Cumulative absolute velocity prediction method, apparatus, device, medium, and program product
By constructing a group of earthquake monitoring stations and utilizing graph neural networks, combining node features and edge features, the problem of inaccurate cumulative absolute velocity prediction was solved, achieving higher prediction accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING UNIV OF CIVIL ENG & ARCHITECTURE
- Filing Date
- 2025-03-06
- Publication Date
- 2026-06-16
AI Technical Summary
Existing technologies cannot comprehensively consider both historical earthquake records and measured data from earthquake monitoring stations, resulting in inaccurate predictions of cumulative absolute velocity.
A group of seismic monitoring stations is constructed, and the node features and edge features of each station are obtained. The historical station group feature map is trained by a graph neural network. The prediction is performed using a dynamic graph attention layer and a global average pooling layer. The initial results are output by combining a fully connected layer. Finally, the cumulative absolute velocity of the station to be predicted is predicted by the trained graph neural network.
The accuracy of cumulative absolute velocity prediction has been improved by comprehensively analyzing historical earthquake data and measured data from surrounding stations, thus enhancing the precision of the prediction.
Smart Images
Figure CN120065333B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of earthquake data prediction technology, and in particular to a cumulative absolute velocity prediction method, apparatus, equipment, medium, and program product. Background Technology
[0002] Cumulative Absolute Velocity (CAV) is crucial data for earthquake assessment. Current CAV predictions primarily rely on interpolation methods and a combination of earthquake prediction algorithms, making it difficult to comprehensively consider both historical earthquake records and measured data from seismic monitoring stations. This results in inaccuracies in current CAV predictions. Summary of the Invention
[0003] This invention provides a method, apparatus, device, medium, and program product for predicting cumulative absolute velocity, which addresses the shortcomings of existing methods for predicting cumulative absolute velocity as being inaccurate and improves the accuracy of cumulative absolute velocity prediction.
[0004] This invention provides a method for predicting cumulative absolute velocity, comprising the following steps:
[0005] Construct a group of earthquake monitoring stations; the group of earthquake monitoring stations includes target monitoring stations and surrounding stations.
[0006] Obtain the node features of each station in the earthquake monitoring station group, and obtain the edge features between the target monitoring station and the surrounding stations;
[0007] A historical station group feature map is constructed based on the node features and the edge features; the historical station group feature map is used to train a graph neural network.
[0008] The feature map of the station group corresponding to the station to be predicted is input into the trained graph neural network to obtain the cumulative absolute velocity prediction result of the station to be predicted.
[0009] According to a cumulative absolute velocity prediction method provided by the present invention, the construction of a seismic monitoring station group includes:
[0010] Obtain the type and number of historical earthquake monitoring stations, as well as the distance between any two historical earthquake monitoring stations;
[0011] Identify the target monitoring station and surrounding stations among historical earthquake monitoring stations; the surrounding stations are stations around the target monitoring station; the target monitoring station and the surrounding stations are surface stations; the number of historical records of the target monitoring station and the surrounding stations is greater than a first threshold; the distance between the target monitoring station and the surrounding stations is less than a second threshold; the number of surrounding stations is greater than a third threshold.
[0012] Based on the target monitoring station and the surrounding stations, an earthquake monitoring station group is constructed.
[0013] According to the cumulative absolute velocity prediction method provided by the present invention, obtaining the nodal characteristics of each station in the seismic monitoring station group includes:
[0014] Obtain the first node features of the target monitoring station; the first node features include the magnitude, epicentral distance, soil shear wave velocity, and focal depth in the historical records of the target monitoring station;
[0015] Obtain the second node features of the surrounding stations; the second node features include soil shear wave velocity, epicentral distance, and cumulative absolute velocity from the historical records of the surrounding stations.
[0016] According to the cumulative absolute velocity prediction method provided by the present invention, obtaining the edge features between the target monitoring station and the surrounding stations includes:
[0017] Based on differences in site conditions, the first edge characteristics between the target monitoring station and the surrounding stations are determined; the differences in site conditions are determined based on the absolute difference in soil shear wave velocity between the target monitoring station and the surrounding stations.
[0018] Based on spatial relationships, the second-side characteristics between the target monitoring station and the surrounding stations are determined; the spatial relationships are determined based on the absolute difference in epicentral distance between the target monitoring station and the surrounding stations.
[0019] Based on the distance between the target monitoring station and the surrounding stations, the third side feature between the target monitoring station and the surrounding stations is determined.
[0020] According to the cumulative absolute velocity prediction method provided by the present invention, the step of constructing a historical station group feature map based on the node features and the edge features includes:
[0021] By transforming the differences in site conditions and the spatial relationships, nonlinear characteristics are obtained;
[0022] Based on the nonlinear characteristics and the third-side characteristics, the connection relationship between the target monitoring station and the surrounding stations is determined;
[0023] Based on the connection relationship and target topology, a historical station group feature map is constructed; in the target topology, the target monitoring station is connected to the surrounding stations, and no two surrounding stations are connected to each other.
[0024] According to a cumulative absolute velocity prediction method provided by the present invention, the graph neural network includes a dynamic graph attention layer, a global average pooling layer, and a fully connected layer; the cumulative absolute velocity prediction method further includes:
[0025] The feature map of the historical station group is input into the dynamic graph attention layer to obtain the relationships between nodes and the adjusted edge feature weights.
[0026] The relationships between nodes and the adjusted edge feature weights are input into the global average pooling layer to obtain the global feature vector;
[0027] The global feature vector is input into the fully connected layer to obtain the initial prediction result;
[0028] The graph neural network is trained based on the initial prediction results; the training process includes network architecture adjustment and activation function adjustment.
[0029] The present invention also provides a cumulative absolute velocity prediction device, comprising the following modules:
[0030] An earthquake monitoring station group construction module is used to construct an earthquake monitoring station group; the earthquake monitoring station group includes a target monitoring station and surrounding stations;
[0031] The feature acquisition module is used to acquire the node features of each station in the earthquake monitoring station group and to acquire the edge features between the target monitoring station and the surrounding stations.
[0032] A historical station group feature map construction module is used to construct a historical station group feature map based on the node features and the edge features; the historical station group feature map is used to train a graph neural network.
[0033] The cumulative absolute velocity prediction module is used to input the feature map of the station group corresponding to the station to be predicted into the trained graph neural network to obtain the cumulative absolute velocity prediction result of the station to be predicted.
[0034] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the cumulative absolute velocity prediction method as described above.
[0035] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the cumulative absolute velocity prediction method as described above.
[0036] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the cumulative absolute velocity prediction method as described above.
[0037] The cumulative absolute velocity prediction method, apparatus, equipment, medium, and program products provided by this invention construct a group of seismic monitoring stations including a target monitoring station and surrounding stations. Using historical monitoring data recorded by these stations, node features of each monitoring station and edge features between the target monitoring station and surrounding stations are extracted. Then, a historical station feature map is constructed using the node and edge features. This constructed station feature map is used to train a graph neural network with a structure suitable for station feature map analysis. Finally, the feature map of the station group corresponding to the station to be predicted in the region lacking cumulative absolute velocity is input into the trained graph neural network to obtain the cumulative absolute velocity prediction result for the station to be predicted. This application improves the accuracy of cumulative absolute velocity prediction by comprehensively analyzing historical seismic data and measured data from surrounding stations through the prediction of cumulative absolute velocity data of a region using a trained graph neural network. Attached Figure Description
[0038] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0039] Figure 1 This is one of the flowcharts illustrating the cumulative absolute velocity prediction method provided by this invention.
[0040] Figure 2 This is the second flowchart of the cumulative absolute velocity prediction method provided by the present invention.
[0041] Figure 3 This is a schematic diagram of the feature map of the station group provided by the present invention.
[0042] Figure 4 This is a schematic diagram of the cumulative absolute velocity prediction device provided by the present invention.
[0043] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation
[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0045] The following is combined Figures 1-5 This invention describes the cumulative absolute velocity prediction method, apparatus, device, medium, and program products.
[0046] Figure 1 This is one of the flowcharts illustrating the cumulative absolute velocity prediction method provided by this invention, such as... Figure 1 As shown, the method includes the following:
[0047] Step 100: Construct a group of earthquake monitoring stations; the group of earthquake monitoring stations includes the target monitoring station and surrounding stations;
[0048] Specifically, the main contents of the cumulative absolute velocity prediction method provided by the present invention include the following steps:
[0049] Step 1: Select a target monitoring station from among numerous historical earthquake monitoring stations. This station is designated as being in an area lacking CAV monitoring data, but it contains historical CAV data that can be used for training. Select multiple stations with measured data (i.e., surrounding stations in this embodiment) around the target monitoring station to form a station group (i.e., the earthquake monitoring station group in this embodiment). Then, based on the historical earthquake database, establish a sample database containing multiple groups of earthquake monitoring stations for training the graph neural network.
[0050] Step 200: Obtain the node features of each station in the earthquake monitoring station group, and obtain the edge features between the target monitoring station and the surrounding stations;
[0051] Specifically, the main contents of the cumulative absolute velocity prediction method provided by the present invention also include the following steps:
[0052] Step 2: Encode the physical characteristics of the constructed seismic monitoring station group to obtain the node features of each monitoring station and the edge features between the target monitoring station and surrounding stations. Construct a feature map of the seismic monitoring station group based on the node features and edge features to capture the interrelationships and characteristics between the seismic monitoring stations.
[0053] Step 300: Construct a historical station group feature map based on the node features and the edge features; the historical station group feature map is used to train a graph neural network;
[0054] Specifically, the main contents of the cumulative absolute velocity prediction method provided by the present invention also include the following steps:
[0055] Step 3: Using the feature maps of the seismic monitoring station group as input to the graph neural network and the CAV at the target monitoring station as output, a deep learning model based on a graph neural network (GNN) is constructed. Trained using the constructed feature maps of multiple historical station groups, this deep learning model can predict the CAV at the target location (i.e., the station to be predicted).
[0056] Step 400: Input the feature map of the station group corresponding to the station to be predicted into the trained graph neural network to obtain the cumulative absolute velocity prediction result of the station to be predicted.
[0057] Specifically, the main contents of the cumulative absolute velocity prediction method provided by the present invention also include the following steps:
[0058] Step 4: After constructing and training the model for predicting the CAV at the target location, input the feature map of the station group corresponding to the station to be predicted into the trained graph neural network to obtain the CAV at the station to be predicted output by the trained graph neural network.
[0059] This embodiment constructs a seismic monitoring station group including the target monitoring station and surrounding stations. Using historical monitoring data recorded by these stations, it extracts node features for each station and edge features between the target station and surrounding stations. Then, it constructs a historical station feature map using the node and edge features. This constructed feature map is used to train a graph neural network with a structure suitable for analyzing station feature maps. Finally, the feature map of the station group corresponding to the station to be predicted in the region lacking cumulative absolute velocity is input into the trained graph neural network to obtain the cumulative absolute velocity prediction results for the station to be predicted. This application improves the accuracy of cumulative absolute velocity prediction by comprehensively analyzing historical seismic data and measured data from surrounding stations through the prediction of cumulative absolute velocity data of the region using a trained graph neural network.
[0060] Figure 2 This is the second flowchart of the cumulative absolute velocity prediction method provided by the present invention, as shown below. Figure 2 As shown, the method may further include:
[0061] Step 110: Obtain the type and number of historical earthquake monitoring stations, as well as the distance between any two historical earthquake monitoring stations;
[0062] Step 120: Determine the target monitoring station and surrounding stations among the historical earthquake monitoring stations; the surrounding stations are stations around the target monitoring station; the target monitoring station and the surrounding stations are surface stations; the number of historical records of the target monitoring station and the surrounding stations is greater than a first threshold; the distance between the target monitoring station and the surrounding stations is less than a second threshold; the number of surrounding stations is greater than a third threshold;
[0063] Step 130: Based on the target monitoring station and the surrounding stations, construct an earthquake monitoring station group.
[0064] Specifically, when selecting the target monitoring station from historical earthquake monitoring stations, as well as the surrounding stations, the screening criteria mainly include the type of station, the historical data of the station, and the distance between stations. For example, the target monitoring station and its surrounding stations are surface stations; the historical data of the target monitoring station and its surrounding stations must be greater than a certain number (e.g., 20, i.e., the first threshold in this embodiment); and the number of surrounding stations less than 40 kilometers away from the selected target monitoring station (i.e., the second threshold in this embodiment) must be greater than 3 (i.e., the third threshold in this embodiment).
[0065] like Figure 3 As shown, taking an example where the number of surrounding stations of the target monitoring station is 4, if the number of surrounding stations less than the second threshold of the target monitoring station is greater than 4, then 4 surrounding stations of the target monitoring station can be selected in order from closest to furthest. Each constructed seismic monitoring station group includes one target monitoring station and multiple surrounding stations.
[0066] This embodiment constructs an earthquake monitoring station group using data from various historical earthquake monitoring stations, enabling the model trained based on the earthquake monitoring station group to comprehensively learn the relationships and characteristics between the stations.
[0067] In one embodiment, the cumulative absolute velocity prediction method provided by this invention may further include:
[0068] Step 210: Obtain the first node features of the target monitoring station; the first node features include the magnitude, epicentral distance, soil shear wave velocity, and focal depth in the historical records of the target monitoring station;
[0069] Step 220: Obtain the second node features of the surrounding stations; the second node features include soil shear wave velocity, epicentral distance, and cumulative absolute velocity in the historical records of the surrounding stations.
[0070] Specifically, the physical characteristics of the seismic monitoring station group are encoded, and the encoded characteristics are designed as graph features of the seismic monitoring station group. This includes: combining the node features, edge features, and graph topology of the seismic monitoring station group into graph features that can be input into a graph neural network; the node features are used to describe the attributes of the seismic monitoring stations, wherein the node features of the target monitoring station are composed of attributes such as magnitude, epicentral distance, soil shear wave velocity, and focal depth, and the node features of multiple surrounding stations are composed of soil shear wave velocity, epicentral distance, and cumulative absolute velocity.
[0071] This embodiment obtains the node features for constructing the feature map of the seismic monitoring station group by encoding the physical characteristics of the seismic monitoring station group.
[0072] In one embodiment, the cumulative absolute velocity prediction method provided by this invention may further include:
[0073] Step 230: Based on the differences in site conditions, determine the first edge characteristics between the target monitoring station and the surrounding stations; the differences in site conditions are determined based on the absolute difference in soil shear wave velocity between the target monitoring station and the surrounding stations.
[0074] Step 240: Based on spatial relationships, determine the second-side features between the target monitoring station and the surrounding stations; the spatial relationships are determined based on the absolute difference in epicentral distance between the target monitoring station and the surrounding stations.
[0075] Step 250: Based on the distance between the target monitoring station and the surrounding stations, determine the third side feature between the target monitoring station and the surrounding stations.
[0076] Specifically, edge features are mainly used to quantify the relationship between the target monitoring station and surrounding stations, and are constructed based on differences in multiple physical attributes. One type of physical attribute difference is the absolute difference in soil shear wave velocity, which represents the difference in site conditions between the target monitoring station and surrounding stations; the second type of physical attribute difference is the absolute difference in epicentral distance, which reflects the spatial relationship between stations; and the third type of physical attribute difference is the distance between surrounding stations and the target monitoring station.
[0077] This embodiment obtains the edge features of the station group feature map by quantifying the relationship between the target monitoring station and the surrounding stations.
[0078] In one embodiment, the cumulative absolute velocity prediction method provided by this invention may further include:
[0079] Step 310: Transform the differences in site conditions and the spatial relationships to obtain nonlinear characteristics;
[0080] Step 320: Based on the nonlinear characteristics and the third-side characteristics, determine the connection relationship between the target monitoring station and the surrounding stations;
[0081] Step 330: Based on the connection relationship and target topology, construct a historical station group feature map; in the target topology, the target monitoring station is connected to the surrounding stations, and no two surrounding stations are connected to each other.
[0082] Specifically, to enhance the modeling capability for nonlinear relationships, logarithmic and square root transformations are applied to the aforementioned differential features to generate extended nonlinear features, namely the nonlinear features and third-side features in this embodiment. For example... Figure 3 As shown, a star topology can be adopted for the topological design of the historical station group feature map, where the target monitoring station serves as the central node and multiple surrounding stations serve as peripheral nodes. The connection method in the map is bidirectional, meaning the central node is interconnected with each peripheral node. This star topology simplifies the station group feature map and reduces its structural complexity while considering the local relationships between seismic monitoring stations.
[0083] This embodiment reduces the structural complexity of the station group feature graph by constructing graph features and graph topology.
[0084] In one embodiment, the cumulative absolute velocity prediction method provided by this invention may further include:
[0085] Step 410: Input the feature map of the historical station group into the dynamic graph attention layer to obtain the relationships between nodes and the adjusted edge feature weights;
[0086] Step 420: Input the node relationships and adjusted edge feature weights into the global average pooling layer to obtain the global feature vector;
[0087] Step 430: Input the global feature vector into the fully connected layer to obtain the initial prediction result;
[0088] Step 440: Train the graph neural network based on the initial prediction results; the training process includes network architecture adjustment and activation function adjustment.
[0089] Specifically, the steps for constructing a deep learning network based on a graph neural network (i.e., the graph neural network in this embodiment) include:
[0090] The deep learning network architecture is designed based on GNN. The first and second layers of the network can adopt a graph attention network based on a multi-head attention mechanism to dynamically adjust the edge feature weights and capture complex relationships between nodes. The third layer uses graph convolution based on Transformer (encoder-decoder) to further improve the ability to model global dependencies. After multi-layer feature extraction, the network aggregates all node features into a fixed-length global feature vector through Global Mean Pooling (GMP). Finally, a two-layer fully connected network is used.
[0091] The network input is the station group feature map obtained above, and the network output is the CAV at the target monitoring station. By adjusting the network architecture, activation function and optimization method, a network architecture suitable for predicting the CAV at the target monitoring station is constructed.
[0092] Specifically, considering the characteristics of CAV prediction at target monitoring stations, a network architecture and optimization strategy were designed. Key parameters of the graph neural network were adjusted based on the distribution characteristics of CAV data from the monitoring stations. The network depth and activation function type of the graph neural network were optimized based on the complexity of the station group feature maps. A suitable combination of network hyperparameters for predicting CAV at the target monitoring stations was determined based on the prediction accuracy and physical plausibility. During the training process of the graph neural network, specific training criteria adapted to the CAV prediction requirements at the target monitoring stations were introduced to optimize the network prediction error. Finally, the validation performance of the trained model on the validation dataset was determined, showing that the model can accurately predict the CAV at the location of the target monitoring stations.
[0093] This embodiment constructs a network architecture based on GNN to train a model for predicting CAV at stations to be predicted, and makes predictions for stations lacking CAV data, thereby improving the accuracy of CAV prediction in areas lacking monitoring data.
[0094] The cumulative absolute velocity prediction device provided by the present invention is described below. The cumulative absolute velocity prediction device described below can be referred to in correspondence with the cumulative absolute velocity prediction method described above.
[0095] Please refer to Figure 4 The present invention also provides a cumulative absolute velocity prediction device, comprising:
[0096] Earthquake monitoring station group construction module 401 is used to construct an earthquake monitoring station group; the earthquake monitoring station group includes target monitoring stations and surrounding stations.
[0097] The feature acquisition module 402 is used to acquire the node features of each station in the earthquake monitoring station group and to acquire the edge features between the target monitoring station and the surrounding stations.
[0098] The historical station group feature map construction module 403 is used to construct a historical station group feature map based on the node features and the edge features; the historical station group feature map is used to train a graph neural network.
[0099] The cumulative absolute velocity prediction module 404 is used to input the feature map of the station group corresponding to the station to be predicted into the trained graph neural network to obtain the cumulative absolute velocity prediction result of the station to be predicted.
[0100] Optionally, the earthquake monitoring station group construction module includes:
[0101] The acquisition unit is used to acquire the type and number of historical earthquake monitoring stations, as well as the distance between any two historical earthquake monitoring stations.
[0102] A station determination unit is used to determine the target monitoring station and surrounding stations among historical earthquake monitoring stations; the surrounding stations are stations around the target monitoring station; the target monitoring station and the surrounding stations are surface stations; the number of historical records of the target monitoring station and the surrounding stations is greater than a first threshold; the distance between the target monitoring station and the surrounding stations is less than a second threshold; and the number of surrounding stations is greater than a third threshold.
[0103] The earthquake monitoring station group construction unit is used to construct an earthquake monitoring station group based on the target monitoring station and the surrounding stations.
[0104] Optionally, the feature acquisition module includes:
[0105] The first node feature acquisition unit is used to acquire the first node features of the target monitoring station; the first node features include the magnitude, epicentral distance, soil shear wave velocity and focal depth in the historical records of the target monitoring station;
[0106] The second node feature acquisition unit is used to acquire the second node features of the surrounding stations; the second node features include soil shear wave velocity, epicentral distance, and cumulative absolute velocity in the historical records of the surrounding stations.
[0107] Optionally, the feature acquisition module further includes:
[0108] The first side feature acquisition unit is used to determine the first side features between the target monitoring station and the surrounding stations based on the differences in site conditions; the differences in site conditions are determined based on the absolute difference in soil shear wave velocity between the target monitoring station and the surrounding stations.
[0109] The second-side feature determination unit is used to determine the second-side features between the target monitoring station and the surrounding stations based on spatial relationships; the spatial relationships are determined based on the absolute difference in epicentral distance between the target monitoring station and the surrounding stations.
[0110] The third-side feature determination unit is used to determine the third-side feature between the target monitoring station and the surrounding stations based on the distance between the target monitoring station and the surrounding stations.
[0111] Optionally, the historical station group feature map construction module includes:
[0112] A nonlinear feature determination unit is used to transform the site condition differences and the spatial relationship to obtain nonlinear features;
[0113] A connection relationship determination unit is used to determine the connection relationship between the target monitoring station and the surrounding stations based on the nonlinear feature and the third-side feature;
[0114] The historical station group feature map construction unit is used to construct a historical station group feature map based on the connection relationship and the target topology; in the target topology, the target monitoring station is connected to the surrounding stations, and no two surrounding stations are connected to each other.
[0115] Optionally, the cumulative absolute velocity prediction device further includes:
[0116] The edge feature weight adjustment module is used to input the historical station group feature map into the dynamic graph attention layer to obtain the relationship between nodes and the adjusted edge feature weights.
[0117] The global average pooling module is used to input the relationship between nodes and the adjusted edge feature weights into the global average pooling layer to obtain a global feature vector.
[0118] The initial prediction result determination module is used to input the global feature vector into the fully connected layer to obtain the initial prediction result;
[0119] The graph neural network training module is used to train the graph neural network based on the initial prediction results; the training process includes network architecture adjustment and activation function adjustment.
[0120] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communications interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a cumulative absolute velocity prediction method, which includes: constructing a seismic monitoring station group; the seismic monitoring station group includes a target monitoring station and surrounding stations; acquiring node features of each station in the seismic monitoring station group, and acquiring edge features between the target monitoring station and the surrounding stations; constructing a historical station group feature map based on the node features and the edge features; the historical station group feature map is used to train a graph neural network; and inputting the station group feature map corresponding to the station to be predicted into the trained graph neural network to obtain the cumulative absolute velocity prediction result of the station to be predicted.
[0121] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0122] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the cumulative absolute velocity prediction method provided by the above methods. The method includes: constructing a seismic monitoring station group; the seismic monitoring station group includes a target monitoring station and surrounding stations; obtaining the node features of each station in the seismic monitoring station group, and obtaining the edge features between the target monitoring station and the surrounding stations; constructing a historical station group feature map based on the node features and the edge features; using the historical station group feature map to train a graph neural network; and inputting the station group feature map corresponding to the station to be predicted into the trained graph neural network to obtain the cumulative absolute velocity prediction result of the station to be predicted.
[0123] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the cumulative absolute velocity prediction method provided by the methods described above. The method includes: constructing a seismic monitoring station group; the seismic monitoring station group including a target monitoring station and surrounding stations; acquiring node features of each station in the seismic monitoring station group, and acquiring edge features between the target monitoring station and the surrounding stations; constructing a historical station group feature map based on the node features and the edge features; the historical station group feature map being used to train a graph neural network; and inputting the station group feature map corresponding to the station to be predicted into the trained graph neural network to obtain the cumulative absolute velocity prediction result of the station to be predicted.
[0124] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0125] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting cumulative absolute velocity, characterized in that, include: Construct a group of earthquake monitoring stations; the group of earthquake monitoring stations includes target monitoring stations and surrounding stations. Obtain the node characteristics of each station in the earthquake monitoring station group, and determine the first edge characteristics between the target monitoring station and the surrounding stations based on the differences in site conditions; The site condition differences are determined based on the absolute difference in soil shear wave velocity between the target monitoring station and the surrounding stations; Based on spatial relationships, the second side features between the target monitoring station and the surrounding stations are determined; The spatial relationship is determined based on the absolute difference in epicentral distance between the target monitoring station and the surrounding stations; Based on the distance between the target monitoring station and the surrounding stations, the third side feature between the target monitoring station and the surrounding stations is determined; By transforming the differences in site conditions and the spatial relationships, nonlinear characteristics are obtained; Based on the nonlinear characteristics and the third-side characteristics, the connection relationship between the target monitoring station and the surrounding stations is determined; Based on the connection relationship and target topology, a historical station group feature map is constructed; in the target topology, the target monitoring station is connected to the surrounding stations, and no two surrounding stations are connected; the historical station group feature map is used to train a graph neural network. The feature map of the station group corresponding to the station to be predicted is input into the trained graph neural network to obtain the cumulative absolute velocity prediction result of the station to be predicted.
2. The cumulative absolute velocity prediction method according to claim 1, characterized in that, The earthquake monitoring station group includes: Obtain the type and number of historical earthquake monitoring stations, as well as the distance between any two historical earthquake monitoring stations; Identify the target monitoring station and surrounding stations among historical earthquake monitoring stations; the surrounding stations are stations around the target monitoring station; the target monitoring station and the surrounding stations are surface stations; the number of historical records of the target monitoring station and the surrounding stations is greater than a first threshold; the distance between the target monitoring station and the surrounding stations is less than a second threshold; the number of surrounding stations is greater than a third threshold. Based on the target monitoring station and the surrounding stations, an earthquake monitoring station group is constructed.
3. The cumulative absolute velocity prediction method according to claim 2, characterized in that, The acquisition of node characteristics of each station in the earthquake monitoring station group includes: Obtain the first node features of the target monitoring station; the first node features include the magnitude, epicentral distance, soil shear wave velocity, and focal depth in the historical records of the target monitoring station; Obtain the second node features of the surrounding stations; the second node features include soil shear wave velocity, epicentral distance, and cumulative absolute velocity from the historical records of the surrounding stations.
4. The cumulative absolute velocity prediction method according to claim 1, characterized in that, The graph neural network includes a dynamic graph attention layer, a global average pooling layer, and a fully connected layer; the cumulative absolute velocity prediction method further includes: The feature map of the historical station group is input into the dynamic graph attention layer to obtain the relationships between nodes and the adjusted edge feature weights. The relationships between nodes and the adjusted edge feature weights are input into the global average pooling layer to obtain the global feature vector; The global feature vector is input into the fully connected layer to obtain the initial prediction result; The graph neural network is trained based on the initial prediction results; the training process includes network architecture adjustment and activation function adjustment.
5. A cumulative absolute velocity prediction device, characterized in that, include: An earthquake monitoring station group construction module is used to construct an earthquake monitoring station group; the earthquake monitoring station group includes a target monitoring station and surrounding stations; The feature acquisition module is used to acquire the node features of each station in the earthquake monitoring station group and determine the first edge features between the target monitoring station and the surrounding stations based on the differences in site conditions. The site condition differences are determined based on the absolute difference in soil shear wave velocity between the target monitoring station and the surrounding stations; Based on spatial relationships, the second side features between the target monitoring station and the surrounding stations are determined; The spatial relationship is determined based on the absolute difference in epicentral distance between the target monitoring station and the surrounding stations; the third side feature between the target monitoring station and the surrounding stations is determined based on the distance between them. The historical station group feature map construction module is used to transform the differences in site conditions and the spatial relationships to obtain nonlinear features; Based on the nonlinear features and the third-side features, the connection relationship between the target monitoring station and the surrounding stations is determined; based on the connection relationship and the target topology, a historical station group feature map is constructed; in the target topology, the target monitoring station is connected to the surrounding stations, and no two surrounding stations are connected; the historical station group feature map is used to train a graph neural network. The cumulative absolute velocity prediction module is used to input the feature map of the station group corresponding to the station to be predicted into the trained graph neural network to obtain the cumulative absolute velocity prediction result of the station to be predicted.
6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the cumulative absolute velocity prediction method as described in any one of claims 1 to 4.
7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the cumulative absolute velocity prediction method as described in any one of claims 1 to 4.
8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the cumulative absolute velocity prediction method as described in any one of claims 1 to 4.