Quality evaluation method for earthquake disaster equipment monitoring data

Through the DBSCAN algorithm, multivariate linear regression attribute fusion method, K-means algorithm, principal component analysis method and improved convolutional neural network, the problem of single indicator of earthquake monitoring data evaluation method in the existing technology is solved, and the earthquake monitoring data is realized The comprehensive quality evaluation of the data improves the operability and practicality of the data and provides support for earthquake prediction and disaster reduction.

CN119064994APending Publication Date: 2024-12-03HAINAN KELIQIANFANG TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410541667.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-04-30
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing seismic monitoring data evaluation methods usually only focus on a single indicator, ignore the comprehensiveness of the data, and lack operability and practicality, so it is impossible to comprehensively evaluate the quality of seismic monitoring data.

Method used

The DBSCAN algorithm is used to detect waveform anomaly data, combine the multivariate linear regression attribute fusion method and the K-means algorithm for data fusion and cluster analysis, use principal component analysis method and improved convolutional neural network for feature extraction, and according to the quality evaluation standard Carry out comprehensive quality evaluation.

Benefits of technology

Effectively divide seismic waveform data, comprehensively evaluate the completeness, accuracy, reliability and timeliness of earthquake monitoring data, improve the operability and practicality of data, and provide support for earthquake prediction and disaster reduction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119064994A_ABST
    Figure CN119064994A_ABST
Patent Text Reader

Abstract

The invention provides a quality evaluation method for earthquake disaster equipment monitoring data, and the method comprises the following steps: S1, collecting earthquake monitoring data from different earthquake disaster detection equipment, and carrying out the noise reduction of the collected earthquake monitoring data; s2, introducing a DBSCAN algorithm to carry out waveform abnormal data detection on the seismic monitoring data subjected to noise reduction processing; s3, performing data fusion and clustering analysis on the seismic monitoring data subjected to noise reduction processing and waveform abnormal data monitoring; s4, based on a principal component analysis method and a convolutional neural network, performing feature extraction on the seismic monitoring data subjected to data fusion; s5, manually setting a quality evaluation standard and a quality inspection condition, and performing quality evaluation on the feature extraction result of the seismic monitoring data; and S6, visually displaying the quality evaluation result. According to the method, the integrity, the reliability and the timeliness of the earthquake monitoring data are comprehensively considered, the overall quality of the data can be comprehensively evaluated, and the operability and the practicability are considered at the same time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of earthquake monitoring data quality evaluation, and in particular to a quality evaluation method for earthquake disaster equipment monitoring data. Background Art

[0002] When an earthquake occurs, earthquake disaster monitoring equipment can obtain various earthquake-related data in real time, such as earthquake magnitude, focal depth, epicenter location, etc., and the quality of these data is directly related to the accuracy of earthquake disaster assessment. In existing technologies, some evaluation methods for earthquake monitoring data focus on the integrity of the data, such as whether the data is complete and whether there are omissions. Other methods focus more on the accuracy of the data, such as whether the data conforms to the actual situation and whether there are errors. Still other methods focus on the reliability of the data, such as whether the data is credible and whether it is repeatable. Finally, some methods also consider the timeliness of the data, that is, whether the data reflects the situation of earthquake disasters in a timely manner. However, existing evaluation methods often only focus on a single indicator, ignore the importance of other indicators, and cannot comprehensively evaluate different types of data; at the same time, existing evaluation methods do not have operability and practicality in practical applications. Summary of the invention

[0003] In view of this, an object of the present invention is to provide a quality evaluation method for earthquake disaster equipment monitoring data to at least solve the above problems.

[0004] To achieve the above-mentioned object of the invention, the present invention provides a quality evaluation method for earthquake disaster equipment monitoring data, the method comprising the following steps:

[0005] S1. Collect earthquake monitoring data from different earthquake disaster detection equipment and perform noise reduction on the collected earthquake monitoring data;

[0006] S2, introduce DBSCAN algorithm to detect waveform abnormal data of seismic monitoring data that has completed noise reduction processing;

[0007] S3, performing data fusion and cluster analysis on the earthquake monitoring data that has completed noise reduction processing and waveform abnormality data monitoring;

[0008] S4. Based on principal component analysis and convolutional neural network, feature extraction is performed on the earthquake monitoring data after data fusion;

[0009] S5. Artificially set quality assessment standards and quality inspection conditions to evaluate the quality of feature extraction results of earthquake monitoring data;

[0010] S6. Visualize the quality evaluation results.

[0011] It should be noted that step S2 specifically includes:

[0012] S21, forming a data set from all the seismic waveform data in the seismic monitoring data, wherein any seismic waveform data is a data point;

[0013] S22, filtering the seismic waveform data in the data set through a filter;

[0014] S23, using short-time Fourier transform to extract features from the seismic waveform data in the data set to obtain each feature vector;

[0015] S24, using Euclidean distance to calculate the distance matrix between each feature vector, which is used to represent the domain relationship of each data point in the data set;

[0016] S25, using the K-distance graph, determining the neighborhood radius ε value and the density threshold minPts value;

[0017] S26. The determined ε value and minPts value are used as parameters and introduced into the DBSCAN algorithm for clustering.

[0018] It should be further explained that step S22 specifically includes:

[0019] S221, using a Gaussian window to move to different positions of the seismic waveform data, performing window processing on the data at each position to obtain a segment of intercepted data, and performing fast Fourier transform on each segment of intercepted data to obtain a frequency spectrum of each segment of data;

[0020] S222, repeating step S221 until the window function moves to the end of the seismic waveform data, obtaining the frequency spectra of all segments of data, and arranging all the frequency spectra in the order of time and frequency to form a two-dimensional matrix, and calculating and obtaining the eigenvector of the matrix;

[0021] It should be further explained that step S25 specifically includes:

[0022] S251, set k value, let minPts = k + 1, where k = 2d-1 value, and d is the characteristic of seismic waveform data;

[0023] S252, calculating the distance from any data point to the kth nearest neighbor point;

[0024] S253. Arrange the k-distances of all data points in descending order and draw a curve, wherein the k-distance value corresponding to the first inflection point position in the k-distance graph is taken as the ε value.

[0025] It should be further explained that step S26 specifically includes:

[0026] S261, for each data point in the data set, calculate the number of points in its ε-neighborhood and compare them, wherein when the number of points in the ε-neighborhood of the data point exceeds the minPts value, the data point is marked as a core point, otherwise it is temporarily marked as a noise point;

[0027] S262, for each core point, connect the core point with all core points in its ε-neighborhood to form a cluster, and mark all non-core points in its ε-neighborhood as boundary points and attribute them to the cluster;

[0028] S263, repeating step S24, steps S251 to S253, and steps S261 to S262 until all core points are processed;

[0029] S264: Mark the unattributed noise points and obtain the final clustering result.

[0030] It should be noted that step S3 specifically includes:

[0031] S31, integrating earthquake monitoring data from different sources that have completed noise reduction processing and waveform abnormality data monitoring, and separating different attributes of the earthquake data, wherein the earthquake attributes of the earthquake monitoring data include waveform, frequency, amplitude, time, location, magnitude, focal depth, damage caused by the earthquake, and impact range;

[0032] S32, using a multivariate linear regression attribute fusion method to perform feature fusion on different attributes of the earthquake monitoring data to obtain predicted values ​​of different attributes;

[0033] S33, using cosine similarity to determine the similarity between each prediction value;

[0034] S34. According to the K-means algorithm, each predicted value is divided into K clusters to obtain the final cluster division result.

[0035] It should be noted that the specific steps of performing feature fusion in step S32 include:

[0036] S321, determining a target attribute and a predicted attribute, wherein the target attribute is the severity of the earthquake event, and the predicted attribute is different attributes of the earthquake;

[0037] S322, establish a multiple linear regression model, the formula of the multiple linear regression model is:

[0038] y=β0+β1x1+β2x2+...+β k x k +ε

[0039] Among them, y is the target attribute, x1, x2, ..., x kare the predicted attributes, β0, β1, ..., β k is the model parameter, ε is the error term;

[0040] S323, using the least squares method to estimate the value of the model parameters, the process is achieved by minimizing the residual sum of squares, and the process is mathematically expressed as:

[0041]

[0042] S324, using the estimated model parameters, substituting different attributes of each earthquake monitoring data into a multivariate linear regression model for prediction to obtain a predicted value;

[0043] The specific steps of step S34 include:

[0044] S341, randomly selecting K predicted values ​​as initial cluster centers;

[0045] S342, for each predicted value, calculating its distance from each cluster center, and assigning it to the cluster where the nearest cluster center is located;

[0046] S343, recalculating the cluster center of each cluster;

[0047] S344, repeat B02 and B03 until the cluster center no longer changes or the maximum number of iterations is reached, and the final cluster division result is obtained.

[0048] It should be noted that step S4 specifically includes:

[0049] S41. Using principal component analysis (PCA) to extract features from the seismic monitoring data after data fusion, a characteristic mother wave time history is obtained, and the error of the characteristic mother wave time history is adjusted based on amplitude, spectrum and duration;

[0050] S42, locally improving the convolutional neural network, used to adapt to the input of waveform data in earthquake monitoring data and the picking of the first arrival time of earthquake P waves, wherein the specific steps of locally improving the convolutional neural network are:

[0051] A01. Use one-dimensional convolutional layers instead of two-dimensional convolutional layers to adapt to the one-dimensionality of waveform data and maintain multi-channel input and output.

[0052] A02. Use the bidirectional recurrent neural network layer BiRNN as the last layer of the convolutional neural network;

[0053] A03. Input the seismic waveform data into the locally improved convolutional neural network to extract the first arrival time of the seismic P wave.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention provides a quality evaluation method for earthquake disaster equipment monitoring data. By using the DBSCAN algorithm, earthquake waveform data is effectively divided; the fusion and cluster analysis of earthquake monitoring data are completed by the multivariate linear regression attribute fusion method and the K-means algorithm, so that subsequent relevant staff can better understand the internal structure and law of earthquake data, thereby providing strong support for earthquake prediction and disaster reduction; the principal component analysis method PCA and the improved convolutional neural network are used to better adapt to the content and format of earthquake monitoring data, complete the feature extraction of earthquake monitoring data, and facilitate subsequent staff to perform quality evaluation according to the set quality assessment standards and quality inspection conditions. The present invention comprehensively considers the integrity, accuracy, reliability and timeliness of earthquake monitoring data, and can comprehensively evaluate the overall quality of earthquake monitoring data, while taking into account operability and practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0057] Figure 1 It is a schematic diagram of the overall process of a quality evaluation method for earthquake disaster equipment monitoring data provided by an embodiment of the present invention.

[0058] Figure 2 It is a specific flow chart of step S2 in a quality evaluation method for earthquake disaster equipment monitoring data provided by an embodiment of the present invention.

[0059] Figure 3 It is a specific flow chart of step S3 in a quality evaluation method for earthquake disaster equipment monitoring data provided by an embodiment of the present invention.

[0060] Figure 4 It is a flowchart diagram of specific steps of locally improving a convolutional neural network in a quality evaluation method for earthquake disaster equipment monitoring data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The enumerated embodiments are only used to explain the present invention and are not used to limit the scope of the present invention.

[0062] Reference Figure 1 This embodiment provides a method for evaluating the quality of earthquake disaster equipment monitoring data, the method comprising the following steps:

[0063] S1. Collect earthquake monitoring data from different earthquake disaster detection equipment and perform noise reduction on the collected earthquake monitoring data;

[0064] S2, introduce DBSCAN algorithm to detect waveform abnormal data of seismic monitoring data that has completed noise reduction processing;

[0065] S3, performing data fusion and cluster analysis on the earthquake monitoring data that has completed noise reduction processing and waveform abnormality data monitoring;

[0066] S4. Based on principal component analysis and convolutional neural network, feature extraction is performed on the earthquake monitoring data after data fusion;

[0067] S5. Artificially set quality assessment standards and quality inspection conditions to evaluate the quality of feature extraction results of earthquake monitoring data;

[0068] S6. Visualize the quality evaluation results.

[0069] It should be noted that step S2 specifically includes:

[0070] S21, forming a data set from all the seismic waveform data in the seismic monitoring data, wherein any seismic waveform data is a data point;

[0071] S22, filtering the seismic waveform data in the data set through a filter;

[0072] S23, using short-time Fourier transform to extract features from the seismic waveform data in the data set to obtain each feature vector;

[0073] S24, using Euclidean distance to calculate the distance matrix between each feature vector, which is used to represent the domain relationship of each data point in the data set;

[0074] S25, using the K-distance graph, determining the neighborhood radius ε value and the density threshold minPts value;

[0075] S26. The determined ε value and minPts value are used as parameters and introduced into the DBSCAN algorithm for clustering.

[0076] It should be further explained that step S22 specifically includes:

[0077] S221, using a Gaussian window to move to different positions of the seismic waveform data, performing window processing on the data at each position to obtain a segment of intercepted data, and performing fast Fourier transform on each segment of intercepted data to obtain a frequency spectrum of each segment of data;

[0078] S222, repeating step S221 until the window function moves to the end of the seismic waveform data, obtaining the frequency spectra of all segments of data, and arranging all the frequency spectra in the order of time and frequency to form a two-dimensional matrix, and calculating and obtaining the eigenvector of the matrix;

[0079] It should be further explained that step S25 specifically includes:

[0080] S251, set k value, let minPts = k + 1, where k = 2d-1 value, and d is the characteristic of seismic waveform data;

[0081] S252, calculating the distance from any data point to the kth nearest neighbor point;

[0082] S253. Arrange the k-distances of all data points in descending order and draw a curve, wherein the k-distance value corresponding to the first inflection point position in the k-distance graph is taken as the ε value.

[0083] It should be further explained that step S26 specifically includes:

[0084] S261, for each data point in the data set, calculate the number of points in its ε-neighborhood and compare them, wherein when the number of points in the ε-neighborhood of the data point exceeds the minPts value, the data point is marked as a core point, otherwise it is temporarily marked as a noise point;

[0085] S262, for each core point, connect the core point with all core points in its ε-neighborhood to form a cluster, and mark all non-core points in its ε-neighborhood as boundary points and attribute them to the cluster;

[0086] S263, repeating step S24, steps S251 to S253, and steps S261 to S262 until all core points are processed;

[0087] S264: Mark the unattributed noise points and obtain the final clustering result.

[0088] It should be noted that the k-distance graph is used in S25 to determine the neighborhood radius ε value and the density threshold minPts value. First, a k value needs to be determined. The k value is 2 multiplied by the dimension of the data minus 1, that is, k=2d-1, where d is the number of features. Then, for each data point in the data set, the distance from the data point to the kth nearest neighbor is calculated, that is, the k-distance of the data point. Arrange the k-distances of all data points in descending order and draw a curve. The k-distance value corresponding to the first inflection point in the k-distance graph is a good ε value. The value of minPts is determined based on the k value, and minPts=k+1 is taken, which ensures that there are at least k points in the neighborhood of the core point.

[0089] It should be noted that the DBSCAN algorithm is a density-based spatial clustering algorithm, which is a method of unsupervised learning. In unsupervised learning, no pre-labeled targets are used to cluster data points. The DBSCAN algorithm does not require the user to pre-set the number of clusters, and can find clusters with complex shapes, and can find points that do not belong to any cluster. The DBSCAN algorithm can effectively detect outliers in seismic waveform data.

[0090] It should be noted that step S3 specifically includes:

[0091] S31, integrating earthquake monitoring data from different sources that have completed noise reduction processing and waveform abnormality data monitoring, and separating different attributes of the earthquake data, wherein the earthquake attributes of the earthquake monitoring data include waveform, frequency, amplitude, time, location, magnitude, focal depth, damage caused by the earthquake, and impact range;

[0092] S32, using a multivariate linear regression attribute fusion method to perform feature fusion on different attributes of the earthquake monitoring data to obtain predicted values ​​of different attributes;

[0093] S33, using cosine similarity to determine the similarity between each prediction value;

[0094] S34. According to the K-means algorithm, each predicted value is divided into K clusters to obtain the final cluster division result.

[0095] It should be noted that the specific steps of performing feature fusion in step S32 include:

[0096] S321, determining a target attribute and a predicted attribute, wherein the target attribute is the severity of the earthquake event, and the predicted attribute is different attributes of the earthquake;

[0097] S322, establish a multiple linear regression model, the formula of the multiple linear regression model is:

[0098] y=β0+β1x1+β2x2+...+β k x k +ε

[0099] Among them, y is the target attribute, x1, x2, ..., x k are the predicted attributes, β0, β1, ..., β k is the model parameter, ε is the error term;

[0100] S323, using the least squares method to estimate the value of the model parameters, the process is achieved by minimizing the residual sum of squares, and the process is mathematically expressed as:

[0101]

[0102] S324, using the estimated model parameters, substituting different attributes of each earthquake monitoring data into a multivariate linear regression model for prediction to obtain a predicted value;

[0103] The specific steps of step S34 include:

[0104] S341, randomly selecting K predicted values ​​as initial cluster centers;

[0105] S342, for each predicted value, calculating its distance from each cluster center, and assigning it to the cluster where the nearest cluster center is located;

[0106] S343, recalculating the cluster center of each cluster;

[0107] S344, repeat B02 and B03 until the cluster center no longer changes or the maximum number of iterations is reached, and the final cluster division result is obtained.

[0108] It should be noted that the multivariate linear regression attribute fusion method can make full use of the structural and lithological information contained in the seismic attributes, overcome the shortcomings of insufficient display of a single seismic attribute and the inability of a single attribute color to highlight regional anomalies, and can also improve the ability to extract geological bodies from multiple attributes, making the image display clearer.

[0109] It should be noted that the fusion and cluster analysis of earthquake monitoring data are completed through the multivariate linear regression attribute fusion method and K-means algorithm, which makes it easier for subsequent relevant staff to better understand the internal structure and laws of earthquake data, thereby providing strong support for earthquake prediction and disaster reduction.

[0110] It should be noted that step S4 specifically includes:

[0111] S41. Using principal component analysis (PCA) to extract features from the seismic monitoring data after data fusion, a characteristic mother wave time history is obtained, and the error of the characteristic mother wave time history is adjusted based on amplitude, spectrum and duration;

[0112] It should be noted that the characteristic mother wave time series obtained by principal component analysis (PCA) can retain the main information of the original data to a certain extent.

[0113] It should be noted that because microseismic waveform data can be regarded as one-dimensional multi-channel image data, each channel corresponds to the record of a detector, and each pixel corresponds to the amplitude value of a sampling point, the traditional convolutional neural network is partially improved to adapt to the characteristics of seismic waveform data.

[0114] S42, locally improving the convolutional neural network, used to adapt to the input of waveform data in earthquake monitoring data and the picking of the first arrival time of earthquake P waves, wherein the specific steps of locally improving the convolutional neural network are:

[0115] A01. Use one-dimensional convolutional layers instead of two-dimensional convolutional layers to adapt to the one-dimensionality of waveform data and maintain multi-channel input and output.

[0116] A02. Use the bidirectional recurrent neural network layer BiRNN as the last layer of the convolutional neural network;

[0117] A03. Input the seismic waveform data into the locally improved convolutional neural network to extract the first arrival time of the seismic P wave.

[0118] It should be noted that the quality assessment standard is manually configured based on national standards such as GB / T 33685-2023 Technical Specification for Seismic Exploration Data Processing and GB / T 33684-2017 Technical Specification for Seismic Exploration Data Interpretation. At the same time, the staff configures the quality inspection conditions based on the configured quality standards. Then, based on the quality standards and quality inspection conditions, the quality evaluation of the feature extraction results of the seismic monitoring data is performed. Finally, the quality evaluation results are visualized.

[0119] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A quality evaluation method for earthquake disaster equipment monitoring data, characterized in that: The method comprises the following steps: S1. Collect earthquake monitoring data from different earthquake disaster detection equipment and perform noise reduction on the collected earthquake monitoring data; S2, introduce DBSCAN algorithm to detect waveform abnormal data of seismic monitoring data that has completed noise reduction processing; S3, performing data fusion and cluster analysis on the earthquake monitoring data that has completed noise reduction processing and waveform abnormality data monitoring; S4. Based on principal component analysis and convolutional neural network, feature extraction is performed on the earthquake monitoring data after data fusion; S5. Artificially set quality assessment standards and quality inspection conditions to evaluate the quality of feature extraction results of earthquake monitoring data; S6. Visualize the quality evaluation results.

2. A quality evaluation method for earthquake disaster equipment monitoring data according to claim 1, characterized in that: Step S2 specifically includes: S21, forming a data set from all the seismic waveform data in the seismic monitoring data, wherein any seismic waveform data is a data point; S22, filtering the seismic waveform data in the data set through a filter; S23, using short-time Fourier transform to extract features from the seismic waveform data in the data set to obtain each feature vector; S24, using Euclidean distance to calculate the distance matrix between each feature vector, which is used to represent the domain relationship of each data point in the data set; S25, using the K-distance graph, determining the neighborhood radius ε value and the density threshold minPts value; S26. The determined ε value and minPts value are used as parameters and introduced into the DBSCAN algorithm for clustering.

3. A quality evaluation method for earthquake disaster equipment monitoring data according to claim 2, characterized in that: Step S22 specifically includes: S221, using a Gaussian window to move to different positions of the seismic waveform data, performing window processing on the data at each position to obtain a segment of intercepted data, and performing fast Fourier transform on each segment of intercepted data to obtain a frequency spectrum of each segment of data; S222, repeat step S221 until the window function moves to the end of the seismic waveform data, obtain the frequency spectra of all segments of data, and arrange all the frequency spectra in the order of time and frequency to form a two-dimensional matrix, and calculate and obtain the eigenvector of the matrix.

4. A quality evaluation method for earthquake disaster equipment monitoring data according to claim 3, characterized in that: Step S25 specifically includes: S251, set k value, let minPts = k + 1, where k = 2d-1 value, and d is the characteristic of seismic waveform data; S252, calculating the distance from any data point to the kth nearest neighbor point; S253. Arrange the k-distances of all data points in descending order and draw a curve, wherein the k-distance value corresponding to the first inflection point position in the k-distance graph is taken as the ε value.

5. A quality evaluation method for earthquake disaster equipment monitoring data according to claim 4, characterized in that: Step S26 specifically includes: S261, for each data point in the data set, calculate the number of points in its ε-neighborhood and compare them, wherein when the number of points in the ε-neighborhood of the data point exceeds the minPts value, the data point is marked as a core point, otherwise it is temporarily marked as a noise point; S262, for each core point, connect the core point with all core points in its ε-neighborhood to form a cluster, and mark all non-core points in its ε-neighborhood as boundary points and attribute them to the cluster; S263, repeating step S24, steps S251 to S253, and steps S261 to S262 until all core points are processed; S264: Mark the unattributed noise points and obtain the final clustering result.

6. A quality evaluation method for earthquake disaster equipment monitoring data according to claim 5, characterized in that: Step S3 specifically includes: S31, integrating earthquake monitoring data from different sources that have completed noise reduction processing and waveform abnormality data monitoring, and separating different attributes of the earthquake data, wherein the earthquake attributes of the earthquake monitoring data include waveform, frequency, amplitude, time, location, magnitude, focal depth, damage caused by the earthquake, and impact range; S32, using a multivariate linear regression attribute fusion method to perform feature fusion on different attributes of the earthquake monitoring data to obtain predicted values ​​of different attributes; S33, using cosine similarity to determine the similarity between each prediction value; S34. According to the K-means algorithm, each predicted value is divided into K clusters to obtain the final cluster division result.

7. A quality evaluation method for earthquake disaster equipment monitoring data according to claim 6, characterized in that: The specific steps of feature fusion in step S32 include: S321, determining a target attribute and a predicted attribute, wherein the target attribute is the severity of the earthquake event, and the predicted attribute is different attributes of the earthquake; S322, establish a multiple linear regression model, the formula of the multiple linear regression model is: y=β0+β1x1+β2x2+...+β k x k +e Among them, y is the target attribute, x1, x2, ..., x k are the predicted attributes, β0, β1, ..., β k is the model parameter, ε is the error term; S323, using the least squares method to estimate the value of the model parameters, the process is achieved by minimizing the residual sum of squares, and the process is mathematically expressed as: S324, using the estimated model parameters, substituting different attributes of each earthquake monitoring data into a multivariate linear regression model for prediction to obtain a predicted value; The specific steps of step S34 include: S341, randomly selecting K predicted values ​​as initial cluster centers; S342, for each predicted value, calculating its distance from each cluster center, and assigning it to the cluster where the nearest cluster center is located; S343, recalculating the cluster center of each cluster; S344, repeat B02 and B03 until the cluster center no longer changes or the maximum number of iterations is reached, and the final cluster division result is obtained.

8. A quality evaluation method for earthquake disaster equipment monitoring data according to claim 7, characterized in that: Step S4 specifically includes: S41. Using principal component analysis (PCA) to extract features from the seismic monitoring data after data fusion, a characteristic mother wave time history is obtained, and the error of the characteristic mother wave time history is adjusted based on amplitude, spectrum and duration; S42, locally improving the convolutional neural network, used to adapt to the input of waveform data in earthquake monitoring data and the picking of the first arrival time of earthquake P waves, wherein the specific steps of locally improving the convolutional neural network are: A01. Use one-dimensional convolutional layers instead of two-dimensional convolutional layers to adapt to the one-dimensionality of waveform data and maintain multi-channel input and output. A02. Use the bidirectional recurrent neural network layer BiRNN as the last layer of the convolutional neural network; A03. Input the seismic waveform data into the locally improved convolutional neural network to extract the first arrival time of the seismic P wave.

Citation Information

Cited By

  • Seismic activity analysis method and system based on Matlab software

    CN120294817A