Seismic hazard tracking method and system based on source parameters and fault structure
Patent Information
- Application Number
- CN202311194089.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-15
- Publication Date
- 2026-09-18
- Estimated Expiration
- 2043-09-15
AI Technical Summary
上述对于地震危险性的分析是针对单一场点进行的，每次分析的结果都是单一场点的结果，无法同时得到多个场点危险性分析参数，若想要实现多场点系统的地震危险性分析，需要面临大量的数据计算处理过程，所消耗的时间较长
[0045] This invention provides a method for tracking seismic hazard based on source parameters and fault structures. The method includes: acquiring source parameters under different fault structures, including fractured fault zones, banded fault zones, large strike-slip fault zones, and low-stress environments; determining the hazard coefficients corresponding to different seismically active fault segments based on the source parameters under the different fault structures; establishing a deep neural network model; using the source parameters under the different fault structures as input and the hazard coefficients corresponding to the seismically active fault segments as output, inputting them into the deep neural network model for training and optimization to obtain an optimized seismic hazard prediction model; acquiring source parameters for a designated area; and tracking the seismic hazard of the designated area based on the optimized seismic hazard prediction model and the source parameters of the designated area. This method can shorten data analysis time and improve the accuracy of seismic hazard tracking.
Smart Images
Figure CN117233828B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of earthquake prediction, and in particular to a method and system for tracking earthquake hazard based on source parameters and fault structures. Background Technology
[0002] In related technologies, seismic hazard analysis generally employs traditional probabilistic seismic hazard analysis methods. These methods typically analyze a single seismic site, sequentially analyzing the impact of all seismic sources on that site, and then synthesizing the combined effects of all sources to obtain the hazard exceedance probability for that site. However, this seismic hazard analysis is performed on a single site, and each analysis yields results for only that site. It cannot simultaneously obtain hazard analysis parameters for multiple sites. To achieve seismic hazard analysis for a multi-site system, extensive data computation and processing are required, resulting in a significant time consumption. Summary of the Invention
[0003] The purpose of this invention is to provide a method and system for tracking seismic hazard based on source parameters and fault structures, which can shorten data analysis time and improve the accuracy of seismic hazard tracking.
[0004] To achieve the above objectives, the present invention provides the following solution:
[0005] A seismic hazard tracking method based on source parameters and fault structure includes:
[0006] Obtain source parameters under different fracture structures, including fractured fracture zones, strip-shaped fracture zones, large strike-slip fracture zones, and low-stress environments;
[0007] Based on the source parameters under different fault structures, the hazard coefficients corresponding to different active fault segments are determined;
[0008] Establish a deep neural network model;
[0009] The source parameters under different fault structures are used as inputs, and the hazard coefficients corresponding to the active fault segments are used as outputs. These are then input into the deep neural network model for training and optimization to obtain an optimized earthquake hazard prediction model.
[0010] Obtain the seismic source parameters for a specified region;
[0011] The seismic hazard of the designated area is tracked based on the optimized seismic hazard prediction model and the source parameters of the designated area.
[0012] Optionally, determining the hazard factor corresponding to different seismically active fault segments based on the source parameters under different fault structures specifically includes:
[0013] The source parameters under different fault structures are further subdivided to generate subdivision results;
[0014] Based on the subdivision results, different weights are set for multiple active seismic fault segments to generate weight setting results;
[0015] Based on the subdivision results and the weight setting results, the weights corresponding to each active seismic fault segment are added together to obtain the risk coefficients corresponding to multiple active seismic fault segments.
[0016] Optionally, the step of inputting the source parameters into the deep neural network model for training and optimization to obtain the optimized seismic hazard prediction model further includes:
[0017] The deep neural network model is optimized using the particle swarm optimization algorithm to obtain the optimized deep neural network model.
[0018] Optionally, the step of optimizing the deep neural network model using a particle swarm optimization algorithm to obtain an optimized deep neural network model specifically includes:
[0019] S1: Randomly initialize the particle swarm;
[0020] S2: Set the fitness of each particle in the particle swarm;
[0021] S3: Select the particle's current position as the initial individual extreme value p i Find the particle with the lowest fitness from the population as the initial global extremum p. g ;
[0022] S4: Compare the current fitness with the initial individual extreme value p. i The fitness of the initial individual, if the extreme value p i If the fitness is better, then maintain it; if the current fitness is better, then update the initial individual extreme value p. i ;
[0023] S5: Compare the initial individual extreme values p for each particle. i The fitness and the initial global extremum p g The fitness, if the initial global extremum p g If the fitness is better, then the initial global extremum p is maintained. g If the initial individual extreme value p i If the fitness is better, then update the initial global extremum p. g ;
[0024] S6: Update the velocity and position of each particle;
[0025] S7: Repeat steps S4 to S6 until an acceptable satisfactory solution is found or the maximum number of iterations is reached;
[0026] S8: Set the initial global extremum p g The corresponding particles are used as parameters for the deep neural network to build an optimized deep neural network model.
[0027] Optionally, the step of inputting the source parameters into the deep neural network model for training and optimization to obtain an optimized seismic hazard prediction model specifically includes:
[0028] The source parameters are divided into a source parameter training set and a source parameter test set;
[0029] The optimized deep neural network model is used to train the source parameter training set to obtain the trained earthquake hazard prediction model.
[0030] The trained earthquake hazard prediction model is optimized based on the source parameter test set to obtain the optimized earthquake hazard prediction model.
[0031] A seismic hazard tracking system based on source parameters and fault structure includes:
[0032] The source parameter acquisition module is used to acquire source parameters under different fault structures, including fractured fault zones, strip-shaped fault zones, large strike-slip fault zones, and low-stress environments.
[0033] The hazard factor determination module is used to determine the hazard factor corresponding to different seismically active fault segments based on the source parameters under the different fault structures.
[0034] The deep neural network model building module is used to build deep neural network models;
[0035] The optimized earthquake hazard prediction model determination module is used to take the source parameters under different fault structures as input and the hazard coefficient corresponding to the active fault segment as output, and input them into the deep neural network model for training and optimization to obtain the optimized earthquake hazard prediction model.
[0036] The module for acquiring seismic source parameters for a specified region is used to acquire seismic source parameters for that region.
[0037] The earthquake hazard tracking module is used to track the earthquake hazard of a set area based on the optimized earthquake hazard prediction model and the source parameters of the set area.
[0038] Optionally, the risk factor determination module specifically includes:
[0039] The subdivision result generation unit is used to subdivide the source parameters under different fault structures and generate subdivision results.
[0040] The weight setting result generation unit is used to set different weights for multiple seismic active fault segments according to the subdivision result and generate weight setting results;
[0041] The hazard coefficient determination unit is used to add up the weights corresponding to each of the active seismic fault segments according to the subdivision results and the weight setting results, and to obtain the hazard coefficients corresponding to multiple active seismic fault segments.
[0042] Optionally, it also includes:
[0043] The deep neural network model optimization module is used to optimize the deep neural network model using the particle swarm optimization algorithm to obtain an optimized deep neural network model.
[0044] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:
[0045] This invention provides a method for tracking seismic hazard based on source parameters and fault structures. The method includes: acquiring source parameters under different fault structures, including fractured fault zones, banded fault zones, large strike-slip fault zones, and low-stress environments; determining the hazard coefficients corresponding to different seismically active fault segments based on the source parameters under the different fault structures; establishing a deep neural network model; using the source parameters under the different fault structures as input and the hazard coefficients corresponding to the seismically active fault segments as output, inputting them into the deep neural network model for training and optimization to obtain an optimized seismic hazard prediction model; acquiring source parameters for a designated area; and tracking the seismic hazard of the designated area based on the optimized seismic hazard prediction model and the source parameters of the designated area. This method can shorten data analysis time and improve the accuracy of seismic hazard tracking. Attached Figure Description
[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0047] Figure 1 This is a flowchart of the seismic hazard tracking method based on source parameters and fault structures according to the present invention;
[0048] Figure 2This is a structural diagram of the seismic hazard tracking system based on source parameters and fault structures according to the present invention. Detailed Implementation
[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0050] The purpose of this invention is to provide a method and system for tracking seismic hazard based on source parameters and fault structures, which can shorten data analysis time and improve the accuracy of seismic hazard tracking.
[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0052] Example 1:
[0053] Figure 1 This is a flowchart of the seismic hazard tracking method based on source parameters and fault structure according to the present invention. Figure 1 As shown, a seismic hazard tracking method based on source parameters and fault structures includes:
[0054] Step 101: Obtain source parameters under different fault structures, including fractured fault zones, banded fault zones, large strike-slip fault zones, and low-stress environments. The source parameters include scalar seismic moment, source scale, and stress drop. This invention primarily uses the differences in stress drop across different fault structures for hazard tracking analysis.
[0055] This invention avoids the problem of low accuracy caused by simply analyzing data from a single field point by acquiring and analyzing source parameters under different fault structures. This invention analyzes and processes source parameters under different fault structures to obtain multiple field point hazard analysis parameters.
[0056] After screening the source parameters under different fault structures, invalid data and duplicate samples were eliminated, and 100 sets of typical case data were finally selected as the research samples for seismic hazard tracking.
[0057] Step 102: Determine the hazard coefficient corresponding to different seismically active fault segments based on the source parameters under the different fault structures.
[0058] Step 102 specifically includes:
[0059] The source parameters under different fault structures are further subdivided to generate subdivision results;
[0060] Based on the subdivision results, different weights are set for multiple active seismic fault segments to generate weight setting results;
[0061] Based on the subdivision results and the weight setting results, the weights corresponding to each active seismic fault segment are added together to obtain the risk coefficients corresponding to multiple active seismic fault segments.
[0062] Step 103: Establish a deep neural network model;
[0063] Step 104: The source parameters under different fault structures are used as inputs, and the hazard coefficients corresponding to the active fault segments are used as outputs. These are then input into the deep neural network model for training and optimization to obtain the optimized earthquake hazard prediction model.
[0064] This step specifically includes:
[0065] The source parameters are divided into a source parameter training set and a source parameter test set;
[0066] The optimized deep neural network model is used to train the source parameter training set to obtain the trained earthquake hazard prediction model.
[0067] The trained earthquake hazard prediction model is optimized based on the source parameter test set to obtain the optimized earthquake hazard prediction model.
[0068] The 100 collected samples were cyclically divided in a 98:2 ratio, with each sample serving as a separate validation set and the remaining 98 samples as the training set. The training set was used for parameter updates and model fitting, while the validation set was used to select the parameters corresponding to the best-performing model for parameter adjustment. The minimum fitness function, the globally optimal weights ω and variance σ, the validation set sample classification accuracy, and the sample mean squared error were recorded after each training iteration. After 100 training iterations, the average sample mean squared error and the sample classification accuracy of the entire validation set were calculated. The model with the closest average sample mean squared error and sample classification accuracy was selected as the seismic hazard assessment model. The final average mean squared error (average minimum fitness function) of the training set was 0.031, and the average mean squared error of the validation set was 0.075, with an accuracy of 97.23%. The optimal ω and σ of the model were selected based on these averages. This method of model training effectively avoids overfitting and yields the best prediction model.
[0069] Step 105: Obtain the seismic source parameters for the designated area;
[0070] Step 106: Track the seismic hazard of the designated area based on the optimized seismic hazard prediction model and the source parameters of the designated area.
[0071] The steps following step 103 and before step 104 also include:
[0072] The deep neural network model is optimized using the particle swarm optimization algorithm to obtain the optimized deep neural network model.
[0073] This step specifically includes:
[0074] S1: Randomly initialize the particle swarm;
[0075] S2: Set the fitness of each particle in the particle swarm;
[0076] S3: Select the particle's current position as the initial individual extreme value p i Find the particle with the lowest fitness from the population as the initial global extremum p. g ;
[0077] S4: Compare the current fitness with the initial individual extreme value p. i The fitness of the initial individual, if the extreme value p i If the fitness is better, then maintain it; if the current fitness is better, then update the initial individual extreme value p. i ;
[0078] S5: Compare the initial individual extreme values p for each particle. i The fitness and the initial global extremum p g The fitness, if the initial global extremum p g If the fitness is better, then the initial global extremum p is maintained. g If the initial individual extreme value p i If the fitness is better, then update the initial global extremum p. g ;
[0079] S6: Update the velocity and position of each particle;
[0080] S7: Repeat steps S4 to S6 until an acceptable satisfactory solution is found or the maximum number of iterations is reached;
[0081] S8: Set the initial global extremum p g The corresponding particles are used as parameters for the deep neural network to build an optimized deep neural network model.
[0082] S2 specifically includes:
[0083] According to the formula The fitness of each particle in the particle swarm is set;
[0084] Where MSE is the fitness, and n is the number of training samples; Y i For reference output; y i This is the actual output.
[0085] S6 specifically includes:
[0086] Through formula and Update the velocity and position of each particle;
[0087] Where i = 1, 2, ..., m, and m is the total number of particles; ω represents the current velocity of the i-th particle at iteration k; k c1 and c2 are inertial weights; c1 and c2 are learning factors, the former being the individual learning factor and the latter the social learning factor; k is the number of iterations; r1 and r2 are random numbers in the interval [0, 1]. Let be the current position of the i-th particle at iteration k. The formula for calculating the inertia weight is: Where, k max The maximum number of iterations is determined by using a trial-and-error algorithm to find the optimal initial and final inertia weights. Finally, ω1 = 0.9 and ω2 = 0.4 are determined to be the initial and final inertia weights, respectively.
[0088] The performance of deep neural networks largely depends on the effectiveness of the parameter training method. Particle swarm optimization (PSO) is a bio-inspired method in the field of computational intelligence, belonging to swarm intelligence optimization algorithms. It's a global stochastic search algorithm based on swarm intelligence, proposed by simulating the behavior of flocks of birds flying and foraging. In PSO, each bird in the search space is treated as a solution to an optimization problem, called a "particle." The particle's flight direction and distance are determined by its velocity and fitness value, which is provided by the objective function being optimized. In each iteration of PSO, the particles track two extreme values (individual extreme value p). i and global extremum p g It updates its speed and position.
[0089] The choice of fitness function directly affects the convergence speed of particle swarm optimization and whether it can find the optimal solution. The mean square error (MSE) of the neural network is defined as the fitness calculation function for particle swarm optimization.
[0090] This invention uses a combination of deep neural networks and particle swarm optimization to analyze and process data from multiple field points, and establishes a seismic hazard prediction model. This solves the problem of processing large amounts of data, shortens data analysis time, and improves the accuracy of seismic hazard tracking.
[0091] Example 2:
[0092] Figure 2 This is a structural diagram of the seismic hazard tracking system based on source parameters and fault structures according to the present invention. Figure 2 As shown, a seismic hazard tracking system based on source parameters and fault structures includes:
[0093] The source parameter acquisition module 201 is used to acquire source parameters under different fault structures, including fractured fault zones, strip-shaped fault zones, large strike-slip fault zones, and low-stress environments.
[0094] The hazard factor determination module 202 is used to determine the hazard factor corresponding to different seismically active fault segments based on the source parameters under the different fault structures.
[0095] Deep neural network model building module 203 is used to build deep neural network models;
[0096] The optimized earthquake hazard prediction model determination module 204 is used to take the source parameters under different fault structures as input and the hazard coefficient corresponding to the active fault segment as output, and input them into the deep neural network model for training and optimization to obtain the optimized earthquake hazard prediction model.
[0097] The earthquake source parameter acquisition module 205 for a set area is used to acquire the earthquake source parameters for a set area;
[0098] The earthquake hazard tracking module 206 is used to track the earthquake hazard of a set area based on the optimized earthquake hazard prediction model and the source parameters of the set area.
[0099] The risk factor determination module 202 specifically includes:
[0100] The subdivision result generation unit is used to subdivide the source parameters under different fault structures and generate subdivision results.
[0101] The weight setting result generation unit is used to set different weights for multiple seismic active fault segments according to the subdivision result and generate weight setting results;
[0102] The hazard coefficient determination unit is used to add up the weights corresponding to each of the active seismic fault segments according to the subdivision results and the weight setting results, and to obtain the hazard coefficients corresponding to multiple active seismic fault segments.
[0103] A seismic hazard tracking system based on source parameters and fault structures also includes:
[0104] The deep neural network model optimization module is used to optimize the deep neural network model using the particle swarm optimization algorithm to obtain an optimized deep neural network model.
[0105] Example 3:
[0106] This embodiment provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor runs the computer program to enable the electronic device to execute the seismic hazard tracking method based on source parameters and fault structures of Embodiment 1.
[0107] Alternatively, the aforementioned electronic device may be a server.
[0108] In addition, this embodiment of the invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the seismic hazard tracking method based on source parameters and fault structures of Embodiment 1.
[0109] Embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may 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.
[0110] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. 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, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0111] 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, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0112] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0113] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.
[0114] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.
Claims
1. A method for tracking seismic hazard based on source parameters and fault structures, characterized in that, include: Obtain source parameters under different fracture structures, including fractured fracture zones, strip-shaped fracture zones, large strike-slip fracture zones, and low-stress environments; Based on the source parameters under different fault structures, the hazard coefficients corresponding to different active fault segments are determined; Establish a deep neural network model; The source parameters under different fault structures are used as inputs, and the hazard coefficients corresponding to the active fault segments are used as outputs. These are then input into the deep neural network model for training and optimization to obtain an optimized earthquake hazard prediction model. Obtain the seismic source parameters for a specified region; The seismic hazard of the designated area is tracked based on the optimized seismic hazard prediction model and the source parameters of the designated area; The determination of the hazard coefficients corresponding to different seismically active fault segments based on the source parameters under different fault structures specifically includes: The source parameters under different fault structures are further subdivided to generate subdivision results; Based on the subdivision results, different weights are set for multiple active seismic fault segments to generate weight setting results; Based on the subdivision results and the weight setting results, the weights corresponding to each active seismic fault segment are added together to obtain the risk coefficients corresponding to multiple active seismic fault segments. Before inputting the aforementioned source parameters into the deep neural network model for training and optimization to obtain the optimized seismic hazard prediction model, the method further includes: optimizing the deep neural network model using a particle swarm optimization algorithm to obtain the optimized deep neural network model, specifically: S1: Randomly initialize the particle swarm; S2: According to the formula The fitness of each particle in the particle swarm is set, as shown in the formula: MSE For fitness, n The number of training samples, For reference output, This is the actual output; S3: Select the particle's current position as the initial individual extreme value. The particle with the lowest fitness in the population is selected as the initial global extremum. ; S4: Compare the current fitness with the initial individual extreme value. The fitness of the initial individual extreme value If the fitness is better, then maintain it; if the current fitness is better, then update the initial individual extreme value. ; S5: Compare the initial individual extreme values of each particle. The fitness and the initial global extremum The fitness, if the initial global extremum If the fitness is better, then the initial global extremum is maintained. If the initial individual extreme value If the fitness is better, then update the initial global extremum. ; S6: Through formula and The velocity and position of each particle are updated, as shown in the formula: , m The total number of particles; For the first i Individual particles k The current speed at the next iteration Inertial weight; These are learning factors; the former refers to individual learning factors, and the latter to social learning factors. k For the number of iterations, A random number in the interval [0,1]. For the first i Individual particles k The current position at the next iteration; the formula for calculating the inertia weight is: ,in, To maximize the number of iterations, the optimal initial and final inertia weights are found through trial and error. and These are the initial inertia weight and the final inertia weight, respectively. S7: Repeat steps S4 to S6 until an acceptable satisfactory solution is found or the maximum number of iterations is reached; S8: Set the initial global extremum The corresponding particles are used as parameters for the deep neural network to build an optimized deep neural network model.
2. The seismic hazard tracking method based on source parameters and fault structure according to claim 1, characterized in that, The source parameters are input into the deep neural network model for training and optimization to obtain an optimized seismic hazard prediction model, specifically including: The source parameters are divided into a source parameter training set and a source parameter test set; The optimized deep neural network model is used to train the source parameter training set to obtain a trained earthquake hazard prediction model; the trained earthquake hazard prediction model is then optimized using the source parameter test set to obtain an optimized earthquake hazard prediction model.
3. A seismic hazard tracking system based on source parameters and fault structures, characterized in that, The method for tracking seismic hazard based on source parameters and fault structures as described in any one of claims 1-2 includes: The source parameter acquisition module is used to acquire source parameters under different fault structures, including fractured fault zones, strip-shaped fault zones, large strike-slip fault zones, and low-stress environments. The hazard factor determination module is used to determine the hazard factor corresponding to different seismically active fault segments based on the source parameters under different fault structures; the deep neural network model building module is used to build a deep neural network model. The optimized earthquake hazard prediction model determination module is used to take the source parameters under different fault structures as input and the hazard coefficient corresponding to the active fault segment as output, and input them into the deep neural network model for training and optimization to obtain the optimized earthquake hazard prediction model. The module for acquiring seismic source parameters for a specified region is used to acquire seismic source parameters for that region. The earthquake hazard tracking module is used to track the earthquake hazard of a set area based on the optimized earthquake hazard prediction model and the source parameters of the set area.