Multi-step collaborative seismic event processing and classifying method
By combining the PhaseNet model, Gamma-Gaussian mixture algorithm, and HypoDD relocation calibration with the CEB model in a multi-step collaborative processing approach, the problems of low phase picking accuracy, large positioning error, and inaccurate classification were solved, achieving high-precision seismic event processing.
Patent Information
- Application Number
- CN202511222764.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-29
- Publication Date
- 2025-12-05
AI Technical Summary
In existing technologies, problems such as insufficient phase picking accuracy, large positioning errors, limited relocation calibration effects, and low accuracy in event classification lead to incomplete and inefficient earthquake event processing.
PhaseNet model is used for seismic phase picking, combined with Gamma Gaussian mixture algorithm and HypoDD relocation calibration, and CEB model is used for multi-dimensional data fusion to form a multi-step collaborative seismic event processing workflow, which improves the accuracy of seismic phase picking, location accuracy and classification accuracy.
This improved the accuracy of seismic phase picking, reduced positioning errors, enhanced the relocation and calibration effect, and improved the accuracy of event classification, thus forming a complete and efficient seismic event processing workflow.
Smart Images
Figure CN121069475A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of earthquake monitoring, in particular to a multi-step coordinated earthquake event processing and classification method. BACKGROUND
[0002] In the field of earthquake monitoring, accurate processing of earthquake events is the basis for earthquake early warning and disaster assessment. Phase picking is the first step in earthquake event processing, which aims to identify the arrival time of P waves (longitudinal waves) and S waves (transverse waves) in seismic waves, providing key data for subsequent positioning. Positioning determines the occurrence location and time of the earthquake event based on phase data. Relocation calibration is used to further optimize the positioning results and improve accuracy. Event classification helps to distinguish natural earthquakes from non-natural earthquakes such as explosions, which is of great significance to earthquake monitoring and safety control.
[0003] Currently, the commonly used methods for phase picking include traditional algorithms based on waveform features (such as STA / LTA algorithm) and models based on deep learning (such as PhaseNet model). Among them, PhaseNet model can better identify phases through deep learning training of seismic waveforms, but when used alone, it does not form a coordinated optimization with subsequent positioning and classification steps.
[0004] In terms of positioning, Gaussian mixture algorithm is often used to process phase data for positioning, but when used alone, it is greatly affected by phase picking errors, and the positioning accuracy is limited. HypoDD algorithm is a commonly used relocation method that uses residual information between earthquakes to optimize positioning results, but in existing technology, it is used as an independent step and not closely combined with the previous positioning algorithm.
[0005] In terms of earthquake event classification, traditional methods are mostly based on single features such as magnitude and focal depth to make judgments, and for natural earthquakes and non-natural earthquakes with similar features, the classification accuracy is low, while the CEB model has certain applications in multi-feature classification, but it does not form a data linkage with the previous phase picking, positioning and other steps.
[0006] The main shortcomings of the existing technology include:
[0007] 1. Phase picking: traditional algorithms have insufficient ability to identify weak phases, and PhaseNet model alone does not have its output results processed to adapt to subsequent positioning algorithms, affecting positioning accuracy.
[0008] 2. Positioning: single Gamma Gaussian mixture algorithm is greatly affected by phase picking error transmission, resulting in large positioning result errors; and lacks coordination with subsequent relocation steps, unable to provide favorable data basis for subsequent calibration at the initial stage of positioning.
[0009] 3. Relocation calibration aspect: HypoDD algorithm is used independently and does not form data interaction and cooperative optimization with the previous positioning algorithm, resulting in limited calibration effect and difficulty in significantly reducing positioning error.
[0010] 4. Event classification aspect: relying on single feature or independent classification model, not fully utilizing multi-dimensional data generated in the previous phase of seismic phase picking and positioning, and lacking classification accuracy and robustness. SUMMARY
[0011] This section is intended to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification of the present application to avoid obscuring the purpose of this section, abstract and title, and such simplifications or omissions cannot be used to limit the scope of the present application.
[0012] 1. Technical problem to be solved:
[0013] In order to solve the problems of insufficient seismic phase picking accuracy, large positioning error, limited relocation calibration effect and low event classification accuracy in the earthquake event processing, specifically, the traditional seismic phase picking method has poor recognition ability for weak seismic phase, resulting in inaccurate subsequent positioning basis data; a single positioning algorithm cannot balance positioning speed and accuracy; the relocation calibration step does not smoothly connect with the previous positioning, and cannot effectively reduce the error; the classification of seismic events relies on a single feature and cannot accurately distinguish natural earthquakes from non-natural earthquakes. The present application aims to solve the problems of low seismic phase picking accuracy, large positioning error, poor calibration effect and inaccurate classification through multi-step cooperative processing, and proposes the present application.
[0014] Therefore, the purpose of the present application is to provide a multi-step cooperative earthquake event processing and classification method, to improve the seismic phase picking accuracy, especially the recognition ability for weak seismic phase, and to make the picking result more suitable for subsequent positioning algorithm; to combine Gamma Gaussian mixture algorithm and HypoDD relocation calibration to reduce positioning error and improve earthquake event positioning accuracy; to use multi-dimensional data generated in the previous processing process to improve the classification accuracy of natural earthquakes and non-natural earthquakes through CEB model; to realize the cooperative linkage of seismic phase picking, positioning, relocation calibration and classification steps, and to form a complete and efficient earthquake event processing flow.
[0015] 2. Technical solution:
[0016] To solve the above technical problems, according to one aspect of the present application, the present application provides the following technical solution:
[0017] A multi-step cooperative earthquake event processing and classification method, comprising the following steps:
[0018] S1: a step of picking up a seismic phase, adopting a PhaseNet model to process input original seismic waveform data to obtain original seismic phase features;
[0019] S2: a positioning step, inputting original seismic phase feature arrival time data obtained in the step S1 into a Gamma Gaussian mixture algorithm to obtain a preliminary positioning result;
[0020] S3: a relocation calibration step, inputting the preliminary positioning result obtained in the step S2 as an initial value into a HypoDD algorithm to perform relocation calibration to obtain a relocation calibrated feature;
[0021] S4: an event classification step.
[0022] As a preferred scheme of the multi-step cooperative seismic event processing and classification method, in the step S1, the original seismic waveform data is preprocessed to remove noise interference, and then the preprocessed waveform data is input into the trained PhaseNet model, and the model outputs new seismic phase features.
[0023] As a preferred scheme of the multi-step cooperative seismic event processing and classification method, in the step S1, the original seismic waveform data is preprocessed to remove noise interference, and then the preprocessed waveform data is input into the trained PhaseNet model, and the model outputs new seismic phase features.
[0024] As a preferred scheme of the multi-step cooperative seismic event processing and classification method, in the step S3, the HypoDD algorithm uses relative travel time residuals between seismic events to iteratively optimize the preliminary positioning result.
[0025] As a preferred scheme of the multi-step cooperative seismic event processing and classification method, in the step S4, multi-dimensional data such as new seismic phase features and the relocation calibrated features in the step S3 are collected, and the data is input into the CEB model.
[0026] As a preferred scheme of the multi-step cooperative seismic event processing and classification method, in the step S1, the original seismic waveform data includes three-component seismic records.
[0027] As a preferred scheme of the multi-step cooperative seismic event processing and classification method, in the step S1, the original seismic waveform data includes three-component seismic records.
[0028] As a preferred solution of the multi-step cooperative seismic event processing and classification method of the application, the features of the relocation calibration in step S3 include focal depth and epicenter position.
[0029] 3. Beneficial effects:
[0030] Compared with the prior art, the application has the beneficial effects that:
[0031] The multi-step cooperative seismic event processing and classification method:
[0032] 1. Higher precision of phase picking: through the PhaseNet model combined with confidence screening, not only the model's ability to identify phases is retained, but also low-quality data is eliminated, improving the quality of the basis data for subsequent positioning;
[0033] 2. Higher positioning accuracy: the preliminary positioning of the Gamma Gaussian mixture algorithm and the HypoDD relocation calibration work cooperatively, the HypoDD algorithm takes the preliminary positioning result as the initial value, reducing the difficulty of iterative optimization and improving the relocation efficiency and accuracy;
[0034] 3. Higher classification accuracy: the CEB model integrates multi-dimensional features of phases and positioning, which can more accurately distinguish natural earthquakes from non-natural earthquakes compared with single feature classification;
[0035] 4. Stronger process integrity: each step forms an organic whole with good data linkage, avoiding the low efficiency and poor effect caused by independent operation of each step in the prior art, and improving the quality and efficiency of seismic event processing as a whole. BRIEF DESCRIPTION OF DRAWINGS
[0036] In order to more clearly illustrate the technical solutions of the embodiments of the application, the application will be described in detail below with reference to the drawings and detailed embodiments. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor. Among them:
[0037] Figure 1 The flowchart of the multi-step cooperative seismic event processing and classification method of the application. DETAILED DESCRIPTION
[0038] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the specific embodiments of the application will be described in detail below with reference to the drawings.
[0039] The present application is described in detail in conjunction with the schematic diagram, and in the detailed description of the embodiments of the present application, the cross-sectional view of the device structure is partially enlarged without the general proportion for the convenience of illustration, and the schematic diagram is only an example, which should not limit the scope of protection of the present application herein. In addition, the three-dimensional spatial dimensions of length, width and depth should be included in actual production.
[0040] The orientation or positional relationship indicated in the term is based on the orientation or positional relationship shown in the drawing, which is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.
[0041] The connection mode in the term should be understood broadly, for example, "connection" can be fixed connection, can be detachable connection, or integral connection; can be mechanical connection, can be electrical connection; can be directly connected, can be indirectly connected through an intermediate medium, and can be internal communication of two elements. For those skilled in the art, the specific meaning of the above-mentioned term in the present application can be understood according to the specific circumstances.
[0042] The embodiments of the present application will be further described in detail below in conjunction with the drawings.
[0043] The present application provides a schematic diagram of the overall structure of an embodiment of a multi-step cooperative seismic event processing and classification method, which comprises:
[0044] Please refer to Figure 1 The multi-step cooperative seismic event processing and classification method of the present embodiment comprises the following steps:
[0045] S1: a phase picking step, using a PhaseNet model to process the input original seismic waveform data to obtain original phase features;
[0046] S2: a positioning step, inputting the original phase feature arrival time data obtained in step S1 into a Gamma Gaussian mixture algorithm to obtain a preliminary positioning result;
[0047] S3: a relocation calibration step, inputting the preliminary positioning result obtained in step S2 as an initial value into a HypoDD algorithm for relocation calibration to obtain a relocation calibrated feature;
[0048] S4: an event classification step.
[0049] It is worth noting that specifically, in step S1, the original seismic waveform data is preprocessed to remove noise interference, and then the preprocessed waveform data is input into the trained PhaseNet model, and the model outputs new phase features.
[0050] Then, specifically, the new seismic phase feature data is input into the Gamma Gaussian mixture algorithm, a probability model is constructed according to the new seismic phase feature data, it is assumed that the error of the seismic phase arrival time obeys Gaussian distribution, the actual data distribution is fitted by mixing multiple Gaussian distributions, and then the possible position and occurrence time of the seismic event are calculated based on the model to obtain the preliminary positioning result.
[0051] Meanwhile, specifically, in step S3, the HypoDD algorithm uses the relative travel time residual between seismic events to iteratively optimize the preliminary positioning result.
[0052] Further, specifically, in step S4, multi-dimensional data such as new seismic phase features and the features calibrated by relocation in step S3 are collected, and the data is input into the CEB model.
[0053] Meanwhile, specifically, the original seismic waveform data in step S1 includes three-component seismic records.
[0054] Then, specifically, the original seismic phase features in step S1 are high-quality P-wave and S-wave arrival times and picking results.
[0055] It should be noted that, specifically, the features calibrated by relocation in step S3 include focal depth and epicenter position.
[0056] Embodiment 1:
[0057] S1: Seismic phase picking step, using the PhaseNet model to process the input original seismic waveform data to obtain original seismic phase features;
[0058] S2: Positioning step, inputting the original seismic phase feature arrival time data obtained in step S1 into the Gamma Gaussian mixture algorithm to obtain a preliminary positioning result;
[0059] S3: Relocation calibration step, inputting the preliminary positioning result obtained in step S2 as an initial value into the HypoDD algorithm for relocation calibration to obtain features calibrated by relocation;
[0060] S4: Event classification step.
[0061] It should be noted that, specifically, in step S1, the original seismic waveform data is preprocessed to remove noise interference, and then the preprocessed waveform data is input into the trained PhaseNet model, the model outputs new seismic phase features, to improve the subsequent positioning accuracy, the new seismic phase results output by the model are screened, and seismic phase data with a confidence lower than a set threshold (such as 0.8) are removed, to obtain high-quality P-wave and S-wave picking results.
[0062] Then, specifically, the new seismic phase feature data is input into the Gamma Gaussian mixture algorithm, a probability model is constructed according to the new seismic phase feature data, it is assumed that the error of the seismic phase arrival time obeys Gaussian distribution, the actual data distribution is fitted by mixing multiple Gaussian distributions, and then the possible position and occurrence time of the seismic event are calculated based on the model to obtain the preliminary positioning result.
[0063] Meanwhile, specifically, in step S3, the HypoDD algorithm uses the relative travel time residual between seismic events to iteratively optimize the preliminary positioning result, specifically, the residual information between different seismic event pairs is calculated to construct an objective function, the position and occurrence time of the seismic event are adjusted by minimizing the objective function to obtain the calibrated high-precision positioning result.
[0064] Further, specifically, in step S4, multi-dimensional data such as new seismic phase features and the repositioned and calibrated features in step S3 are collected, and the data is input into the CEB model, the CEB model outputs the classification result of the seismic event as a natural earthquake or a non-natural earthquake by learning and training the multi-feature data.
[0065] Meanwhile, specifically, the original seismic waveform data in step S1 includes three-component seismic records.
[0066] Then, specifically, the original seismic phase features in step S1 are high-quality P-wave and S-wave arrival times and picking results.
[0067] It should be noted that, specifically, the repositioned and calibrated features in step S3 include focal depth and epicenter position.
[0068] Embodiment 2:
[0069] It should be noted that, specifically, in step S1, the original seismic waveform data is preprocessed to remove noise interference, and then the preprocessed waveform data is input into the trained PhaseNet model, and the model outputs new seismic phase features. In order to improve the subsequent positioning accuracy, the new seismic phase results output by the model are screened, and seismic phase data with a confidence level lower than a set threshold (such as 0.8) are removed, to obtain high-quality P-wave and S-wave picking results. The training data set of the PhaseNet model and the main parameter settings (such as the number of network layers and the number of iterations).
[0070] Next, specifically, the new seismic phase feature data is input into the Gamma Gaussian mixture algorithm, a probability model is constructed according to the new seismic phase feature data, it is assumed that the error of the seismic phase arrival time obeys Gaussian distribution, the actual data distribution is fitted by mixing multiple Gaussian distributions, then the possible position and the time of the earthquake event are calculated based on the model to obtain the preliminary positioning result, and the selection basis and parameter optimization method of the number of Gaussian distributions in the Gamma Gaussian mixture algorithm.
[0071] Meanwhile, specifically, in step S3, the HypoDD algorithm uses the relative travel time residual between earthquake events to iteratively optimize the preliminary positioning result, specifically, the residual information between different earthquake event pairs is calculated to construct an objective function, the position and the time of the earthquake event are adjusted by minimizing the objective function to obtain the calibrated high-precision positioning result, and the setting standard of the number of iterations and the residual threshold in the HypoDD algorithm.
[0072] Further, specifically, in step S4, new seismic phase features, features repositioned and calibrated in step S3 and multi-dimensional data are collected, the data is input into the CEB model, the CEB model learns and trains the multi-feature data, and outputs the classification result of the earthquake event as a natural earthquake or a non-natural earthquake, and the specific types and extraction methods of the multi-feature data used by the CEB model.
[0073] Embodiment 3
[0074] In the seismic phase picking step, other deep learning models (such as the EQTransformer model) can be used instead of the PhaseNet model to achieve the seismic phase picking function, and only the input and output formats of the model need to be adapted;
[0075] In the positioning step, the K-means clustering algorithm can be used instead of the Gamma Gaussian mixture algorithm for preliminary positioning, although there is a slight difference in probability distribution fitting, but it can still be used as a preliminary positioning means to provide an initial value for subsequent repositioning and calibration;
[0076] In the event classification step, the SVM (Support Vector Machine) model can be used instead of the CEB model to classify natural earthquakes and non-natural earthquakes by training multi-feature data, and the parameters of the model need to be adjusted to adapt to the input features.
[0077] Embodiment 4
[0078] The flow chart of the multi-step cooperative seismic event processing and classification method comprises the connection relationship of the above four steps, which are in sequence: phase picking step (relying on PhaseNet model) → positioning step (relying on Gamma Gaussian mixture algorithm positioning) → relocation calibration step (relying on HypoDD algorithm) → event classification step (relying on CEB model classification). Each step is represented by a box, and the arrow represents the data flow direction.
[0079] Taking the flow chart of the multi-step cooperative seismic event processing and classification method as an example, the seismic waveform data first enters the PhaseNet model module, the output P wave and S wave data flow into the Gamma Gaussian mixture algorithm module, the preliminary positioning result obtained after processing enters the HypoDD relocation calibration module, and the final positioning result and the phase feature are jointly input into the CEB model module, and the classification result is output.
[0080] Although the present application has been described with reference to the embodiments above, various modifications can be made thereto and equivalents can be substituted therefor without departing from the scope of the present application. In particular, features of the disclosed embodiments can be combined together in any manner, provided that there is no structural conflict. The combinations are not exhaustively described in the specification only for the purpose of omitting the length and saving resources. Therefore, the present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.
Claims
1. A multi-step synergistic seismic event processing and classification method, characterized in that, The method comprises the following steps: S1: a phase picking step, wherein a PhaseNet model is used to process inputted original seismic waveform data to obtain original phase features; S2: a positioning step, wherein original phase feature arrival time data obtained in step S1 is inputted into a Gamma Gaussian mixture algorithm to obtain a preliminary positioning result; S3: a relocation calibration step, wherein the preliminary positioning result obtained in step S2 is inputted into a HypoDD algorithm as an initial value to perform relocation calibration to obtain a relocated and calibrated feature; S4: an event classification step.
2. The multi-step synergistic seismic event processing and classification method of claim 1, wherein, In step S1, the original seismic waveform data is preprocessed to remove noise interference, and then the preprocessed waveform data is inputted into a trained PhaseNet model, and the model outputs new phase features.
3. The multi-step synergistic seismic event processing and classification method of claim 2, wherein, The data of the new phase features is inputted into a Gamma Gaussian mixture algorithm, a probability model is constructed according to the data of the new phase features, it is assumed that the error of phase arrival time obeys a Gaussian distribution, a plurality of Gaussian distributions are mixed to fit the actual data distribution, and then the possible position and the occurrence time of a seismic event are calculated based on the model to obtain a preliminary positioning result.
4. The multi-step synergistic seismic event processing and classification method of claim 1, wherein, In step S3, the HypoDD algorithm uses the relative travel time residual between seismic events to iteratively optimize the preliminary positioning result.
5. The multi-step synergistic seismic event processing and classification method of claim 2, wherein, In step S4, multi-dimensional data such as new phase features and the relocated and calibrated features in step S3 are collected, and the data is inputted into a CEB model.
6. The multi-step synergistic seismic event processing and classification method of claim 1, wherein, The original seismic waveform data in step S1 comprises three-component seismic records.
7. The multi-step synergistic seismic event processing and classification method of claim 1, wherein, The original phase features in step S1 are high-quality P-wave and S-wave arrival times and picking results.
8. The multi-step synergistic seismic event processing and classification method of claim 1, wherein, The relocated and calibrated features in step S3 comprise a focal depth and a hypocenter position.
Citation Information
Cited By
Multi-source fusion seismic event automatic construction system and method
CN121806114A