Natural earthquake hazard analysis method and system based on artificial neural network, and medium

Abstract

Disclosed in the present invention is a natural earthquake hazard analysis method based on an artificial neural network, the method comprising: S1, acquiring a target analysis region, and dividing the target analysis region into a plurality of sub-regions; S2, acquiring a historical earthquake dataset corresponding to the target analysis region and performing preprocessing; S3, performing sub-region labeling on the preprocessed dataset to obtain a plurality of focus sub-regions and non-focus sub-regions; S4, acquiring data subsets corresponding to the focus sub-regions and the non-focus sub-regions, and inputting the data subsets into an artificial neural network for feature extraction; S5, calculating the similarities between the non-focus sub-regions and the plurality of focus sub-regions; S6, using the historical earthquake dataset to train a convolutional neural network, so as to obtain an earthquake prediction model; and S7, inputting the data subsets of different regions into the earthquake prediction model for prediction, so as to obtain an earthquake prediction result, and outputting an earthquake hazard analysis report on the basis of the earthquake prediction result. The method realizes earthquake hazard analysis for multiple regions and potential earthquake regions.

Description

Methods, systems, and media for natural earthquake hazard analysis based on artificial neural networks Technical Field

[0001] This invention relates to the field of earthquake analysis technology, specifically to a method, system, equipment, and medium for analyzing the hazard of natural earthquakes based on artificial neural networks. Background Technology

[0002] In existing technologies, the analysis of seismic hazard generally adopts the traditional probabilistic seismic hazard analysis method. The traditional probabilistic seismic hazard analysis method usually analyzes a single area, analyzes the impact of all seismic sources on the area in turn, and then combines the impact of each seismic source on the area to obtain the hazard exceedance probability of the area.

[0003] The above analysis of seismic hazard is conducted for a single region, and the results of each analysis are for a single region. It is not possible to obtain hazard analysis parameters for multiple regions simultaneously. If a multi-regional systematic seismic hazard analysis is to be carried out, a large amount of data calculation and processing is required, which consumes a lot of time. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, device, and medium for natural earthquake hazard analysis based on artificial neural networks. By using artificial neural networks for efficient data processing, it is possible to perform hazard analysis on multiple regions simultaneously and to adopt different monitoring strategies for different region types, thereby achieving personalized analysis.

[0005] This invention is achieved through the following technical solution:

[0006] In a first aspect, embodiments of the present invention provide a method for natural earthquake hazard analysis based on artificial neural networks, comprising:

[0007] S1: Obtain the target analysis region and divide the target analysis region into multiple sub-regions;

[0008] S2: Obtain the historical earthquake dataset corresponding to the target analysis area and preprocess it to obtain the preprocessed historical earthquake dataset, which includes regional data and earthquake data;

[0009] S3: Sub-region labeling is performed on the preprocessed historical earthquake dataset to obtain multiple source sub-regions and multiple non-source sub-regions;

[0010] S4: Obtain the data subsets corresponding to the source sub-region and non-source sub-region, input the data subsets into the artificial neural network for feature extraction, and obtain the feature vectors of the source sub-region and non-source sub-region.

[0011] S5: Calculate the similarity between non-source sub-regions and multiple source sub-regions, mark non-source sub-regions with similarity greater than a preset threshold as potential source sub-regions, and mark non-source sub-regions with similarity less than or equal to the preset threshold as sourceless sub-regions;

[0012] S6: The convolutional neural network is trained using historical earthquake datasets to obtain an earthquake prediction model;

[0013] S7: Different monitoring strategies are adopted for the source sub-region, potential source sub-region and sourceless sub-region. According to the monitoring strategy, the corresponding regional data subsets are collected from the source sub-region, potential source sub-region and sourceless sub-region. The regional data subsets are input into the earthquake prediction model for prediction to obtain the earthquake prediction results. Based on the earthquake prediction results, an earthquake hazard analysis report is output.

[0014] Furthermore, the specific method for obtaining and preprocessing the historical earthquake dataset corresponding to the target analysis area includes:

[0015] The abnormal data points in the historical earthquake dataset are removed by a point-by-point elimination method to obtain the remaining data, and the missing data in the remaining data is filled in.

[0016] Furthermore, S21: Assume X = {x0, x1, x2, ..., x} n Let} be the interpolation sample set, Y = {y0, y1, y2, ..., y...} n} represents the corresponding X function value, n represents the number of nodes, and x represents the interpolation nodes;

[0017] S22: Input n+1 interpolation sample points x0, x1, x2, ..., xn n and the corresponding function values ​​y0, y1, y2, ..., y n ;

[0018] S23: Calculate the nth-order Lagrange basis function:

[0019] Where i = 0, 1, 2, ..., n;

[0020] S24: Calculate the nth-order Lagrange interpolation function:

[0021] Among them, l i (x) represents the nth-order Lagrange basis function;

[0022] S25: Input the interpolation point x and substitute it into the interpolation function in step S24 to obtain the calculation result.

[0023] Furthermore, the specific method for sub-region labeling based on the preprocessed historical earthquake dataset includes:

[0024] Determine whether the location information of the earthquake source in the historical earthquake dataset is located within a certain sub-region. If so, mark the sub-region as the earthquake source sub-region; otherwise, mark the sub-region as a non-earthquake source sub-region.

[0025] Furthermore, the step of obtaining data subsets corresponding to the seismic source sub-region and non-seismic source sub-regions, and inputting the data subsets into an artificial neural network for feature extraction, specifically includes:

[0026] Obtain source datasets corresponding to source sub-regions and non-source datasets corresponding to non-source sub-regions from historical earthquake datasets. The source datasets include earthquake data subsets and source region data subsets, and the non-source datasets include non-source region data subsets.

[0027] The earthquake data subset and the source region data subset are respectively input into the artificial neural network for feature extraction, and the earthquake feature vector and source region feature vector corresponding to each source sub-region are obtained.

[0028] The non-source region data subset is input into an artificial neural network for feature extraction, resulting in the non-source region feature vector corresponding to each non-source sub-region.

[0029] Furthermore, the step of training a convolutional neural network using historical earthquake datasets to obtain an earthquake prediction model specifically includes:

[0030] The source dataset and non-source dataset are used to form a training sample set according to a preset ratio. The regional data subset in the remaining source dataset is used as the test set, and the earthquake data subset is used as the validation set.

[0031] A convolutional neural network (CNN) is used as the model to be trained. The training sample set is input into the CNN for training. The validation set is used to observe whether the loss function of the CNN converges during the training process to determine whether to terminate the training. The test set is used to test the prediction accuracy of the CNN. The CNN is trained using the stochastic gradient descent method until the loss function of the neural network converges to the minimum value. The trained CNN is then used as the earthquake prediction model.

[0032] Furthermore, the monitoring strategy involves collecting regional data within sub-regions using different collection frequencies, with the collection frequency for the seismic source sub-region being greater than that for the potential seismic source sub-region, and the collection frequency for the potential seismic source sub-region being greater than that for the non-seismic source sub-region.

[0033] Secondly, another embodiment of the present invention provides a natural earthquake hazard analysis system based on artificial neural networks, used to implement the natural earthquake hazard analysis method based on artificial neural networks described in the above embodiments, comprising:

[0034] The region division module is used to obtain the target analysis region and divide the target analysis region into multiple sub-regions;

[0035] The data processing module is used to acquire and preprocess the historical earthquake dataset corresponding to the analysis area to obtain the preprocessed historical earthquake dataset, which includes regional data and earthquake data.

[0036] The region labeling module is used to label sub-regions of the preprocessed historical earthquake dataset, resulting in multiple source sub-regions and multiple non-source sub-regions. It is also used to calculate the similarity between non-source sub-regions and multiple source sub-regions, labeling non-source sub-regions with similarity greater than a preset threshold as potential source sub-regions, and labeling non-source sub-regions with similarity less than or equal to the preset threshold as sourceless sub-regions.

[0037] The feature extraction module is used to obtain the corresponding data subsets of the source sub-region and the non-source sub-region, input the data subsets into the artificial neural network for feature extraction, and obtain the feature vectors of the source sub-region and the non-source sub-region.

[0038] The model training module is used to train a convolutional neural network based on historical earthquake datasets to obtain an earthquake prediction model.

[0039] The regional monitoring module is used to adopt different monitoring strategies for the source sub-region, potential source sub-region, and sourceless sub-region. According to the monitoring strategy, it collects the corresponding regional data subsets from the source sub-region, potential source sub-region, and sourceless sub-region, inputs the regional data subsets into the earthquake prediction model for prediction, obtains the earthquake prediction results, and outputs an earthquake hazard analysis report based on the earthquake prediction results.

[0040] Thirdly, another embodiment of the present invention provides a natural earthquake hazard analysis device based on an artificial neural network. The device includes a memory and a processor, the memory and the processor being coupled together. The memory stores program instructions, which, when executed by the processor, cause the device to perform the method described in the above embodiments.

[0041] Fourthly, another embodiment of the present invention provides a computer-readable storage medium including a computer program that, when run on an electronic device, causes the electronic device to perform the methods described in the above embodiments.

[0042] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0043] This invention provides a natural earthquake hazard analysis method based on artificial neural networks. This method enables earthquake hazard analysis across multiple regions. By accurately extracting the characteristics of the epicenter region through artificial neural networks, it identifies potential earthquake zones within areas where earthquakes have not yet occurred, effectively reducing the earthquake hazard in these potential zones. Furthermore, it uses convolutional neural networks for earthquake prediction, generating earthquake hazard analysis reports that provide a scientific basis for managers to make earthquake prevention decisions, further reducing the harm caused by earthquakes. In addition, this invention employs different monitoring strategies to achieve personalized monitoring of different types of areas, improving the precision of earthquake hazard analysis.

[0044] The embodiments of the present invention are based on deep learning for model training, which reduces computational costs, enables efficient use of historical data resources, and improves the efficiency of natural earthquake hazard analysis.

[0045] The natural earthquake hazard analysis system, equipment and medium based on artificial neural networks provided in this embodiment of the invention have the same inventive concept and the same beneficial effects as the natural earthquake hazard analysis method based on artificial neural networks, and will not be described again here. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0047] Figure 1 is a flowchart of a natural earthquake hazard analysis method based on artificial neural networks provided in the first embodiment of the present invention;

[0048] Figure 2 is a structural block diagram of a natural earthquake hazard analysis system based on an artificial neural network provided in the second embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0050] Example 1

[0051] As shown in Figure 1, the first embodiment of the present invention provides a method for natural earthquake hazard analysis based on artificial neural networks, which includes the following steps:

[0052] S1: Obtain the target analysis region and divide the target analysis region into multiple sub-regions.

[0053] The region can be divided according to administrative regions. For example, if the preset analysis region is one or more provinces, each sub-region can correspond to a city or district. If the preset analysis region is one or more countries, each sub-region can correspond to a province or city. Alternatively, the region can be divided manually, with the preset analysis region divided into multiple sub-regions according to geographical location or geographical area. The area of ​​each sub-region can be the same or different. Alternatively, any other method of region division known in the art can be used, and this application does not specifically limit this method.

[0054] S2: Obtain the historical earthquake dataset corresponding to the target analysis area and preprocess it to obtain the preprocessed historical earthquake dataset, which includes regional data and earthquake data.

[0055] Historical earthquake datasets can be obtained through one or more of the following methods:

[0056] 1. Official earthquake monitoring agencies:

[0057] China Earthquake Administration: The China Earthquake Networks Center provides real-time earthquake data, which can be accessed through its official website.

[0058] The United States Geological Survey (USGS) provides access to global earthquake data.

[0059] 2. Data sharing platform:

[0060] The National Earth System Science Data Sharing Service Platform provides a variety of geoscience data, including earthquake data.

[0061] The National Data Sharing and Exchange Platform aggregates data resources from various government departments, including earthquake data.

[0062] 3. Academic research institutions:

[0063] Some university and research institution earthquake laboratories publish earthquake datasets they collect or study.

[0064] 4. International Data Organization:

[0065] IRIS (Incorporated Research Institutions for Seismology): Provides global earthquake data.

[0066] EMSC (European-Mediterranean Seismological Centre): Provides earthquake data for Europe and the Mediterranean region.

[0067] 5. Open source data projects:

[0068] There are also some open-source projects on GitHub that collect and organize earthquake data, which can be downloaded and used for free.

[0069] Historical earthquake datasets can include regional data related to regional information and earthquake data related to earthquake information. Earthquake data related to earthquake information can include earthquake occurrence time, focal longitude, focal latitude, focal depth, and magnitude type, while regional data related to regional information can include the region's geological structure, crustal type, geomorphological features, climate change, and underground fluid activity.

[0070] Geological structure, crustal type, and geomorphological features are all important factors influencing earthquake occurrence; they can act individually or in combination to cause earthquakes. The following is a detailed explanation of how these factors affect earthquake occurrence:

[0071] 1. Geological structure:

[0072] Plate tectonics: The most common cause of earthquakes is the interaction between tectonic plates. The compression, stretching, or friction between plates accumulates stress in the Earth's crust. When the stress exceeds the strength of the rock, a fracture occurs, releasing energy and causing an earthquake.

[0073] Fault activity: A fault is a fracture zone in the Earth's crust. Rocks near an active fault will displace under continuous stress, thus causing earthquakes.

[0074] 2. Type of Earth's crust:

[0075] The thickness and composition of the Earth's crust: The thickness and composition of the Earth's crust affect its strength and brittleness. A thinner or more brittle crust is more prone to fracturing, thus triggering earthquakes.

[0076] Rock type: Different types of rocks respond differently to stress. For example, some rocks may undergo plastic deformation when subjected to stress without immediately fracturing, while others may be more brittle and prone to fracture.

[0077] 3. Landform features:

[0078] Mountain formation: The formation of mountain ranges is often related to the collision of tectonic plates, and this geological activity can lead to earthquakes.

[0079] Geological age: Some landform features, such as the rate of uplift of mountains, can indicate the intensity and frequency of crustal activity, thus affecting the occurrence of earthquakes.

[0080] Landform: Landform can reflect the underground geological structure, such as the surface traces of faults, and these areas are usually high-risk areas for earthquakes.

[0081] In general, geological structure, crustal type, and geomorphological features are all potential factors in earthquake occurrence. They determine the frequency and intensity of earthquakes by influencing the accumulation and release of crustal stress. However, earthquake occurrence is a complex natural process, usually the result of the combined effects of multiple factors. Therefore, earthquake prediction and risk assessment need to comprehensively consider these factors, as well as other possible factors such as underground fluid activity and climate change.

[0082] Preprocessing historical earthquake datasets typically involves correcting or removing anomalies and errors in the acquired parameters from the seismic data system. This is because historical earthquake data often covers parameters from several years, decades, or even centuries, resulting in a massive amount of data. Therefore, a point-by-point elimination method is needed to identify obviously anomalous data points, then determine the normal data, and finally perform regression analysis on all possible subsets of the remaining data within the range of anomalous data. Furthermore, to address potential missing or anomaly data in the dataset, it is also necessary to impute any missing data.

[0083] The Lagrange interpolation algorithm can be used to complete missing data. The specific steps are as follows:

[0084] S21: Assume X = {x0, x1, x2, ..., x} n Let} be the interpolation sample set, Y = {y0, y1, y2, ..., y...} n} represents the corresponding X function value, n represents the number of nodes, and x represents the interpolation nodes;

[0085] S22: Input n+1 interpolation sample points x0, x1, x2, ..., xn n and the corresponding function values ​​y0, y1, y2, ..., y n ;

[0086] S23: The formula for calculating the nth-order Lagrange basis function is:

[0087] Where i = 0, 1, 2, ..., n;

[0088] S24: Calculate the nth-order Lagrange interpolation function:

[0089] Among them, l i (x) represents the nth-order Lagrange basis function.

[0090] S25: Input the interpolation node x and substitute it into the interpolation function in step S24 to obtain the calculation result.

[0091] S3: Sub-region labeling is performed on the preprocessed historical earthquake dataset to obtain multiple source sub-regions and multiple non-source sub-regions.

[0092] Based on the preprocessed historical earthquake dataset, multiple sub-regions are matched. If the location information of a certain earthquake source in the historical earthquake dataset indicates that the source is located in a certain sub-region, then the sub-region is marked as a source sub-region, thus obtaining one or more source sub-regions. The remaining sub-regions are marked as non-source sub-regions.

[0093] S4: Obtain the data subsets corresponding to the source sub-region and the non-source sub-region, input the data subsets into the artificial neural network for feature extraction, and obtain the feature vectors of the source sub-region and the non-source sub-region.

[0094] Obtain the source dataset corresponding to the source sub-region and the non-source dataset corresponding to the non-source sub-region from the historical earthquake dataset. For the source sub-region, since an earthquake has occurred in this sub-region, the corresponding data naturally includes regional data related to regional information and earthquake data related to earthquake information. Therefore, the source dataset is set to include a subset of earthquake data and a subset of source regional data. For the non-source sub-region, since no earthquake has occurred in this sub-region, the corresponding data naturally only includes regional data related to regional information. Therefore, the non-source dataset is set to include only a subset of non-source regional data.

[0095] The source region data subsets and earthquake data subsets corresponding to the source sub-regions are respectively input into an artificial neural network for feature extraction, resulting in the source region feature vector Q corresponding to each source sub-region. m (x1,x2,…,x i ) and earthquake eigenvector P m (y1,y2,…,y j The non-source region data subsets corresponding to the non-source sub-regions are input into an artificial neural network for feature extraction, resulting in the non-source region feature vector Q corresponding to each non-source sub-region. n (x1,x2,…,x i ); where 1≤m≤M and 1≤n≤N, M is the number of source sub-regions, N is the number of non-source sub-regions, M and N are both positive integers greater than or equal to 1, i is the number of regional features, j is the number of seismic features, i and j are both positive integers greater than or equal to 1.

[0096] S5: Calculate the similarity between non-source sub-regions and multiple source sub-regions, mark non-source sub-regions with similarity greater than a preset threshold as potential source sub-regions, and mark non-source sub-regions with similarity less than or equal to the preset threshold as sourceless sub-regions.

[0097] The similarity between sub-regions can be obtained by calculating the similarity between feature vectors. First, calculate the non-source region feature vector Q of any non-source sub-region. nThe source region eigenvector Q of any source sub-region m Similarity between Summing these values ​​again yields the non-source region feature vector Q for any non-source sub-region. n Similarity between source region feature vectors and multiple source sub-regions sim(Q n This can be used as the similarity between a non-source sub-region and multiple source sub-regions.

[0098] If sim(Q n If sim(Q) is greater than a preset threshold, it indicates that the non-source region characteristics of this non-source sub-region are quite similar to the source region characteristics of one or more sub-regions within the source sub-region. This non-source sub-region also has the potential to experience an earthquake, therefore it is marked as a potential source sub-region. n If the value is less than or equal to a preset threshold, it indicates that the non-source region characteristics of this non-source sub-region are not similar to the source region characteristics of each sub-region in the source sub-region. The probability of an earthquake occurring in this non-source sub-region is extremely low, so it is marked as a sourceless sub-region.

[0099] S6: The convolutional neural network is trained using historical earthquake datasets to obtain an earthquake prediction model.

[0100] A portion of the source dataset and non-source dataset are used to form a training sample set according to a preset ratio. A subset of regional data from the remaining source dataset is used as a test set, and a subset of earthquake data is used as a validation set.

[0101] A convolutional neural network (CNN) is used as the model to be trained. The training sample set is input into the CNN for training. The validation set is used to observe whether the loss function of the CNN converges during training to determine whether to terminate training. The test set is used to test the prediction accuracy of the CNN. The stochastic gradient descent method is used to train the CNN until the loss function converges to its minimum value. The trained CNN is then used as the earthquake prediction model.

[0102] S7: Different monitoring strategies are adopted for the source sub-region, potential source sub-region and sourceless sub-region. According to the monitoring strategy, the corresponding regional data subsets are collected from the source sub-region, potential source sub-region and sourceless sub-region. The regional data subsets are input into the earthquake prediction model for prediction to obtain the earthquake prediction results. Based on the earthquake prediction results, an earthquake hazard analysis report is output.

[0103] Different monitoring strategies can refer to collecting regional data within a sub-region based on different acquisition frequencies. For example, for a seismic source sub-region, regional data is collected at a preset acquisition frequency D1; for a potential seismic source sub-region, regional data is collected at a preset acquisition frequency D2; and for a seismic source-free sub-region, regional data is collected at a preset acquisition frequency D3, where D1>D2>D3>0. By inputting the collected regional data into the earthquake prediction model to obtain earthquake prediction results, anomalies in the regional data of the seismic source sub-region or potential seismic source sub-region can be detected in a timely manner, thereby determining whether there is a possibility of an earthquake occurring. By outputting an earthquake hazard analysis report, the management personnel in that sub-region can be alerted to take relevant preventive measures to reduce losses caused by an earthquake.

[0104] In addition, existing monitoring strategies in the current technology can be combined according to actual needs to form multiple monitoring strategies with different priorities, so as to realize personalized monitoring of different types of areas and improve the precision of seismic hazard analysis.

[0105] This invention provides a natural earthquake hazard analysis method based on artificial neural networks, which enables earthquake hazard analysis of multiple regions. By accurately extracting the characteristics of the epicenter region through artificial neural networks, potential earthquake zones in areas where no earthquakes have occurred can be identified, effectively reducing the earthquake hazard of potential earthquake zones. In addition, earthquake prediction is performed through convolutional neural networks to generate earthquake hazard analysis reports, providing a scientific basis for managers to make earthquake prevention decisions and further reducing the harm caused by earthquakes.

[0106] Example 2

[0107] As shown in Figure 2, the second embodiment of the present invention provides a natural earthquake hazard analysis system based on artificial neural networks, used to implement the natural earthquake hazard analysis method based on artificial neural networks described in the above embodiments. The system includes:

[0108] The region division module is used to obtain the target analysis region and divide the target analysis region into multiple sub-regions;

[0109] The data processing module is used to acquire and preprocess the historical earthquake dataset corresponding to the analysis area to obtain the preprocessed historical earthquake dataset, which includes regional data and earthquake data.

[0110] The region labeling module is used to label sub-regions of the preprocessed historical earthquake dataset, resulting in multiple source sub-regions and multiple non-source sub-regions. It is also used to calculate the similarity between non-source sub-regions and multiple source sub-regions, labeling non-source sub-regions with similarity greater than a preset threshold as potential source sub-regions, and labeling non-source sub-regions with similarity less than or equal to the preset threshold as sourceless sub-regions.

[0111] The feature extraction module is used to obtain the corresponding data subsets of the source sub-region and the non-source sub-region, input the data subsets into the artificial neural network for feature extraction, and obtain the feature vectors of the source sub-region and the non-source sub-region.

[0112] The model training module is used to train a convolutional neural network based on historical earthquake datasets to obtain an earthquake prediction model.

[0113] The regional monitoring module is used to adopt different monitoring strategies for the source sub-region, potential source sub-region, and sourceless sub-region. According to the monitoring strategy, it collects the corresponding regional data subsets from the source sub-region, potential source sub-region, and sourceless sub-region, inputs the regional data subsets into the earthquake prediction model for prediction, obtains the earthquake prediction results, and outputs an earthquake hazard analysis report based on the earthquake prediction results.

[0114] The execution process of each module can be carried out according to the steps of the natural earthquake hazard analysis method based on artificial neural network in Embodiment 1, and will not be described in detail in this embodiment.

[0115] Example 3

[0116] The third embodiment of the present invention also provides a natural earthquake hazard analysis device based on artificial neural networks, including a memory and a processor, wherein the memory and the processor are coupled; the memory stores program instructions, and when the program instructions are executed by the processor, the device performs the above-described natural earthquake hazard analysis method based on artificial neural networks.

[0117] Example 4

[0118] The fourth embodiment of the present invention also provides a computer-readable storage medium including a computer program that, when the computer program is run on an electronic device, causes the electronic device to perform the above-described method for natural earthquake hazard analysis based on artificial neural networks.

[0119] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0120] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more flowchart illustrations and / or one or more block diagrams.

[0121] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0122] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0123] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A natural seismic hazard analysis method based on artificial neural networks, characterized by, include: S1: Obtain the target analysis region and divide the target analysis region into multiple sub-regions; S2: Obtain the historical earthquake dataset corresponding to the target analysis area and preprocess it to obtain the preprocessed historical earthquake dataset, which includes regional data and earthquake data; S3: Sub-region labeling is performed on the preprocessed historical earthquake dataset to obtain multiple source sub-regions and multiple non-source sub-regions; S4: Obtain the data subsets corresponding to the source sub-region and non-source sub-region, input the data subsets into the artificial neural network for feature extraction, and obtain the feature vectors of the source sub-region and non-source sub-region. S5: Calculate the similarity between non-source sub-regions and multiple source sub-regions, mark non-source sub-regions with similarity greater than a preset threshold as potential source sub-regions, and mark non-source sub-regions with similarity less than or equal to the preset threshold as sourceless sub-regions; S6: The convolutional neural network is trained using historical earthquake datasets to obtain an earthquake prediction model; S7: Different monitoring strategies are adopted for the source sub-region, potential source sub-region and sourceless sub-region. According to the monitoring strategy, the corresponding regional data subsets are collected from the source sub-region, potential source sub-region and sourceless sub-region. The regional data subsets are input into the earthquake prediction model for prediction to obtain the earthquake prediction results. Based on the earthquake prediction results, an earthquake hazard analysis report is output.

2. The natural seismic hazard analysis method based on artificial neural network according to claim 1, characterized in that, The specific methods for obtaining and preprocessing the historical earthquake dataset corresponding to the target analysis area include: The abnormal data points in the historical earthquake dataset are removed by a point-by-point elimination method to obtain the remaining data, and the missing data in the remaining data is filled in.

3. The natural seismic hazard analysis method based on artificial neural network according to claim 2, characterized in that, The process of filling in missing data in the remaining data uses the Lagrange interpolation algorithm, specifically: S21: Assuming X = {x0, x1, x2,..., xn} is the interpolation sample set, Y = {y0, y1, y2,..., yn} is the corresponding X function value, n represents the number of nodes, and x represents the interpolation node. n n ​​ S22: input n+1 interpolation sample points x0, x1, x2,..., xn and corresponding function values y0, y1, y2,..., yn n ; and n ; S23: Calculate the n-th Lagrange base function: Where i = 0, 1, 2, ..., n; S24: Calculate the Lagrange interpolation function of order n: wherein, l i (x) represents an nth order Lagrange basis function; S25: Input the interpolation point x and substitute it into the interpolation function in step S24 to obtain the calculation result.

4. The natural seismic hazard analysis method based on artificial neural network according to claim 1, characterized in that, The specific method for sub-region labeling of the preprocessed historical earthquake dataset includes: Determine whether the location information of the earthquake source in the historical earthquake dataset is located within a certain sub-region. If so, mark the sub-region as the earthquake source sub-region; otherwise, mark the sub-region as a non-earthquake source sub-region.

5. The natural seismic hazard analysis method based on artificial neural network according to claim 1, characterized in that, The process of obtaining data subsets corresponding to the seismic source sub-region and non-seismic source sub-regions, and inputting these data subsets into an artificial neural network for feature extraction, specifically includes: Obtain source datasets corresponding to source sub-regions and non-source datasets corresponding to non-source sub-regions from historical earthquake datasets. The source datasets include earthquake data subsets and source region data subsets, and the non-source datasets include non-source region data subsets. The earthquake data subset and the source region data subset are respectively input into the artificial neural network for feature extraction, and the earthquake feature vector and source region feature vector corresponding to each source sub-region are obtained. The non-source region data subset is input into an artificial neural network for feature extraction, resulting in the non-source region feature vector corresponding to each non-source sub-region.

6. The natural seismic hazard analysis method based on artificial neural network according to claim 5, wherein, The method of training a convolutional neural network using historical earthquake datasets to obtain an earthquake prediction model specifically includes: The source dataset and non-source dataset are used to form a training sample set according to a preset ratio. The regional data subset in the remaining source dataset is used as the test set, and the earthquake data subset is used as the validation set. A convolutional neural network (CNN) is used as the model to be trained. The training sample set is input into the CNN for training. The validation set is used to observe whether the loss function of the CNN converges during the training process to determine whether to terminate the training. The test set is used to test the prediction accuracy of the CNN. The CNN is trained using the stochastic gradient descent method until the loss function of the neural network converges to the minimum value. The trained CNN is then used as the earthquake prediction model.

7. The natural seismic hazard analysis method based on artificial neural network according to claim 1, characterized in that, The monitoring strategy involves collecting regional data within sub-regions at different collection frequencies. The collection frequency for the seismic source sub-region is higher than that for the potential seismic source sub-region, and the collection frequency for the potential seismic source sub-region is higher than that for the sub-region without a seismic source.

8. A natural earthquake hazard analysis system based on artificial neural networks, used to implement the natural earthquake hazard analysis method based on artificial neural networks as described in claim 1, characterized in that, include: The region division module is used to obtain the target analysis region and divide the target analysis region into multiple sub-regions; The data processing module is used to acquire and preprocess the historical earthquake dataset corresponding to the analysis area to obtain the preprocessed historical earthquake dataset, which includes regional data and earthquake data. The region labeling module is used to label sub-regions of the preprocessed historical earthquake dataset, resulting in multiple source sub-regions and multiple non-source sub-regions. It is also used to calculate the similarity between non-source sub-regions and multiple source sub-regions, labeling non-source sub-regions with similarity greater than a preset threshold as potential source sub-regions, and labeling non-source sub-regions with similarity less than or equal to the preset threshold as sourceless sub-regions. The feature extraction module is used to obtain the corresponding data subsets of the source sub-region and the non-source sub-region, input the data subsets into the artificial neural network for feature extraction, and obtain the feature vectors of the source sub-region and the non-source sub-region. The model training module is used to train a convolutional neural network based on historical earthquake datasets to obtain an earthquake prediction model. The regional monitoring module is used to adopt different monitoring strategies for the source sub-region, potential source sub-region, and sourceless sub-region. According to the monitoring strategy, it collects the corresponding regional data subsets from the source sub-region, potential source sub-region, and sourceless sub-region, inputs the regional data subsets into the earthquake prediction model for prediction, obtains the earthquake prediction results, and outputs an earthquake hazard analysis report based on the earthquake prediction results.

9. A natural earthquake hazard analysis device based on artificial neural networks, characterized in that, The device includes: a memory and a processor, the memory and the processor being coupled; the memory stores program instructions that, when executed by the processor, cause the device to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Includes a computer program that, when run on an electronic device, causes the electronic device to perform the method as described in any one of claims 1 to 7.

Images