Method and system for predicting seismic parameters at unobserved locations

Abstract

The application relates to the technical field of earthquake monitoring, in particular to a non-observation position earthquake parameter prediction method and a prediction system, and solves the technical problem of low prediction accuracy of ground motion parameters in the prior art without setting a strong earthquake station position. The non-observation position earthquake parameter prediction method provided by the application maintains the original measured data record of the data, fully utilizes the spatial position relationship between the target position and the surrounding position and the earthquake propagation path, can fully characterize the complex changes in the earthquake event, and thus improves the accuracy of the prediction value, so as to meet the requirements of rapid earthquake damage assessment, structure response analysis and engineering seismic application.

Description

Technical Field

[0001] This application relates to the field of earthquake monitoring technology, and more specifically, to methods and systems for predicting earthquake parameters at non-observation locations. Background Technology

[0002] Seismic ground motion parameters are crucial fundamental data in earthquake engineering, seismic damage assessment, and structural seismic analysis. They are essential for understanding earthquake focal characteristics, formation mechanisms, and predicting strong ground motion and seismic activity trends. Common parameters include peak ground acceleration (PGA), peak ground velocity (PGV), cumulative absolute velocity (CAV), Arias intensity, and response spectrum acceleration at different periods. After an actual earthquake, seismic ground motion parameters are typically calculated from strong-motion stations. However, due to the limited and uneven spatial distribution of these stations, some geographical locations lack strong-motion stations, making it difficult to directly obtain seismic ground motion parameters for these locations. Therefore, it is necessary to predict the seismic ground motion parameters for these geographical locations based on real-time seismic ground motion parameters detected by strong-motion stations located in the surrounding areas.

[0003] Existing technologies typically employ methods such as seismic motion prediction equations, spatial interpolation methods, machine learning, and graph neural networks to predict seismic motion parameters at locations where strong-motion stations have not been set up.

[0004] However, the prediction data are all based on ground motion parameters already detected by surrounding strong-motion stations, such as peak ground acceleration (PGA), peak ground velocity (PGV), and cumulative absolute velocity (CAV). These ground motion parameters are scalar results obtained by time-history compression of ground motion acceleration, discarding detailed information from the original measured records. This makes it difficult to fully reflect the dynamic characteristics of ground motion, such as waveform evolution, frequency composition, energy accumulation, and duration, resulting in incomplete prediction data. Furthermore, ground motion is influenced by source characteristics, propagation path, local site conditions, and spatial correlation. Relying solely on ground motion indicators such as magnitude, epicentral distance, and site parameters is insufficient to fully characterize the complex changes within a single earthquake event, thus reducing the prediction accuracy of ground motion parameters in locations where strong-motion stations are not located. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for predicting earthquake parameters at non-observation locations, which solves the technical problem of low accuracy in predicting ground motion parameters at locations where strong-motion stations are not located in the prior art.

[0006] As a first aspect of this application, this application provides a method for predicting seismic parameters at non-observation locations, including: Based on the time histories of ground motion accelerations in two horizontal directions corresponding to earthquake events detected by multiple observation stations around the target location, the time histories embedding features of each of the observation stations are determined. Construct a station group map between the target location and the multiple observed stations, the station group map being used to represent the physical relationship between the target location and the multiple observed stations; The node features of the target location are constructed based on the time history embedding features of the observed stations, the target location, and the station physical information of the multiple observed stations. The node features of the target location and the station group map are input into the edge feature-aware graph attention network for information aggregation to obtain the comprehensive feature vector of the target location; Based on the comprehensive feature vector of the target location, the predicted value of the ground motion intensity index of the target location is determined; wherein, the ground motion intensity index includes at least one of: peak ground acceleration, peak ground velocity, cumulative absolute velocity, Arias intensity, and spectral acceleration at different periods.

[0007] In one embodiment of this application, the time history embedding features of each of the observed stations are determined based on the time histories of dual horizontal ground motion accelerations detected by multiple observed stations around the target location, including: The time histories of the dual horizontal ground motion accelerations detected by each observed station are input into a time history feature extraction network for calculation to extract the time history embedding features of each observed station.

[0008] In one embodiment of this application, the temporal feature extraction network includes a one-dimensional convolutional residual network; Specifically, the time histories of the dual horizontal ground motion accelerations detected by each observed station are input into a time history feature extraction network for calculation, in order to extract the time history embedding features of each observed station, including: The time histories of the dual horizontal ground motion accelerations detected by each observed station are input into the one-dimensional convolutional residual network for joint encoding to obtain the time histories embedding features of each observed station. The time histories embedding features include at least one of amplitude evolution, frequency distribution, energy accumulation, and vibration duration.

[0009] In one embodiment of this application, constructing a station group map between the target location and multiple observed stations includes: Based on the differences in site parameters, distance parameters, and propagation paths between the target location and the observed stations, and the spatial distance between the target location and the observed stations, edge features between the target location and the observed stations are constructed. Construct an edge index between the target location and the observed stations based on the connection relationship between the target location and the multiple observed stations; Based on the edge features between the target location and the observed stations, and the edge indexes between the target location and the observed stations, a station group map is constructed between the target location and the observed stations.

[0010] In one embodiment of this application, constructing the edge features between the target location and the observed stations includes: Based on the target location and the site parameters of the observed stations, calculate the initial site parameter difference between the target location and the observed stations; Based on the distance parameters between the target location and the observed stations, calculate the initial distance parameter difference between the target location and the observed stations; Obtain the initial spatial distance between the target location and the observed stations; The initial site parameter differences, the initial distance parameter differences, and the initial station spatial distance are subjected to nonlinear transformation to obtain the site parameter differences, distance parameter differences, and station spatial distances.

[0011] In one embodiment of this application, the station map includes a first directed edge from the target location to each of the observed stations, and a second directed edge from each of the observed stations to the target location.

[0012] In one embodiment of this application, the node features of the target location are constructed based on the time-history embedding features of the observed stations, the target location, and the station physical information of the plurality of observed stations, including: Based on the target location and the station physical information of multiple observed stations, a target node auxiliary feature is constructed for the target location. The target node auxiliary feature includes at least two combinations of magnitude, focal depth, distance parameter, and site parameter. The temporal embedding features of each of the observed stations are fused with the auxiliary features of the target node at the target location to obtain the node features of the target location.

[0013] In one embodiment of this application, the distance parameters include the epicentral distance of the target location and the epicentral distance between observed stations; The site parameters include the average shear wave velocity within 30 meters below the surface at the target location and the average shear wave velocity within 30 meters below the surface at the observed stations.

[0014] In one embodiment of this application, determining the predicted value of the seismic intensity index of the target location based on the comprehensive feature vector of the target location includes: Based on the comprehensive feature vector of the target location, the calculated values ​​of the seismic intensity index in the two horizontal directions of the target location are determined respectively; Calculate the geometric mean of the calculated values ​​of the seismic intensity index in two horizontal directions at the target location, where the geometric mean is the predicted value of the seismic intensity index at the target location.

[0015] As a second aspect of this application, this application also provides a seismic parameter prediction system for non-observation locations, comprising: A time history feature extraction network is applied to determine the time history embedding features of each of the observed stations based on the time histories of ground motion accelerations in two horizontal directions in earthquake events detected by multiple observed stations around the target location. A construction unit is used to construct a station group map between the target location and the multiple observed stations, the station group map being used to represent the physical relationship between the target location and the multiple observed stations; The feature fusion unit is used to construct the node features of the target location based on the time history embedding features of the observed stations, the target location, and the station physical information of the multiple observed stations. An edge-feature-aware graph attention network is used to aggregate information from the node features of the target location and the station map to obtain a comprehensive feature vector of the target location. The calculation unit determines the predicted value of the ground motion intensity index of the target location based on the comprehensive feature vector of the target location; wherein the ground motion intensity index includes at least one of peak ground acceleration, peak ground velocity, cumulative absolute velocity, Arias intensity, and spectral acceleration at different periods.

[0016] The method for predicting seismic parameters at non-observation locations provided in this application first obtains the time histories of ground acceleration in both horizontal directions corresponding to seismic events detected by observed stations around the target location (i.e., the non-observation location). Then, a station map is constructed based on the physical information of the target location and surrounding observed stations to reflect the spatial relationship and propagation path between the target location and the surrounding observed stations. Finally, the time histories of ground acceleration in both horizontal directions corresponding to seismic events detected by each observed station, along with the station map, are input into a pre-constructed edge-feature-aware graph attention network for calculation, obtaining a comprehensive feature vector of the target location. Based on this comprehensive feature vector, the predicted value of the ground motion intensity index at the target location is determined. This application uses the initial data (time histories of ground acceleration in both horizontal directions) that were actually detected, preserving the original measured data, and fully utilizes the spatial relationship between the target location and surrounding locations, as well as the seismic propagation path. This allows for a comprehensive characterization of the complex changes in seismic events, thereby improving the accuracy of the predicted values ​​to meet the needs of rapid seismic damage assessment, structural response analysis, and seismic engineering applications. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 The diagram shown is a flowchart illustrating a seismic parameter prediction method for non-observation locations provided in an embodiment of this application.

[0019] Figure 2 The diagram shown is a flowchart illustrating a seismic parameter prediction method for non-observation locations provided in another embodiment of this application.

[0020] Figure 3 The diagram shown is a flowchart illustrating a seismic parameter prediction method for non-observation locations provided in another embodiment of this application.

[0021] Figure 4 The diagram shown is a flowchart illustrating a seismic parameter prediction method for non-observation locations provided in another embodiment of this application.

[0022] Figure 5 The diagram shown is a flowchart illustrating a seismic parameter prediction method for non-observation locations provided in another embodiment of this application.

[0023] Figure 6 The figure shown is a comparison between the predicted PGA and the measured PGA obtained by the seismic parameter prediction method at non-observation locations in this application.

[0024] Figure 7 The figure shown is a comparison between the predicted Arias obtained by the seismic parameter prediction method at non-observation locations in this application and the measured Arias.

[0025] Figure 8 The figure shown is a comparison between the predicted CAV and the measured CAV obtained by the seismic parameter prediction method at non-observation locations in this application.

[0026] Figure 9 The figure shown is a comparison between the predicted PGV and the measured PGV obtained by the seismic parameter prediction method at non-observation locations in this application.

[0027] Figure 10 The diagram shown is a working block diagram of a seismic parameter prediction system for non-observation locations provided in an embodiment of this application.

[0028] Figure 11 The diagram shown is a block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0029] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] In the description of this application, it should be noted that the terms "upper", "lower", "front", "horizontal", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0031] In the description of this application, it should be noted that, unless otherwise explicitly specified and limited, the term "installation" should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral connection; it can refer to a mechanical connection or an electrical connection; it can refer to a direct connection or an indirect connection through an intermediate medium; and it can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0032] The present application will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0033] Exemplary methods As a first aspect of this application, this application provides a method for predicting seismic parameters at non-observation locations. Figure 1 The diagram shown is a flowchart illustrating a seismic parameter prediction method for non-observation locations according to an embodiment of this application. Figure 1 As shown, an embodiment of this application provides a method for predicting seismic parameters at non-observation locations, including the following steps S100-S500: S100: Determine the time history embedding features of each observed station based on the time histories of ground motion accelerations in two horizontal directions corresponding to the seismic events detected by multiple observed stations around the target location. Specifically, the observed stations refer to the locations where strong-motion stations are set up to detect seismic waves at the corresponding locations when an earthquake occurs. Seismic waves include transverse waves and longitudinal waves.

[0034] The target location refers to the location where no strong seismograph station has been set up, also known as the non-observation location.

[0035] Two horizontal directions refer to two directions that form a certain angle with each other, such as the horizontal direction and the vertical direction, which are perpendicular to each other.

[0036] Seismic ground acceleration time history refers to the ground ground acceleration at each detection time in a certain direction during an earthquake. For example, the seismic ground acceleration time histories corresponding to two horizontal directions can be: the seismic ground acceleration time history in the transverse direction and the seismic ground acceleration time history in the longitudinal direction.

[0037] Specifically, time-history embedding features refer to the feature values ​​of relevant features calculated based on the time history of ground motion acceleration as a function of each detection time. For example, time-history embedding features can include amplitude evolution (i.e., the process of amplitude change over time), frequency distribution (the distribution of waves of different frequencies contained in the seismic wave), energy accumulation (the integral of seismic energy over time), and duration of vibration.

[0038] Optionally, the number of observation stations located around the target location can be determined based on the distance and positional relationship of the existing observation stations around the target location. When there are a large number of observation stations around the target location, multiple stations are selected from among them. For example, if there are four observation stations, the area around the target location is divided into four regions centered on the target location, and the four observation stations are located within these four regions respectively. It should be noted that the straight-line distance between each observation station and the target location can be all the same, all different, or partially the same and partially different.

[0039] S200: Construct a station group map between the target location and multiple observed stations. The station group map is used to represent the physical relationship between the target location and multiple observed stations. Optionally, the station group map includes a first directed edge from the target location to each observed station and a second directed edge from each observed station to the target location, but no edge is established between two adjacent observed stations, that is, the station group map is a two-way star map structure.

[0040] The features corresponding to each edge in the station group map represent the physical relationship between the target location and the observed stations. The physical relationship may include: the distance between the target location and the observed stations, the difference in epicentral distance between the target location and the observed stations, the difference in average shear wave velocity within 30 meters below the ground surface, etc.

[0041] S300: Construct node features of the target location based on the time history embedding features of the observed stations, the target location, and the physical information of multiple observed stations; Specifically, the physical information of the observed stations includes at least two combinations of magnitude, focal depth, distance parameters, and site parameters.

[0042] The distance parameters include the epicentral distance of the target location, the epicentral distance of each observed station, and the distance between the target location and the observed stations. The difference in epicentral distance between the target location and the observed stations is calculated using the epicentral distance from the distance parameters.

[0043] Site parameters include: the average shear wave velocity within a 30-meter radius below the surface at the target location (i.e., V S30 The average shear wave velocity within 30 meters below the surface of each observed station (i.e., V S30 The distance between the target location and the observed stations can be calculated using site parameters. V S30 Difference (i.e., the difference in average shear wave velocity within a 30-meter radius below the Earth's surface), for example, at the target location V S30 The speed is 500 m / s, and the observed stations are... V S30 If the speed is 300 m / s, then the position of the target is relative to the position of the observed stations. V S30 The difference is equal to |500-300|.

[0044] S400: Input the node features of the target location and the station group map into the edge feature-aware graph attention network for information aggregation to obtain the comprehensive feature vector of the target location; Specifically, the edge-feature-aware graph attention network was pre-trained and validated. The training and validation sets were constructed based on the ground motion parameters from observation stations of known seismic events.

[0045] The comprehensive feature vector of the target location is the output data of an edge-feature-aware graph attention network, which integrates the target location's own physical information, the physical information of multiple surrounding observed stations, the seismic acceleration time histories of multiple surrounding observed stations, the distance between the target location and each observed station, the epicentral distance difference, and other information. V S30 difference.

[0046] S500: Based on the comprehensive feature vector of the target location, determine the predicted value of the seismic intensity index at the target location; Once the comprehensive feature vector is obtained, the predicted value of the seismic intensity index at the target location can be determined based on the comprehensive feature vector. The seismic intensity index includes at least one of the following: Peak Ground Acceleration (PGA), Peak Ground Velocity (PGV), Cumulative Absolute Velocity (CAV), Arias intensity, and spectral acceleration at different periods. For example, the PGA and PGV at the target location can be obtained.

[0047] Optionally, S500 (determining the predicted value of the seismic intensity index at the target location based on the comprehensive feature vector of the target location) specifically includes the following steps: S510: Based on the comprehensive feature vector of the target location, determine the calculated values ​​of the seismic intensity index in the two horizontal directions of the target location respectively; S520: Calculate the geometric mean of the calculated values ​​of the seismic intensity index in two horizontal directions at the target location. The geometric mean is the predicted value of the seismic intensity index at the target location.

[0048] The predicted value is the geometric mean of the calculated values ​​of the earthquake intensity indices in two horizontal directions.

[0049] The method for predicting seismic parameters at non-observation locations provided in this application first obtains the time histories of ground acceleration in both horizontal directions corresponding to seismic events detected by observed stations around the target location (i.e., the non-observation location). Then, a station map is constructed based on the physical information of the target location and surrounding observed stations to reflect the spatial relationship and propagation path between the target location and the surrounding observed stations. Finally, the time histories of ground acceleration in both horizontal directions corresponding to seismic events detected by each observed station, along with the station map, are input into a pre-constructed edge-feature-aware graph attention network for calculation, obtaining a comprehensive feature vector of the target location. Based on this comprehensive feature vector, the predicted value of the ground motion intensity index at the target location is determined. This application uses the initial data (time histories of ground acceleration in both horizontal directions) that were actually detected, preserving the original measured data, and fully utilizes the spatial relationship between the target location and surrounding locations, as well as the seismic propagation path. This allows for a comprehensive characterization of the complex changes in seismic events, thereby improving the accuracy of the predicted values ​​to meet the needs of rapid seismic damage assessment, structural response analysis, and seismic engineering applications.

[0050] In one embodiment of this application, as Figure 2 As shown, S100 (determining the time history embedding features of each observed station based on the time histories of dual horizontal ground motion accelerations detected by multiple observed stations around the target location) specifically includes the following step S110: S110: Input the time history of the dual horizontal ground motion acceleration detected by each observed station into the time history feature extraction network for calculation, so as to extract the time history embedding features of each observed station.

[0051] Specifically, the temporal feature extraction network was pre-built.

[0052] Optionally, the temporal feature extraction network includes a one-dimensional convolutional residual network; S110 (inputting the time histories of the dual horizontal ground motion accelerations detected by each observed station into the time history feature extraction network for calculation, in order to extract the time history embedding features of each observed station) specifically includes the following step S111: S111: Input the time history of the dual horizontal ground motion acceleration detected by each observed station into a one-dimensional convolutional residual network for joint encoding to obtain the time history embedding features of each observed station. The time history embedding features include at least one of amplitude evolution, frequency distribution, energy accumulation, and vibration duration.

[0053] The amplitude evolution refers to the process of the peak value changing over time, and may include at least one of the following: the evolution trend of the peak ground acceleration (PGA) over time, the evolution trend of the peak ground velocity (PGV) over time, the evolution trend of the peak effective peak acceleration (EPA) over time, the evolution trend of the peak intensity (SI) over time, and the evolution trend of the peak displacement (Sd) over time. The peak evolution process can be used to determine structural parameters such as the maximum inertial force at unobserved locations. For example, the PGA can reflect the maximum inertial force experienced at unobserved locations, and the degree of damage and damage mode at unobserved locations can be determined based on the evolution process.

[0054] Frequency distribution refers to the distribution of wave content of different frequencies included in a seismic wave. It is usually represented by Fourier spectrum or response spectrum.

[0055] Energy accumulation refers to the integral of the square of the seismic wave acceleration over time, representing the total energy input from the seismic motion into the structural system. It can reflect whether the structure will collapse due to insufficient energy dissipation.

[0056] The duration of a seismic event refers to the length of time from the start to the end of an earthquake. It typically refers to the effective duration, which is the length of the strong vibrations in the middle after excluding the weak initial and final vibrations. The duration of the seismic event reflects the number of repeated loading cycles a structure has undergone. For example, if the seismic event lasts longer than 30 seconds, the structure will experience fatigue and cumulative damage due to repeated tension and compression.

[0057] In one embodiment of this application, as Figure 3 As shown, S200 (constructing a station group map between the target location and multiple observed stations) specifically includes the following steps S210-S230: S210: Based on the differences in site parameters, distance parameters, and propagation paths between the target location and the observed stations, and the spatial distance between the target location and the observed stations, construct the edge features between the target location and the observed stations; For example, site parameters include the average shear wave velocity within a 30-meter radius below the surface. Distance parameters include the epicentral distance.

[0058] Correspondingly, S210 constructs the edge features between the target location and the observed stations based on the differences in average shear wave velocity, epicentral distance, and propagation path within 30 meters below the surface between the target location and the observed stations, as well as the spatial distance between the target location and the observed stations. S220: Construct an edge index between the target location and multiple observed stations based on the connection relationship between the target location and multiple observed stations; S230: Construct a station group map between the target location and the observed stations based on the edge features and edge indexes between the target location and the observed stations.

[0059] Optionally, S210 (constructing the edge features between the target location and the observed stations) specifically includes the following steps S211-S214: S211: Calculate the initial site parameter difference between the target location and the observed stations based on the site parameters of the target location and the observed stations; For example, calculate the difference in average shear wave velocity within an initial 30-meter range below the Earth's surface between the target location and the observed stations.

[0060] S212: Calculate the difference in initial distance parameters between the target location and the observed stations based on the distance parameters between the target location and the observed stations; Specifically, the distance parameters include the initial epicentral distance of the target location, the initial epicentral distance of the observed stations, and the initial spatial distance between the target location and the observed stations.

[0061] For example, calculate the difference in initial epicentral distance and initial propagation path between the target location and the observed stations.

[0062] The differences in initial epicentral distance and initial propagation path can be calculated based on the initial epicentral distance of the target location and the initial epicentral distance of the observed stations. S213: Obtain the initial spatial distance between the target location and the observed stations; The initial spatial distance between stations is calculated based on the initial spatial distance between the target location and the observed stations.

[0063] S214: Perform nonlinear transformation on the initial site parameter differences, initial distance parameter differences, and initial station spatial distances to obtain the site parameter differences, distance parameter differences, and station spatial distances.

[0064] Specifically, nonlinear transformations include at least one of logarithmic transformation, square root transformation, power transformation, or normalization transformation.

[0065] In one embodiment of this application, as Figure 4As shown, S300 (constructing the node features of the target location based on the time history embedding features of the observed stations, the target location, and the physical information of multiple observed stations) specifically includes the following steps S310-S320: S310: Construct target node auxiliary features for the target location based on the target location and the station physical information of multiple observed stations. The target node auxiliary features include at least two combinations of magnitude, focal depth, distance parameters, and site parameters. Specifically, the auxiliary features of the target node include at least two combinations of magnitude, focal depth, distance parameters, and site parameters, for example: (1) The auxiliary features of the target node include two combinations of magnitude, focal depth, distance parameter and site parameter. For example, it may include magnitude and focal depth; or magnitude and distance parameter; or magnitude and site parameter; or focal depth and distance parameter; or focal depth and site parameter; or distance parameter and site parameter.

[0066] (2) The auxiliary features of the target node include three combinations of magnitude, focal depth, distance parameter and site parameter. For example, it may include magnitude, focal depth and site parameter; or it may include magnitude, focal depth and distance parameter; or it may include magnitude, distance parameter and site parameter; or it may include focal depth, distance parameter and site parameter. (3) The auxiliary features of the target node include four combinations of magnitude, focal depth, distance parameters and site parameters, namely magnitude, focal depth, distance parameters and site parameters.

[0067] The distance parameters include the epicentral distance of the target location and the epicentral distance of each observed station. The difference in epicentral distance between the target location and the observed stations is calculated using the epicentral distance from the distance parameters.

[0068] Site parameters include: the average shear wave velocity within a 30-meter radius below the surface at the target location (i.e., V S30 The average shear wave velocity within 30 meters below the surface of each observed station (i.e., V S30 The distance between the target location and the observed stations can be calculated using site parameters. V S30 Difference (i.e., the difference in average shear wave velocity within a 30-meter radius below the Earth's surface), for example, at the target location V S30 The speed is 500 m / s, and the observed stations are... V S30 If the speed is 300 m / s, then the position of the target is relative to the position of the observed stations. V S30 The difference is equal to |500-300|.

[0069] S320: The time history embedding features of each observed station are fused with the auxiliary features of the target node at the target location to obtain the node features at the target location.

[0070] This application integrates the temporal history embedding features of the observation stations around the target location with physical information such as magnitude, focal depth, distance parameters, and site parameters to form the node features of a graph neural network.

[0071] The method for predicting seismic parameters at non-observation locations provided in this application is further described below with specific embodiments.

[0072] Example 1: The distance parameters are the epicentral distance, the distance between the target location and the observed stations, the site parameters are the average shear wave velocity within 30 meters below the ground surface, and the ground motion intensity indicators include: peak ground acceleration (PGA), peak ground velocity (PGV), cumulative absolute velocity (CAV), Arias intensity, and spectral acceleration at different periods.

[0073] Figure 5 The diagram shown is a flowchart illustrating a seismic parameter prediction method for non-observation locations provided in this application. Figure 5 As shown, the seismic parameter prediction method for non-observation locations provided in this application includes the following steps S10-S80: S10: Input the time history of the dual horizontal ground motion acceleration detected by each observed station into the time history feature extraction network, perform joint encoding, and obtain the time history embedding features of each observed station. The time history embedding features include amplitude evolution, frequency distribution, energy accumulation, and vibration duration. S20: Based on the differences in epicentral distance, propagation path, and average shear wave velocity within 30 meters below the surface between the target location and the observed stations, and the spatial distance between the target location and the observed stations, construct the edge features between the target location and the observed stations. S30: Construct an edge index between the target location and multiple observed stations based on the connection relationship between the target location and multiple observed stations; S40: Construct a station group map between the target location and the observed stations based on the edge features and edge indexes between the target location and the observed stations; S50: Based on the target location and the station physical information of multiple observed stations, construct the target node auxiliary features of the target location. The target node auxiliary features include magnitude, focal depth, epicentral distance, and average shear wave velocity within 30 meters below the ground surface. S60: The time history embedding features of each observed station are fused with the auxiliary features of the target node at the target location to obtain the node features at the target location; S70: Input the node features of the target location and the station group map into the edge feature-aware graph attention network for information aggregation calculation to obtain the comprehensive feature vector of the target location; S80: Based on the comprehensive feature vector of the target location, determine the predicted values ​​of the ground motion intensity indices for the target location. These indices include: Peak Ground Acceleration (PGA), Peak Ground Velocity (PGV), Cumulative Absolute Velocity (CAV), Arias intensity, and spectral acceleration (Sa) at different periods.

[0074] The predicted values ​​of the obtained seismic intensity indices are shown in Table 1 below. In Table 1, PGA refers to peak ground acceleration, PGV refers to peak ground velocity, CAV refers to cumulative absolute velocity, Arias Intensity refers to Arias intensity, and Sa refers to spectral acceleration at different periods, for example, Sa _ 0.1 refers to the spectral acceleration with a period of 0.1 seconds, Sa. _ 3.0 refers to the spectral acceleration with a period of 3 seconds. R 2 This refers to the degree of agreement between predicted and actual values. MSE refers to the mean squared error between predicted and actual values, RMSE refers to the root mean square error between predicted and actual values, and MAE refers to the mean absolute error between predicted and actual values. Among these, PGA, Arias intensity, CAV, and PGV correspond to R... 2 MSE, RMSE, and MAE, i.e., the comparison chart between actual and predicted values, are shown below. Figures 6-9 As shown.

[0075] Table 1: Predicted values ​​of seismic motion intensity index

[0076] As shown in Table 1, combined Figures 6-9 As shown, the prediction method of this application has an average R-value of various indicators. 2 The average Pearson correlation coefficient was 0.9160, with a value of 0.8376, indicating good multi-indicator predictive ability and a high degree of consistency between predicted and actual values. Specifically, the R-value of CAV was [missing information]. 2 Reaching 0.9218, the R of PGV 2 The R value of Arias intensity reaches 0.8487. 2 The accuracy reached 0.8906, indicating high prediction accuracy.

[0077] Comparative Example 1: Kriging Interpolation Algorithm The prediction method for predicting the intensity index of ground motion using the Kriging interpolation algorithm includes the following steps: (1) Calculate the value of the ground motion intensity index corresponding to each observed station based on the time history of the dual horizontal ground motion acceleration detected by each observed station; (2) Based on the Kriging interpolation algorithm, the ground motion intensity index at the target location is predicted according to the values ​​of the ground motion intensity index corresponding to each observed station and the distance between the target location and the observed stations, so as to obtain the predicted value of the ground motion intensity index at the target location. Among them, the ground motion intensity index includes: peak ground acceleration PGA, peak ground velocity PGV, cumulative absolute velocity CAV, Arias intensity, and spectral acceleration Sa under different periods.

[0078] Comparative Example 2: IDW Interpolation Algorithm The prediction method for predicting the intensity index of seismic ground motion using the IDW interpolation algorithm includes the following steps: (1) Calculate the value of the ground motion intensity index corresponding to each observed station by taking the time history of the dual horizontal ground motion acceleration detected by each observed station; (2) Based on the IDW interpolation algorithm, the ground motion intensity index at the target location is predicted according to the values ​​of the ground motion intensity index corresponding to each observed station and the distance between the target location and the observed stations, so as to obtain the predicted value of the ground motion intensity index at the target location. Among them, the ground motion intensity index includes: peak ground acceleration PGA, peak ground velocity PGV, cumulative absolute velocity CAV, Arias intensity, and spectral acceleration Sa under different periods.

[0079] The difference between Example 1 and Comparative Examples 1 and 2 is as follows: (1) The data source of Comparative Examples 1 and 2 is the value of the ground motion intensity index corresponding to the observed stations, while the data source of Example 1 is: the original data detected by the observed stations: the time history of the two horizontal ground motion accelerations. (2) Example 1 takes into account the station group map between the target location and the observed stations, while Comparative Examples 1 and 2 only take into account the distance between the target location and the observed stations.

[0080] The comparison results of the predicted values ​​of the seismic intensity index obtained in Example 1 with the predicted values ​​of the seismic intensity index obtained in Comparative Examples 1 and 2 are shown in Table 2 below: Table 2: Comparison of predicted values ​​for Example 1, Comparative Example 1, and Comparative Example 2

[0081] Comparing the predicted values ​​of Example 1 with those of Comparative Example 1 (Kriging interpolation algorithm) and Comparative Example 2 (IDW interpolation algorithm) in Table 2, we can see that: Taking PGA and PGV as examples, Example 1 in R 2 In terms of metrics such as MSE, RMSE, and MAE, it outperforms two traditional interpolation methods: Kriging interpolation and IDW interpolation. Specifically, in PGA prediction, the R value corresponding to Example 1... 2 The value is 0.7924, higher than the 0.5481 for the Kriging interpolation algorithm and the 0.6867 for the IDW interpolation algorithm. In PGV prediction, the R value corresponding to Example 1 is... 2 It is 0.8487, which is higher than the Kriging interpolation algorithm's 0.7378 and the IDW interpolation algorithm's 0.7728.

[0082] Comparative Example 3: DNN Model (Non-Graph Neural Network Model) Prediction The method for predicting seismic parameters at a target location (not a location where observation stations are set up) using a DNN model includes the following steps: (1) Input the time history of the dual horizontal ground motion acceleration detected by each observed station into the time history feature extraction network and perform joint encoding to obtain the time history embedding features of each observed station. The time history embedding features include amplitude evolution, frequency distribution, energy accumulation, and vibration duration. (2) Input the time history embedding features of each observed station into the DNN model for prediction to obtain the predicted values ​​of the ground motion intensity indices at the target location. Among them, the ground motion intensity indices include: peak ground acceleration (PGA), peak ground velocity (PGV), cumulative absolute velocity (CAV), Arias intensity, and spectral acceleration (Sa) at different periods.

[0083] The difference between Comparative Example 3 and Example 1 is that, although both Comparative Example 3 and Example 1 input the time histories of the dual horizontal ground motion accelerations detected by the observed stations into a time history extraction network to extract time history embedding features, Comparative Example 3 does not construct a station group map between the target location and the surrounding observed stations; it only inputs the time history embedding features into a DNN model for prediction. Example 1, on the other hand, inputs both the node features of the target location and the station group map into an edge feature-aware graph attention network for calculation.

[0084] The comparison results of the predicted values ​​of the seismic intensity index obtained in Example 1 and the predicted values ​​of the seismic intensity index obtained in Comparative Example 3 are shown in Table 3 below: Table 3: Comparison of predicted values ​​for Example 1 and Comparative Example 3

[0085] As shown in Table 3, the predicted values ​​of the two ground motion intensity indices, PGA and PGV, obtained in Example 1 are significantly better than those of the DNN model.

[0086] In the prediction of PGA, the corresponding R of this application is 2 It is 0.7924, which is higher than the 0.2726 corresponding to DNN.

[0087] In the prediction of PGV, the corresponding R of this application 2 It is 0.8487, which is higher than the 0.5970 corresponding to DNN.

[0088] As can be seen from the comparative data in Table 3, the prediction method of this application can more effectively express the spatial relationship and influence weight between the target location and the surrounding observed stations, which helps to improve the prediction effect of ground motion parameters at non-observed locations.

[0089] Exemplary System As a second aspect of this application, this application also provides a seismic parameter prediction system for non-observation locations. Figure 10 The diagram shown is a working block diagram of a seismic parameter prediction system for non-observation locations provided in an embodiment of this application. Figure 10 As shown, an embodiment of this application provides a seismic parameter prediction system 1 for non-observation locations, comprising: The time history feature extraction network 11 is applied to determine the time history embedding features of each observed station based on the time histories of ground motion acceleration corresponding to two horizontal directions in the earthquake events detected by multiple observed stations around the target location. Construction unit 12 is used to construct a station group map between the target location and multiple observed stations. The station group map is used to represent the physical relationship between the target location and multiple observed stations. The feature fusion unit 13 is used to construct the node features of the target location based on the time history embedding features of the observed stations, the target location, and the station physical information of multiple observed stations. The edge feature-aware graph attention network 14 is used to aggregate and calculate the node features of the target location and the station group map information to obtain the comprehensive feature vector of the target location. The calculation unit 15 determines the predicted value of the ground motion intensity index of the target location based on the comprehensive feature vector of the target location; wherein the ground motion intensity index includes at least one of the following: peak ground acceleration, peak ground velocity, cumulative absolute velocity, Arias intensity, and spectral acceleration under different periods.

[0090] The earthquake parameter prediction system provided in this application for non-observation locations inputs the time histories of ground motion acceleration in both horizontal directions corresponding to earthquake events detected by observed stations around the target location (i.e., non-observation location) into a time history feature extraction network for calculation, obtaining the time history embedding features of each observed station; then, a station group map is constructed based on the physical information of the target location and the surrounding observed stations to reflect the spatial relationship and propagation path between the target location and the surrounding observed stations; finally, the time histories of ground motion acceleration in both horizontal directions corresponding to earthquake events detected by each observed station and the station group map are input together into a pre-constructed edge feature-aware graph attention network for calculation, obtaining the comprehensive feature vector of the target location, and determining the predicted value of the ground motion intensity index of the target location based on the comprehensive feature vector. This application uses the initial data actually detected (the time histories of ground motion acceleration corresponding to the two horizontal directions), maintains the original measured record of the data, and makes full use of the spatial relationship between the target location and the surrounding location as well as the earthquake propagation path. It can fully characterize the complex changes in the earthquake event, thereby improving the accuracy of the predicted value to meet the needs of rapid earthquake damage assessment, structural response analysis and seismic engineering applications.

[0091] Exemplary electronic devices As a fourth aspect of this application, this application also provides an electronic device, such as... Figure 11 As shown, it includes a processor and a memory. The memory stores a computer program. When the processor executes the computer program, it performs the aforementioned method for predicting seismic parameters at non-observation locations.

[0092] Specifically, the electronic device includes a processor, memory, network interface, and input device connected via a device bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores operating devices and computer programs. The internal memory provides an environment for the operation of the operating devices and computer programs in the non-volatile storage medium. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it follows the steps of a seismic parameter prediction method for non-observation locations according to various embodiments of this specification, as described in the above embodiments.

[0093] The processor may include the main processor, as well as baseband chips, modems, etc.

[0094] The memory stores a program for executing the technical solution of this application, and may also store operating devices and other key business functions. Specifically, the program may include program code, which includes computer operation instructions. More specifically, the memory may include read-only memory (ROM), other types of static storage devices capable of storing static information and instructions, random access memory (RAM), other types of dynamic storage devices capable of storing information and instructions, disk storage, flash memory, etc.

[0095] The processor can be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the execution of the program of the present application. It can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0096] Input devices may include devices that receive data and information input by the user, such as keyboards, mice, cameras, scanners, light pens, voice input devices, touch screens, pedometers, or gravity sensors.

[0097] Output devices may include devices that allow information to be output to the user, such as displays, printers, speakers, etc.

[0098] The communication interface may include any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0099] The processor executes the program stored in the memory and calls other devices, which can be used to implement the various steps of any of the seismic parameter prediction methods for non-observation locations provided in the embodiments described above.

[0100] The electronic device may also include a display component and a voice component. The display component may be a liquid crystal display or an e-ink display. The input device of the controller may be a touch layer covering the display component, or a button, trackball, or touchpad set on the controller housing, or an external keyboard, touchpad, or mouse, etc.

[0101] Those skilled in the art will understand that Figure 11The structures shown are merely block diagrams of a portion of the structure related to the scheme described in this specification, and do not constitute a limitation on the electronic devices to which the scheme described in this specification is applied. Specific electronic devices may include more or fewer components than those shown in the figures, or may combine certain components, or may have different component arrangements.

[0102] Exemplary computer program products and storage media In addition to the methods and devices described above, the seismic parameter prediction method for non-observation locations provided in the embodiments of this specification can also be a computer program product, which includes computer program instructions that, when executed by a processor, cause the processor to perform the steps in the seismic parameter prediction method for non-observation locations according to various embodiments of this specification as described in the "Exemplary Methods" section above.

[0103] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0104] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this specification. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages.

[0105] Furthermore, embodiments of this specification also provide a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor of the steps in a method for predicting seismic parameters at a non-observation location according to various embodiments of this specification as described in the "Exemplary Methods" section above.

[0106] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this specification can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0107] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0108] The embodiments described above are merely illustrative of several implementation methods outlined in this specification. While the descriptions are specific and detailed, they should not be construed as limiting the scope of the solutions provided in this specification. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this specification, and these all fall within the scope of protection of this specification. Therefore, the scope of protection for this patent should be determined by the appended claims.

Claims

1. A method for predicting seismic parameters at non-observation locations, characterized in that, include: Based on the time histories of ground motion accelerations in two horizontal directions corresponding to earthquake events detected by multiple observation stations around the target location, the time histories embedding features of each of the observation stations are determined. Construct a station group map between the target location and the multiple observed stations, the station group map being used to represent the physical relationship between the target location and the multiple observed stations; The node features of the target location are constructed based on the time history embedding features of the observed stations, the target location, and the station physical information of the multiple observed stations. The node features of the target location and the station group map are input into the edge feature-aware graph attention network for information aggregation to obtain the comprehensive feature vector of the target location; Based on the comprehensive feature vector of the target location, the predicted value of the ground motion intensity index of the target location is determined; wherein, the ground motion intensity index includes at least one of: peak ground acceleration, peak ground velocity, cumulative absolute velocity, Arias intensity, and spectral acceleration at different periods.

2. The prediction method according to claim 1, characterized in that, Based on the time histories of dual horizontal ground motion accelerations detected by multiple observed stations around the target location, the time histories embedding features of each of the observed stations are determined, including: The time histories of the dual horizontal ground motion accelerations detected by each observed station are input into a time history feature extraction network for calculation to extract the time history embedding features of each observed station.

3. The prediction method according to claim 2, characterized in that, The temporal feature extraction network includes a one-dimensional convolutional residual network; Specifically, the time histories of the dual horizontal ground motion accelerations detected by each observed station are input into a time history feature extraction network for calculation, in order to extract the time history embedding features of each observed station, including: The time histories of the dual horizontal ground motion accelerations detected by each observed station are input into the one-dimensional convolutional residual network for joint encoding to obtain the time histories embedding features of each observed station. The time histories embedding features include at least one of amplitude evolution, frequency distribution, energy accumulation, and vibration duration.

4. The prediction method according to claim 1, characterized in that, Constructing a station group map between the target location and the multiple observed stations includes: Based on the differences in site parameters, distance parameters, and propagation paths between the target location and the observed stations, and the spatial distance between the target location and the observed stations, edge features between the target location and the observed stations are constructed. Construct an edge index between the target location and the observed stations based on the connection relationship between the target location and the multiple observed stations; Based on the edge features between the target location and the observed stations, and the edge indexes between the target location and the observed stations, a station group map is constructed between the target location and the observed stations.

5. The prediction method according to claim 4, characterized in that, The construction of the edge features between the target location and the observed stations includes: Based on the target location and the site parameters of the observed stations, calculate the initial site parameter difference between the target location and the observed stations; Based on the distance parameters between the target location and the observed stations, calculate the initial distance parameter difference between the target location and the observed stations; Obtain the initial spatial distance between the target location and the observed stations; The initial site parameter differences, the initial distance parameter differences, and the initial station spatial distance are subjected to nonlinear transformation to obtain the site parameter differences, distance parameter differences, and station spatial distances.

6. The prediction method according to claim 1, characterized in that, The station map includes a first directed edge from the target location to each of the observed stations, and a second directed edge from each of the observed stations to the target location.

7. The prediction method according to claim 1, characterized in that, Based on the time-history embedding features of the observed stations, the target location, and the physical information of multiple observed stations, the node features of the target location are constructed, including: Based on the target location and the station physical information of multiple observed stations, a target node auxiliary feature is constructed for the target location. The target node auxiliary feature includes at least two combinations of magnitude, focal depth, distance parameter, and site parameter. The temporal embedding features of each of the observed stations are fused with the auxiliary features of the target node at the target location to obtain the node features of the target location.

8. The prediction method according to claim 7, characterized in that, The distance parameters include the epicentral distance of the target location and the epicentral distance between observed stations; The site parameters include the average shear wave velocity within 30 meters below the surface at the target location and the average shear wave velocity within 30 meters below the surface at the observed stations.

9. The prediction method according to claim 1, characterized in that, Based on the comprehensive feature vector of the target location, the predicted value of the seismic intensity index at the target location is determined, including: Based on the comprehensive feature vector of the target location, the calculated values ​​of the seismic intensity index in the two horizontal directions of the target location are determined respectively; Calculate the geometric mean of the calculated values ​​of the seismic intensity index in two horizontal directions at the target location, where the geometric mean is the predicted value of the seismic intensity index at the target location.

10. A seismic parameter prediction system for non-observation locations, characterized in that, include: A time history feature extraction network is applied to determine the time history embedding features of each of the observed stations based on the time histories of ground motion accelerations in two horizontal directions in earthquake events detected by multiple observed stations around the target location. A construction unit is used to construct a station group map between the target location and the multiple observed stations, the station group map being used to represent the physical relationship between the target location and the multiple observed stations; The feature fusion unit is used to construct the node features of the target location based on the time history embedding features of the observed stations, the target location, and the station physical information of the multiple observed stations. An edge-feature-aware graph attention network is used to aggregate information from the node features of the target location and the station map to obtain a comprehensive feature vector of the target location. The calculation unit determines the predicted value of the ground motion intensity index of the target location based on the comprehensive feature vector of the target location; wherein the ground motion intensity index includes at least one of peak ground acceleration, peak ground velocity, cumulative absolute velocity, Arias intensity, and spectral acceleration at different periods.

Images