction target for each objective variable (11 types of earthquake motion indices) output from each of the 11 earthquake motion evaluation models 13A to 13K.

(Regarding the output processing step by the output processing unit 412)
The output processing unit 412 outputs the predicted value of the seismic motion index predicted by the prediction unit 411 to a visually recognizable 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 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 data on the paper medium. The output processing unit 412 may communicate and output the output data to, for example, a seismic motion prediction map creation system or a hazard map creation system, or may communicate and output the output data to a disaster prevention system for a public facility, building, factory, or the like.

For example, when the earthquake motion characteristic parameters to be predicted received by the receiving unit 410 are multiple earthquake motion characteristic parameters for one anticipated earthquake, for example, with each grid point being a prediction point position, the output processing unit 412 outputs the predicted values of multiple earthquake motion indices at each prediction point position predicted by the prediction unit 411 based on the multiple earthquake motion characteristic parameters to an output medium so as to be superimposed on a map, for example, as a contour diagram or a color-coded mesh diagram.

In addition, when the earthquake motion characteristic parameters to be predicted received by the receiving unit 410 are multiple earthquake motion characteristic parameters including multiple different directional characteristics centered on the prediction point position, the output processing unit 412 represents the predicted values of multiple earthquake motion indices in each direction predicted by the prediction unit 411 based on the multiple earthquake motion characteristic parameters as distances from a reference point on the output medium, and outputs multiple earthquake motion indices to the output medium by assigning multiple directional characteristics to each direction centered on the reference point.

Furthermore, when the earthquake motion characteristic parameters to be predicted received by the receiving unit 410 are multiple earthquake motion characteristic parameters including different epicenter distances, the output processing unit 412 outputs multiple earthquake motion indices to an output medium by expressing the predicted values of the multiple earthquake motion indices predicted by the prediction unit 411 based on the multiple earthquake motion characteristic parameters as distance attenuation characteristics.

As described above, according to the earthquake motion evaluation model providing device 3 and earthquake motion evaluation model providing method of this embodiment, new observation data and simulation results obtained each time an earthquake occurs are subjected to a predetermined preprocessing and then adopted to update the database 10 as needed, while new learning data is obtained from the database 10 to generate an update proposal 130 for the earthquake motion evaluation model 13 as needed, and if the validity is confirmed by a predetermined verification process, the latest version 131 of the earthquake motion evaluation model 13 is updated with the update proposal 130 and provided to the outside. Therefore, the earthquake motion evaluation model 13 is automatically generated, verified, and improved as new data is obtained and provided to the outside, so that big data can be effectively used for evaluating and predicting earthquake motion.

In addition, according to the earthquake motion evaluation device 4 and earthquake motion evaluation method of this embodiment, by using the earthquake motion evaluation model 13 provided from time to time by the earthquake motion evaluation model providing device 3 and the earthquake motion evaluation model generating method, it is possible to evaluate and predict with high accuracy the characteristics of earthquake motion due to earthquakes that are expected to occur in the future, without relying on the experience of individual experts.

Other Embodiments
Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and can be modified as appropriate without departing from the technical concept of the present invention.

For example, in the above embodiment, the earthquake motion evaluation model providing device 3 generates the earthquake motion evaluation model 13 using observation data from 138 observation points installed in one metropolis and six prefectures in the Kanto region, and the earthquake motion evaluation device 4 uses the earthquake motion evaluation model 13 to evaluate and predict earthquake motion in the Kanto region. However, the region covered by the earthquake motion evaluation model 13 is not limited to this, and the range and shape of the target region may also be changed as desired. Furthermore, the earthquake motion evaluation model 13 may not only cover a region, but may also cover an observation point, or may cover an observation point group in which multiple observation points are grouped according to a predetermined classification criterion.

In the above embodiment, the feature quantities constituting the learning data 12 are described as six types of earthquake motion characteristic parameters selected from the 25 types of earthquake motion characteristic parameters shown in FIG. 4, but any earthquake motion characteristic parameters may be selected from the 25 types of earthquake motion characteristic parameters as feature quantities and combined, or earthquake motion characteristic parameters other than the 25 types of earthquake motion characteristic parameters may be further combined as feature quantities.

In the above embodiment, the earthquake motion evaluation model providing program 300 and the earthquake motion evaluation program 400 are described as being stored in the storage units 30 and 40, respectively, but they may be provided by being recorded in an installable or executable format on a computer-readable recording medium such as a CD-ROM, DVD, or USB memory. Also, the earthquake motion evaluation model providing program 300 and the earthquake motion evaluation program 400 may be provided by being stored on a computer connected to a network such as the Internet and downloaded via the network.

1...Earthquake motion evaluation system,
2A: Observation data providing device, 2B: Analysis data providing device,
2C: Underground structure data providing device; 2D: Earthquake motion simulation data providing device;
3...earthquake motion evaluation model providing device, 4...earthquake motion evaluation device, 5...network,
10...database, 11, 11A, 11B...earthquake motion data, 12...learning data, 13, 13A to 13K...earthquake motion evaluation model,
14A: first time history waveform evaluation model, 14B: second time history waveform evaluation model, 14C: third time history waveform 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,
130...Update proposal, 131...Latest version,
300: Earthquake motion evaluation model provision program,
310: database update unit, 311: machine learning unit, 312: model update unit,
400…Earthquake motion evaluation program,
410: Reception unit, 411: Prediction unit, 412: Output processing unit

Claims (4)

A method for providing an earthquake motion evaluation model, comprising the steps of: performing machine learning using earthquake motion data, which is registered in a database in association with earthquake motion characteristic parameters and earthquake motion observation records, as learning data; and providing an updated version of an earthquake motion evaluation model based on the machine learning to an external party, the method comprising the steps of:
a database updating step of collecting new seismic motion data and updating the database with the new seismic motion data obtained by performing a predetermined preprocessing on the new seismic motion data;
a machine learning process for acquiring new learning data from the updated database, performing the machine learning, and generating an update plan for the seismic motion evaluation model by machine learning the correlation between the seismic motion characteristic parameters and the seismic motion index obtained from the seismic motion observation record;
a model updating step of performing a predetermined verification process on the update plan and, if the verification result proves valid, updating the latest version with the update plan ;
The pretreatment is
an exclusion process for analyzing the new seismic motion data and excluding the seismic motion data when it is inappropriate to register the data in the database;
The exclusion process includes:
a first exclusion process for excluding seismic motion data when a signal-to-noise ratio is smaller than a predetermined value in a time history waveform of the seismic motion observation record included in the new seismic motion data;
a second exclusion process for excluding seismic motion data when another earthquake or another event is mixed in the time history waveform; and
and a third exclusion process of excluding seismic motion data when the recording time of the seismic motion observation record is insufficient in the time history waveform.
Method for providing earthquake motion evaluation model. The first exclusion process includes:
inputting image data based on the time history waveform of the seismic motion observation record included in the new seismic motion data into a first time history waveform evaluation model that has been machine-learned to determine a correlation between image data and the abnormal state based on the time history waveform to which the abnormal state related to the signal-to-noise ratio is assigned as a correct answer label, and excluding the seismic motion data based on the abnormal state related to the signal-to-noise ratio output from the first time history waveform evaluation model;
The second exclusion process includes:
inputting image data based on the time history waveform of the seismic motion observation record included in the new seismic motion data into a second time history waveform evaluation model that has been machine-learned to determine a correlation between image data based on a time history waveform to which an abnormal state related to the different earthquake or the different event has been assigned as a correct answer label, and excluding the seismic motion data based on the abnormal state related to the different earthquake or the different event output from the second time history waveform evaluation model;
The third exclusion process includes:
image data based on the time history waveform of the seismic motion observation record included in the new seismic motion data is input to a third time history waveform evaluation model, which is configured to machine-learn a correlation between image data and the abnormal state using image data based on a time history waveform to which the abnormal state related to the recording time is assigned as a correct answer label, and the seismic motion data is excluded based on the abnormal state related to the recording time output from the third time history waveform evaluation model.
The method for providing a seismic motion evaluation model according to claim 1 . A method for providing an earthquake motion evaluation model, comprising the steps of: performing machine learning using earthquake motion data, which is registered in a database in association with earthquake motion characteristic parameters and earthquake motion observation records, as learning data; and providing an updated version of an earthquake motion evaluation model based on the machine learning to an external party, the method comprising the steps of:
a database updating step of collecting new seismic motion data and updating the database with the new seismic motion data obtained by performing a predetermined preprocessing on the new seismic motion data;
a machine learning process for acquiring new learning data from the updated database, performing the machine learning, and generating an update plan for the seismic motion evaluation model by machine learning the correlation between the seismic motion characteristic parameters and the seismic motion index obtained from the seismic motion observation record;
a model updating step of performing a predetermined verification process on the update plan and, if the verification result proves valid, updating the latest version with the update plan ;
The verification process includes:
a first verification process for verifying the update plan by comparing a predicted value of the seismic motion index based on the update plan with a predicted value of the seismic motion index based on the latest version at present or in the past;
Method for providing earthquake motion evaluation model. A computer comprising:
A control unit that executes each step included in the method for providing a seismic motion evaluation model according to any one of claims 1 to 3 .
A device providing a seismic motion evaluation model.

Images