Combined weighting microseismic source location method based on principal component adaptive search

Abstract

The combination weighting microseismic source positioning method based on principal component adaptive search first carries out principal component analysis on monitoring data to obtain a global search domain, then calculates in the range of the global search domain by solving an algorithm to obtain multiple microseismic source approximate solutions, and then obtains a standardized microseismic source approximate solution matrix; then the standardized microseismic source approximate solution matrix is analyzed by using an optimized clustering algorithm, and then multiple clusters and cluster related information are obtained, then the clusters are screened multiple times to retain effective clusters, and then the three-dimensional position coordinates of the microseismic source are obtained according to the weight and the median center of the effective clusters; the global search domain effectively reduces the invalid search space and improves the positioning accuracy; multiple screening removes abnormal solutions in the microseismic source approximate solution, and the positioning stability in the noise environment is significantly improved by weighting; therefore, the positioning accuracy of the microseismic source is high, and the positioning stability in the noise environment is good.

Description

Technical Field

[0001] This invention relates to a method for locating microseismic sources, belonging to the field of microseismic source location technology, and particularly to a combined weighted microseismic source location method based on principal component adaptive search. Background Technology

[0002] Microseismic monitoring is a key technology for rock mass stability assessment and disaster early warning in fields such as mining, hydropower engineering and oil and gas development. Its core lies in the high-precision positioning of microseismic sources, which directly determines the accuracy of monitoring and early warning and the effectiveness of engineering decisions.

[0003] Chinese patent application number 202511334490.3, filed on September 18, 2025, discloses a microseismic source location method and system based on an improved particle swarm optimization algorithm. First, n sensors are deployed in an open-pit mine. Then, the improved particle swarm optimization algorithm is used to solve for the minimum value of the objective function within the defined domain to determine the source coordinates. The improved particle swarm optimization algorithm introduces a hybrid chaotic mapping combining Logistic and Tent mappings in the initial population generation stage to generate a more regular and uniform initial population. After each iteration, the particle fitness is sorted, with populations with fitness below the average value called the dominant population and those with fitness above the average value called the suboptimal population. In the next iteration, an adaptive weighting strategy is used for the dominant population, and a Levy flight combined with an adaptive weighting strategy is used for the suboptimal population. Although this design can monitor the location of microseismic sources, it still has the following drawbacks:

[0004] Microseismic monitoring data contains a certain amount of noise interference. When faced with a multi-peaked and highly nonlinear microseismic source location objective function, the improved particle swarm optimization algorithm will output microseismic source coordinates with low accuracy due to noise interference. This design does not perform anomaly screening for microseismic source coordinates, so the design has low accuracy in locating microseismic sources.

[0005] The information disclosed in this background section is intended only to enhance understanding of the overall background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings and problems of low positioning accuracy of microseismic sources in the prior art, and to provide a combined weighted microseismic source positioning method based on principal component adaptive search with high positioning accuracy.

[0007] To achieve the above objectives, the technical solution of the present invention is:

[0008] A combined weighted microseismic source localization method based on principal component adaptive search, the method comprising the following steps:

[0009] Step 1: First, collect monitoring data, including the average wave velocity of the monitoring area, the three-dimensional position coordinates of multiple stations located within the monitoring area, and the observed values ​​at the time; then, perform principal component analysis on the monitoring data to obtain the global search domain;

[0010] The second step is to first input the monitoring data into the fitness function, then calculate within the global search domain using the solution algorithm to obtain multiple approximate solutions for microseismic sources, and then construct a standardized microseismic source approximate solution matrix based on the multiple microseismic source approximate solutions.

[0011] The third step involves clustering the standardized microseismic source approximation solution matrix using a clustering algorithm to obtain multiple clusters composed of microseismic source approximation solutions, along with the cluster center for each cluster. Then, basic cluster information is obtained for each cluster, followed by a comprehensive quality score for each cluster based on its basic information and cluster center. This comprehensive quality score and basic information are then used to screen each cluster for quality to retain preliminary effective clusters. Spatial anomaly screening is then applied to these preliminary effective clusters to obtain spatial anomaly clusters, followed by statistical anomaly screening to obtain statistical anomaly clusters. Finally, spatial anomaly clusters and statistical anomaly clusters are removed from the preliminary effective clusters to retain the remaining effective clusters.

[0012] Step 4: First, obtain the weights and median center based on the effective clusters, and then obtain the three-dimensional location coordinates of the microseismic source based on the weights and median center, thus completing the location of the microseismic source.

[0013] In the third step, obtaining the comprehensive quality score of each cluster based on the basic information of each cluster and the cluster center means: obtaining the point factor, compactness factor, density factor and shape factor based on the basic information of each cluster and the cluster center; then obtaining the comprehensive quality score of each cluster based on the point factor, compactness factor, density factor and shape factor; and finally, performing threshold screening on the clusters based on the comprehensive quality score and basic information of each cluster to retain the initially effective clusters.

[0014] In the third step, the spatial anomaly screening of the preliminary effective clusters to obtain spatial anomaly clusters refers to: first obtaining the Euclidean distance matrix based on the preliminary effective clusters, then obtaining the mean and standard deviation based on the Euclidean distance matrix, then obtaining the spatial critical value based on the mean and standard deviation, and finally screening the preliminary effective clusters based on the spatial critical value to obtain spatial anomaly clusters.

[0015] In the third step, the statistical anomaly screening of the preliminary effective clusters to obtain statistically abnormal clusters refers to: first obtaining a set of comprehensive quality scores based on the comprehensive quality scores of the clusters, then obtaining a statistical threshold based on the set of comprehensive quality scores, and finally screening the preliminary effective clusters based on the statistical threshold to obtain statistically abnormal clusters.

[0016] In the fourth step, obtaining the weights and median center based on the effective clusters, and then obtaining the three-dimensional location coordinates of the microseismic source based on the weights and median center, means: first, obtaining the normalized weight of each effective cluster based on the comprehensive mass score of the clusters; then, obtaining the mass score weighted center of the effective clusters based on the normalized weight of each cluster; then, obtaining the median center based on the effective clusters; then, obtaining the preliminary solution based on the median center and the mass score weighted center; and finally, restoring the coordinates of the preliminary solution to obtain the three-dimensional location coordinates of the microseismic source.

[0017] In the first step, the principal component analysis of the monitoring data to obtain the global search domain refers to: first, obtaining a decentralized array coordinate matrix based on the monitoring data; then, performing principal component analysis on the decentralized array coordinate matrix to obtain the principal component coordinate system; then, converting the three-dimensional position coordinate data of multiple stations in the monitoring data into the principal component coordinate system to obtain the principal component coordinates of multiple stations; then, determining the basic range based on the principal component coordinates of multiple stations; then, expanding the basic range to obtain the extended range; then, generating a cube based on the extended range, the cube including eight vertex coordinates; then, converting the eight vertex coordinates into eight one-to-one corresponding three-dimensional position coordinates; then, obtaining the minimum and maximum coordinate values ​​among the eight three-dimensional position coordinates; and finally, obtaining the global search domain based on the minimum and maximum coordinate values.

[0018] In the first step, performing principal component analysis on the decentralized array coordinate matrix and then obtaining the principal component coordinate system means: firstly, performing covariance matrix calculation and eigenvalue decomposition on the decentralized array coordinate matrix in sequence, and then obtaining the principal component coordinate system.

[0019] The principal component coordinate system includes a first principal component, a second principal component, and a third principal component. The first principal component, the second principal component, and the third principal component are mutually perpendicular to each other. The first principal component is the direction with the longest extension of the station distribution, the second principal component is the direction with the second longest extension of the station distribution, and the third principal component is the direction with the shortest extension of the station distribution.

[0020] In the first step, determining the basic range based on the principal component coordinates of multiple stations means: obtaining the three minimum and three maximum coordinate values ​​corresponding to the first, second, and third principal components based on the principal component coordinates of multiple stations, and then determining the basic range based on the three sets of minimum and maximum coordinate values.

[0021] The process of expanding the basic range to obtain the extended range refers to: first, obtaining the extended distance based on the maximum arrival time observation value, average wave velocity, and safety factor in the monitoring data; then, subtracting the extended distance from the three minimum coordinate values ​​to obtain the three minimum boundary values; and finally, increasing the extended distance from the three maximum coordinate values ​​to obtain the maximum boundary values. The range between the three maximum boundary values ​​and the three minimum boundary values ​​is the extended range.

[0022] In the third step, the clustering algorithm refers to the DBSCAN algorithm, which optimizes the neighborhood radius and minimum point threshold using the CTCM algorithm.

[0023] In the second step, the solution algorithm refers to a hybrid optimization algorithm that combines the Grey Wolf Optimization Algorithm and the Genetic Algorithm.

[0024] In the second step, the fitness function refers to: first, obtaining a time difference mathematical model based on a specified formula; then, obtaining an optimal model based on the time difference mathematical model; and finally, constructing a ternary quadratic nonlinear fitness function with four-by-four combinations and equal weights through arithmetic summation based on the optimal model. Compared with the prior art, the beneficial effects of this invention are:

[0025] 1. The present invention provides a combined weighted microseismic source localization method based on principal component adaptive search, comprising the following steps: Step 1: Perform principal component analysis on monitoring data to obtain a global search domain; Step 2: Substitute the monitoring data into a fitness function, and then calculate within the global search domain using a solution algorithm to obtain multiple approximate solutions for microseismic sources, and then construct a standardized matrix of approximate solutions for microseismic sources; Step 3: Perform cluster analysis on the standardized matrix of approximate solutions for microseismic sources to obtain multiple clusters composed of approximate solutions for microseismic sources and related information, and then perform quality screening, spatial screening, and statistical screening on the multiple clusters to retain effective clusters; Step 5: First, obtain the weights and median centers based on the effective clusters, and then fuse the weights and median centers to obtain the three-dimensional location coordinates of the microseismic source. The advantages of this invention also include:

[0026] Firstly, this invention can obtain the three-dimensional position coordinates of the microseismic source;

[0027] Secondly, in the third step, multiple clusters composed of approximate solutions from microseismic sources are subjected to quality screening, spatial screening, and statistical screening to remove abnormal clusters and retain effective clusters, thus screening the approximate solutions from microseismic sources; in the fourth step, the most stable and reliable three-dimensional position coordinates of the microseismic source are obtained through weighted averaging; the combination of these two steps significantly improves the positioning stability in noisy environments.

[0028] Thirdly: In the first and second steps, the solution algorithm obtains multiple approximate solutions for microseismic sources within the global search domain. The global search domain limits the search space of the solution algorithm, effectively reducing the invalid search space and avoiding the blind expansion of the search domain, thus improving the accuracy of the obtained approximate solutions for microseismic sources.

[0029] Therefore, the microseismic source of the present invention has high positioning accuracy and good positioning stability in noisy environments.

[0030] 2. In the combined weighted microseismic source localization method based on principal component adaptive search of the present invention, in the third step, the point number factor, compactness factor, density factor, and shape factor are obtained based on the basic information of each cluster. Then, the comprehensive quality score of each cluster is obtained based on the point number factor, compactness factor, density factor, and shape factor. The clusters are threshold-screened based on the comprehensive quality score and basic information of the clusters to retain preliminary effective clusters. The preliminary effective clusters are then screened to remove spatially and statistically anomalous clusters, and then effective clusters are obtained. Then, weighted calculation is performed based on the effective clusters, and then coordinate reconstruction is performed to obtain the three-dimensional position coordinates of the microseismic source. In application, anomalous clusters are removed through multiple screenings, that is, anomalous microseismic source approximate solutions are removed, and effective microseismic source approximate solutions are retained, thus improving the robustness of microseismic source localization. Furthermore, the weighted calculation is used to increase the weight of clusters with higher comprehensive quality scores, further improving the localization robustness. Therefore, the localization robustness of the present invention is good.

[0031] 3. In the combined weighted microseismic source location method based on principal component adaptive search of the present invention, in the first step, a principal component coordinate system is obtained first through principal component analysis based on monitoring data. Then, the three-dimensional position coordinates of multiple stations are transformed into the principal component coordinate system to obtain the basic range of the stations. Next, an extended range is obtained through travel time extension. Then, a cube is generated based on the extended range, and the coordinates of the eight vertices of the cube are converted into eight three-dimensional position coordinates. Finally, the minimum and maximum values ​​of the coordinates are obtained based on the three-dimensional position coordinates, thus obtaining the global search domain. In application, the spatial characteristics of the station distribution can be identified through the aforementioned steps. By aligning the shape and orientation of the global search domain with the station geometry, the amount of invalid search space is effectively reduced. Simultaneously, the directions of the three coordinate axes (i.e., principal component vectors) of the principal component coordinate system correspond sequentially to the strength of the constraint relationship between the station and the seismic source. The extended range obtained through travel-time expansion ensures that the search domain may include the actual seismic source location, while the orientation design based on station geometry avoids blind expansion of the search domain. Finally, this method does not rely on a specific coordinate system or positive / negative value restrictions, naturally supporting a global coordinate system that includes negative values, making it highly suitable for various scenarios such as surface monitoring, underground monitoring, and tunnel monitoring. Therefore, this invention does not restrict the location of the stations.

[0032] 4. In the combined weighted microseismic source localization method based on principal component adaptive search of the present invention, the clustering algorithm in the third step refers to the DBSCAN algorithm, which optimizes the neighborhood radius and minimum point threshold using the CTCM algorithm. In application, the traditional DBSCAN algorithm requires manual parameter tuning before clustering. However, the present invention uses the globally optimal values ​​automatically obtained by the CTCM algorithm as the neighborhood radius and minimum point threshold, reducing manual intervention and improving the adaptiveness of clustering. Therefore, the clustering results of the DBSCAN algorithm can better match the true distribution characteristics of the source approximation solution, that is, it can identify truly effective clusters in different application scenarios, thereby improving the accuracy of the subsequently output microseismic source location coordinates. Therefore, the present invention uses the DBSCAN algorithm without requiring manual parameter tuning.

[0033] 5. In the combined weighted microseismic source localization method based on principal component adaptive search of the present invention, the solution algorithm in the second step refers to a hybrid optimization algorithm combining the Grey Wolf Algorithm and the Genetic Algorithm. When applied, the hybrid optimization algorithm combines the advantages of both algorithms: the fast global search capability of the Grey Wolf Algorithm and the diversity preservation of the Genetic Algorithm. Furthermore, the mutation operation of the Genetic Algorithm is embedded in the iteration process of the Grey Wolf Algorithm, increasing population diversity and preventing getting trapped in local optima. Combined with the fitness evaluation of the Genetic Algorithm, the search range of the Grey Wolf Algorithm can be dynamically adjusted, avoiding blind expansion. This achieves a comprehensive improvement in convergence speed, solution accuracy, and robustness. Therefore, the present invention has good results in solving approximate solutions for microseismic sources.

[0034] 6. In the combined weighted microseismic source location method based on principal component adaptive search of the present invention, in the second step, the fitness function refers to: obtaining an optimized model based on the arrival time difference mathematical model, and then constructing a ternary quadratic nonlinear fitness function with four-by-four combinations and equal weights through arithmetic summation based on the optimized model. In application, the fitness function constructed through the arrival time difference model does not depend on the absolute arrival time of the first arrival wave, but uses the relative time difference of the first arrival wave arriving at different monitoring points for location. Therefore, it does not depend on the picking accuracy of the first arrival wave arrival time data, and can effectively cope with the location anomalies caused by errors in picking the first arrival wave of seismic waves, local inaccuracies in the velocity model, or the mixing of non-tectonic events (such as blasting, collapse). Therefore, the present invention does not depend on the picking accuracy of the first arrival wave arrival time data. Attached Figure Description

[0035] Figure 1 This is a flowchart of the present invention.

[0036] Figure 2 This is a schematic diagram of the quality screening in Example 2.

[0037] Figure 3 This is a schematic diagram of anomaly screening in Example 2.

[0038] Figure 4 This is a schematic diagram of Example 3.

[0039] Figure 5 This is a three-dimensional and two-dimensional schematic diagram of the three-dimensional location of the microseismic source obtained by the application of this invention.

[0040] Figure 6 yes Figure 5 A three-dimensional schematic diagram of the three-dimensional location of the source of a micro-earthquake.

[0041] Figure 7 yes Figure 5 A two-dimensional XY plane schematic diagram of the three-dimensional location of the source of a micro-earthquake.

[0042] Figure 8 yes Figure 5 A two-dimensional YZ plane schematic diagram of the three-dimensional location of the source of a micro-earthquake.

[0043] Figure 9 yes Figure 5 A two-dimensional ZX-plane schematic diagram of the three-dimensional location of the source of a micro-earthquake.

[0044] Figure 10 This is a schematic diagram of the clustering results applied using a three-dimensional coordinate system.

[0045] Figure 11 This is a schematic diagram of the clustering result structure applied by the present invention.

[0046] Figure 12 This is a schematic diagram comparing the three-dimensional location of the microseismic source obtained by applying the present invention with the actual location of the microseismic source. Detailed Implementation

[0047] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] Please see Figure 1 — Figure 12 A combined weighted microseismic source localization method based on principal component adaptive search, the method comprising the following steps:

[0049] Step 1: First, collect monitoring data, including the average wave velocity of the monitoring area, the three-dimensional position coordinates of multiple stations located within the monitoring area, and the observed values ​​at the time; then, perform principal component analysis on the monitoring data to obtain the global search domain;

[0050] The second step is to first input the monitoring data into the fitness function, then calculate within the global search domain using the solution algorithm to obtain multiple approximate solutions for microseismic sources, and then construct a standardized microseismic source approximate solution matrix based on the multiple microseismic source approximate solutions.

[0051] The third step involves clustering the standardized microseismic source approximation solution matrix using a clustering algorithm to obtain multiple clusters composed of microseismic source approximation solutions, along with the cluster center for each cluster. Then, basic cluster information is obtained for each cluster, followed by a comprehensive quality score for each cluster based on its basic information and cluster center. This comprehensive quality score and basic information are then used to screen each cluster for quality to retain preliminary effective clusters. Spatial anomaly screening is then applied to these preliminary effective clusters to obtain spatial anomaly clusters, followed by statistical anomaly screening to obtain statistical anomaly clusters. Finally, spatial anomaly clusters and statistical anomaly clusters are removed from the preliminary effective clusters to retain the remaining effective clusters.

[0052] Step 4: First, obtain the weights and median center based on the effective clusters, and then obtain the three-dimensional location coordinates of the microseismic source based on the weights and median center, thus completing the location of the microseismic source.

[0053] In the third step, obtaining the comprehensive quality score of each cluster based on the basic information of each cluster and the cluster center means: obtaining the point factor, compactness factor, density factor and shape factor based on the basic information of each cluster and the cluster center; then obtaining the comprehensive quality score of each cluster based on the point factor, compactness factor, density factor and shape factor; and finally, performing threshold screening on the clusters based on the comprehensive quality score and basic information of each cluster to retain the initially effective clusters.

[0054] In the third step, the spatial anomaly screening of the preliminary effective clusters to obtain spatial anomaly clusters refers to: first obtaining the Euclidean distance matrix based on the preliminary effective clusters, then obtaining the mean and standard deviation based on the Euclidean distance matrix, then obtaining the spatial critical value based on the mean and standard deviation, and finally screening the preliminary effective clusters based on the spatial critical value to obtain spatial anomaly clusters.

[0055] In the third step, the statistical anomaly screening of the preliminary effective clusters to obtain statistically abnormal clusters refers to: first obtaining a set of comprehensive quality scores based on the comprehensive quality scores of the clusters, then obtaining a statistical threshold based on the set of comprehensive quality scores, and finally screening the preliminary effective clusters based on the statistical threshold to obtain statistically abnormal clusters.

[0056] In the fourth step, obtaining the weights and median center based on the effective clusters, and then obtaining the three-dimensional location coordinates of the microseismic source based on the weights and median center, means: first, obtaining the normalized weight of each effective cluster based on the comprehensive mass score of the clusters; then, obtaining the mass score weighted center of the effective clusters based on the normalized weight of each cluster; then, obtaining the median center based on the effective clusters; then, obtaining the preliminary solution based on the median center and the mass score weighted center; and finally, restoring the coordinates of the preliminary solution to obtain the three-dimensional location coordinates of the microseismic source.

[0057] In the first step, the principal component analysis of the monitoring data to obtain the global search domain refers to: first, obtaining a decentralized array coordinate matrix based on the monitoring data; then, performing principal component analysis on the decentralized array coordinate matrix to obtain the principal component coordinate system; then, converting the three-dimensional position coordinate data of multiple stations in the monitoring data into the principal component coordinate system to obtain the principal component coordinates of multiple stations; then, determining the basic range based on the principal component coordinates of multiple stations; then, expanding the basic range to obtain the extended range; then, generating a cube based on the extended range, the cube including eight vertex coordinates; then, converting the eight vertex coordinates into eight one-to-one corresponding three-dimensional position coordinates; then, obtaining the minimum and maximum coordinate values ​​among the eight three-dimensional position coordinates; and finally, obtaining the global search domain based on the minimum and maximum coordinate values.

[0058] In the first step, performing principal component analysis on the decentralized array coordinate matrix and then obtaining the principal component coordinate system means: firstly, performing covariance matrix calculation and eigenvalue decomposition on the decentralized array coordinate matrix in sequence, and then obtaining the principal component coordinate system.

[0059] The principal component coordinate system includes a first principal component, a second principal component, and a third principal component. The first principal component, the second principal component, and the third principal component are mutually perpendicular to each other. The first principal component is the direction with the longest extension of the station distribution, the second principal component is the direction with the second longest extension of the station distribution, and the third principal component is the direction with the shortest extension of the station distribution.

[0060] In the first step, determining the basic range based on the principal component coordinates of multiple stations means: obtaining the three minimum and three maximum coordinate values ​​corresponding to the first, second, and third principal components based on the principal component coordinates of multiple stations, and then determining the basic range based on the three sets of minimum and maximum coordinate values.

[0061] The process of expanding the basic range to obtain the extended range refers to: first, obtaining the extended distance based on the maximum arrival time observation value, average wave velocity, and safety factor in the monitoring data; then, subtracting the extended distance from the three minimum coordinate values ​​to obtain the three minimum boundary values; and finally, increasing the extended distance from the three maximum coordinate values ​​to obtain the maximum boundary values. The range between the three maximum boundary values ​​and the three minimum boundary values ​​is the extended range.

[0062] In the third step, the clustering algorithm refers to the DBSCAN algorithm, which optimizes the neighborhood radius and minimum point threshold using the CTCM algorithm.

[0063] In the second step, the solution algorithm refers to a hybrid optimization algorithm that combines the Grey Wolf Optimization Algorithm and the Genetic Algorithm.

[0064] In the second step, the fitness function refers to: first, obtaining the time difference mathematical model according to the specified formula, then obtaining the optimal model according to the time difference mathematical model, and then constructing a ternary quadratic nonlinear fitness function with four-by-four combinations and equal weights through arithmetic summation based on the optimal model.

[0065] The following are supplementary descriptions of the present invention:

[0066] The arrival time observation value mentioned in this invention refers to the time data of the arrival of the first arrival wave at the station.

[0067] Example 1:

[0068] Please see Figure 1 — Figure 12 A combined weighted microseismic source localization method based on principal component adaptive search, the method comprising the following steps:

[0069] Step 1: First, collect monitoring data, including the average wave velocity of the monitoring area, the three-dimensional position coordinates of multiple stations located within the monitoring area, and the observed values ​​at the time; then, perform principal component analysis on the monitoring data to obtain the global search domain;

[0070] The second step is to first input the monitoring data into the fitness function, then calculate within the global search domain using the solution algorithm to obtain multiple approximate solutions for microseismic sources, and then construct a standardized microseismic source approximate solution matrix based on the multiple microseismic source approximate solutions.

[0071] The third step involves clustering the standardized microseismic source approximation solution matrix using a clustering algorithm to obtain multiple clusters composed of microseismic source approximation solutions, along with the cluster center for each cluster. Then, basic cluster information is obtained for each cluster, followed by a comprehensive quality score for each cluster based on its basic information and cluster center. This comprehensive quality score and basic information are then used to screen each cluster for quality to retain preliminary effective clusters. Spatial anomaly screening is then applied to these preliminary effective clusters to obtain spatial anomaly clusters, followed by statistical anomaly screening to obtain statistical anomaly clusters. Finally, spatial anomaly clusters and statistical anomaly clusters are removed from the preliminary effective clusters to retain the remaining effective clusters.

[0072] Step 4: First, obtain the weights and median center based on the effective clusters, and then obtain the three-dimensional location coordinates of the microseismic source based on the weights and median center, thus completing the location of the microseismic source.

[0073] Example 2:

[0074] The basic content is the same as in Example 1, except that:

[0075] Please see Figure 1 — Figure 3In the third step, obtaining the comprehensive quality score of each cluster based on its basic cluster information and cluster center means: obtaining the point factor, compactness factor, density factor, and shape factor based on the basic cluster information and cluster center; then obtaining the comprehensive quality score of each cluster based on the point factor, compactness factor, density factor, and shape factor; and finally, performing threshold screening on the clusters based on the comprehensive quality score and basic cluster information to retain initially effective clusters. In the third step, performing spatial anomaly screening on the initially effective clusters to obtain spatially anomalous clusters means: first obtaining the Euclidean distance matrix based on the initially effective clusters; then obtaining the mean and standard deviation based on the Euclidean distance matrix; then obtaining the spatial critical value based on the mean and standard deviation; and finally, screening the initially effective clusters based on the spatial critical value to obtain spatially anomalous clusters. In the third step, performing statistical anomaly screening on the initially effective clusters to obtain statistically anomalous clusters means: first obtaining a set of comprehensive quality scores based on the clusters' comprehensive quality scores; then obtaining a statistical threshold based on the set of comprehensive quality scores; and finally, screening the initially effective clusters based on the statistical threshold to obtain statistically anomalous clusters. In the fourth step, obtaining the weights and median centers based on the effective clusters, and then obtaining the three-dimensional location coordinates of the microseismic source based on the weights and median centers, means: first, obtaining the normalized weights of each effective cluster based on the comprehensive quality score of the clusters; then, obtaining the weighted center of the quality score of the effective clusters based on the normalized weights of each cluster; then, obtaining the median center based on the effective clusters; then, obtaining a preliminary solution based on the median center and the weighted center of the quality score; and finally, restoring the coordinates of the preliminary solution to obtain the three-dimensional location coordinates of the microseismic source. Preferably, in the second step, obtaining the standardized microseismic source approximate solution matrix based on multiple microseismic source approximate solutions with different distributions means: first, obtaining the microseismic source approximate solution matrix based on multiple microseismic source approximate solutions with different distributions; then, performing Z-score standardization on the microseismic source approximate solution matrix to obtain the standardized microseismic source approximate solution matrix.

[0076] In application, in the fourth step, the standardized microseismic source approximate solution matrix is ​​first clustered using the optimized clustering algorithm to obtain the cluster label vector. and Cluster centers The central matrix formed Clustering label vector It includes multiple clusters composed of approximate solutions from microseismic sources, with a central matrix. This includes multiple cluster centers corresponding to each cluster; then, based on these multiple clusters, basic cluster information corresponding to each cluster is obtained, i.e., the number of points within each cluster. Then calculate the point factor:

[0077] ;

[0078] Then calculate the average distance from each point within the cluster to the cluster center. Its opposite number is defined as compactness:

[0079] ;

[0080] In the formula, The value is a random minimum to prevent division by zero in the formula;

[0081] Then calculate the compactness factor:

[0082] ;

[0083] Then calculate the average distance between points within the cluster. Then calculate the density index:

[0084] ;

[0085] At the same time, if the number of points ≥3, use principal component analysis to calculate the shape index, that is, calculate the eigenvalues ​​of the covariance matrix ( Then calculate the elongation:

[0086] ;

[0087] In the formula, The value is a random minimum to prevent division by zero in the formula;

[0088] If the number of points If the elongation is less than 3, then the elongation is:

[0089] ;

[0090] Then calculate the shape factor:

[0091] ;

[0092] Then, based on the point factor, compactness factor, density factor, and shape factor, the overall quality score of each cluster is obtained:

[0093] ;

[0094] Then, all clusters are subjected to quality screening to ensure they meet the requirements. and This constitutes the initial effective clusters; then, in the initial effective cluster set... In the above, calculate the Euclidean distance matrix between any two clusters. Then for all Euclidean distance matrices Calculate the average value and standard deviation Then based on the average value and standard deviation The spatial critical value is obtained:

[0095] ;

[0096] Then filter out all spatial anomaly cluster sets , of which A spatial anomaly cluster It must satisfy the set of all inter-cluster distances of this cluster (based on the Euclidean distance matrix). The minimum value obtained is greater than the spatial critical value, that is:

[0097] ;

[0098] Then, using the comprehensive quality score set of Value and Value filtering to extract statistically abnormal clusters , among which, the Statistical anomaly clusters The cluster must meet the overall quality score. Less than the statistical threshold, that is:

[0099] ;

[0100] In the formula, For the set of comprehensive quality scores The 25th percentile, For the set of comprehensive quality scores Interquartile range;

[0101] Then from the initial effective cluster set Remove spatial anomaly clusters in the middle Statistical anomaly cluster set To obtain an effective cluster set ,Right now:

[0102] ;

[0103] Then, for the effective cluster set The normalized weights of the effective clusters within the cluster are calculated sequentially, i.e.:

[0104] ;

[0105] Effective cluster set The weighted center for quality score is:

[0106] ;

[0107] Then, in the effective cluster set Calculate the median center of the cluster centers (i.e., the cluster centers):

[0108] ;

[0109] Reintegration Quality Score Weighted Center With median center After restoring the coordinates, the final solution will be obtained. :

[0110] ;

[0111] ;

[0112] In the formula, and The value is determined based on empirical values, and can be 0.7 and 0.3, 0.5 and 0.5, etc.

[0113] The coordinate restoration is the inverse operation of the Z-score normalization process in step three.

[0114] Example 3:

[0115] The basic content is the same as in Example 1, except that:

[0116] Please see Figure 1 — Figure 12In the first step, the principal component analysis of the monitoring data to obtain the global search domain refers to: first, obtaining a decentralized array coordinate matrix based on the monitoring data; then, performing principal component analysis on the decentralized array coordinate matrix to obtain the principal component coordinate system; then, converting the three-dimensional position coordinate data of multiple stations in the monitoring data into the principal component coordinate system to obtain the principal component coordinates of multiple stations; then, determining the basic range based on the principal component coordinates of multiple stations; then, expanding the basic range to obtain the extended range; then, generating a cube based on the extended range, the cube including eight vertex coordinates; then, converting the eight vertex coordinates into eight one-to-one corresponding three-dimensional position coordinates; then, obtaining the minimum and maximum coordinate values ​​among the eight three-dimensional position coordinates; and finally, obtaining the global search domain based on the minimum and maximum coordinate values. In the first step, performing principal component analysis on the decentralized array coordinate matrix to obtain the principal component coordinate system means: firstly, calculating the covariance matrix and decomposing the eigenvalues ​​of the decentralized array coordinate matrix sequentially, and then obtaining the principal component coordinate system; the principal component coordinate system includes a first principal component, a second principal component, and a third principal component, which are mutually perpendicular to each other; the first principal component is the direction with the longest extension of the station distribution, the second principal component is the direction with the second longest extension of the station distribution, and the third principal component is the direction with the shortest extension of the station distribution. In the first step, determining the basic range based on the principal component coordinates of multiple stations means: obtaining the three minimum and three maximum coordinate values ​​corresponding to the first, second, and third principal components based on the principal component coordinates of multiple stations, and then determining the basic range based on the three sets of minimum and maximum coordinate values; expanding the basic range to obtain the extended range means: first obtaining the extended distance based on the maximum arrival time observation value, average wave velocity, and safety factor in the monitoring data, then subtracting the extended distance from the three minimum coordinate values ​​to obtain the three boundary minimum values, and then increasing the extended distance from the three maximum coordinate values ​​to obtain the boundary maximum values. The range between the three boundary maximum values ​​and the three boundary minimum values ​​is the extended range.

[0117] When applying the application, first collect all stations (the number of which is...). 3D position coordinate data Compared with the observed value at that time ,in, (Since stations may be located on the surface, underground, or in tunnels, therefore...) (including negative values), and the average wave velocity of the monitored area. Then, based on the three-dimensional position coordinate data of the station... Calculate the geometric center point of all stations. Then, the center point is subtracted from the coordinates of each station to achieve data decentralization, resulting in a decentralized station coordinate matrix. This is to ensure that subsequent principal component analysis can accurately identify the main directions of station distribution; then, the decentralized station coordinate matrix is... Principal component analysis is performed, which involves calculating the variance matrix sequentially. Eigenvalue decomposition (in It is the eigenvector matrix. (It is an eigenvalue diagonal matrix), and then the principal components are obtained. First principal component The direction representing the longest extension of the station distribution is where the constraint of the stations on the microseismic source is weakest. When the microseismic source moves along this direction, the travel time change is not sensitive. (Second principal component) The direction representing the secondary extension of the station distribution indicates that the stations exert moderate constraint on the microseismic sources, and the third principal component... The direction representing the shortest extension of the station distribution is where the station exerts the strongest constraint on the microseismic source. Even a slight movement of the microseismic source will cause significant travel time changes. The direction of the principal component coordinate system is the first principal component. Second principal component With the third principal component Then, the three-dimensional location coordinate data of all stations. Transform sequentially to the principal component coordinate system, i.e. , After matrix transpose, the characteristics of the station distribution become more apparent, i.e., the principal component coordinates of the stations follow the first principal component. The coordinate values ​​change most along the direction of the third principal component. The coordinate values ​​of the direction change the least; then, based on the principal component coordinates of all stations, the first principal component is found. Second principal component With the third principal component Find the maximum and minimum values ​​on the range, and then obtain the boundary of the basic range. , , The area within this boundary is the basic range; then, the boundary of the basic range is extended by travel time constraints, that is, the estimated maximum distance of the microseismic source is calculated based on the travel time observations. In the formula, The maximum value among the observed values ​​at that time. This is a safety factor (typically ranging from 1.5 to 2.0). The average wave speed is then determined based on the estimated maximum distance. Extend the boundary of the base range to obtain the boundary coordinates of the extended range:

[0118] ;

[0119] Then, a cube is generated based on the boundary coordinates of the extended range. The vertices of the cube are as follows:

[0120] ;

[0121] This cube represents the search range in the principal component space; then, the eight vertices of the cube are converted into a three-dimensional coordinate system, i.e. Then, the search range in the three-dimensional coordinate system is obtained. The shape and direction of this search range are aligned with the main features of the array distribution. Finally, the minimum values ​​in the x, y, and z axes are found within the search range in the three-dimensional coordinate system. , and the maximum value minimum value The lower bound and the maximum bound of the global search domain. The global search domain is the upper limit of the search. The shape and orientation of the global search domain depend on the coordinates of the stations. Therefore, the global search domain is an oriented cuboid that matches the distribution characteristics of the stations. The distribution characteristics of the stations are related to the potential location of the microseismic sources. By aligning the global search domain with the distribution characteristics of the station array (and also with the sampling characteristics of the station array), the invalid search space can be reduced (i.e., adaptive search), and the subsequent search efficiency can be improved.

[0122] Example 4:

[0123] The basic content is the same as in Example 1, except that:

[0124] Please see Figure 1 — Figure 4 In the third step, the clustering algorithm refers to the DBSCAN algorithm, which optimizes the neighborhood radius and minimum point threshold using the CTCM algorithm.

[0125] When applying the algorithm, first initialize the CTCM algorithm's population size (number of teams and number of members in each team), maximum number of iterations, search space range, and other parameters, then use the DBSCAN algorithm's proprietary clustering performance values. As a fitness function, the DBSCAN algorithm automatically optimizes two key parameters: neighborhood radius (epsilon) and minimum point threshold (MinPts). After iteration, the globally optimal values ​​are selected as the neighborhood radius and minimum point threshold for the DBSCAN algorithm. By simulating the process of individual competition and cooperation in a team, the CTCM algorithm intelligently searches for the optimal parameter combination, solving the problem of manual parameter tuning required before clustering in the DBSCAN algorithm. When initializing the population size for the CTCM algorithm, the parameters can be determined based on the general principles of metaheuristic optimization algorithms, i.e., empirical values. For example, the population size can be set to 8-12, the maximum number of iterations to 50-100, and the search range for the minimum point threshold (MinPts) to be... The value is [3, 15], the search range of the neighborhood radius (epsilon) is [0.1, 2.0], the cooperation probability is 0.3-0.4, and the competition probability is 0.2-0.3. If the clustering results of the optimized DBSCAN algorithm are still not ideal (such as all noise or too many clusters), the search range of the neighborhood radius (epsilon) and the minimum number of points threshold (MinPts) should be adjusted first based on experience. This is the key to determining the solution space of the microseismic source. If the iterative optimization process is too slow, you can try to appropriately reduce the population size or the maximum number of iterations and observe whether the convergence curve has stabilized in advance. If the CTCM algorithm seems to stagnate (prematurely convergent), you can try to increase the competition probability or cooperation probability to increase the population diversity.

[0126] Example 5:

[0127] The basic content is the same as in Example 1, except that:

[0128] Please see Figure 1 — Figure 4 In the second step, the solution algorithm refers to a hybrid optimization algorithm that combines the gray wolf optimization algorithm and the genetic algorithm.

[0129] In application, genetic algorithms maintain population diversity through crossover and mutation operations, effectively exploring the solution space of microseismic sources, thus enhancing global search capabilities. In contrast, the gray wolf optimization algorithm is prone to getting trapped in local optima in the later stages of iteration. The hybrid optimization algorithm utilizes the mutation mechanism of the genetic algorithm to randomly perturb the positions of individuals, increasing diversity and thus enhancing the ability to escape local optima and avoiding premature convergence. The crossover operation of the genetic algorithm can efficiently propagate high-quality genes and accelerate global convergence. The hybrid algorithm combines the rapid focusing characteristics of the gray wolf optimization algorithm with the global search capability of the genetic algorithm, quickly narrowing the search range in the early stages and refining the search through the genetic algorithm mechanism in the later stages, balancing global exploration and local development. The gray wolf optimization algorithm has a simple structure, few parameters, and is easy to implement, but its adaptability to complex multimodal functions is limited. The genetic algorithm is suitable for handling nonlinear and multimodal problems through probabilistic optimization mechanisms. The hybrid optimization algorithm introduces the mutation and crossover of the genetic algorithm into the framework of the gray wolf optimization algorithm, improving the algorithm's adaptability to complex functions.

[0130] Example 6:

[0131] The basic content is the same as in Example 1, except that:

[0132] Please see Figure 1 — Figure 4 In the second step, the fitness function refers to: first, obtaining the time difference mathematical model according to the specified formula, then obtaining the optimal model according to the time difference mathematical model, and then constructing a ternary quadratic nonlinear fitness function with four-by-four combinations and equal weights through arithmetic summation based on the optimal model.

[0133] In application, the specified formula refers to the seismic wave propagation theory formula, and the content of the formula relevant to this invention is as follows:

[0134] ;

[0135] In the formula, For the first individual stations To microseismic source The absolute value of the spatial distance, where, , The number of stations that received valid seismic wave signals; This refers to the velocity of seismic waves; For the first The station and the first The difference in the time-arrival observation data recorded by each station; For the first The location of the microseismic source from each station and the first The difference between the location of each station and the microseismic source; This represents the actual time of occurrence of the microseismic event;

[0136] Then we obtain the mathematical model of time difference:

[0137] ;

[0138] Then we obtain the optimal model:

[0139] ;

[0140] In the formula, The number of non-repeating combinations of any two stations;

[0141] The closer the minimum value of the optimization model is to 0, the closer the approximate solution of the microseismic source is to the actual source coordinates.

[0142] The above description is only a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. Any equivalent modifications or changes made by those skilled in the art based on the content disclosed in the present invention should be included within the scope of protection set forth in the claims.

Claims

1. A combined weighted microseismic source localization method based on principal component adaptive search, characterized in that: The method includes the following steps: Step 1: First, collect monitoring data, including the average wave velocity of the monitoring area, the three-dimensional position coordinates of multiple stations located within the monitoring area, and the observed values ​​at the time; then, perform principal component analysis on the monitoring data to obtain the global search domain; The second step is to first input the monitoring data into the fitness function, then calculate within the global search domain using the solution algorithm to obtain multiple approximate solutions for microseismic sources, and then construct a standardized microseismic source approximate solution matrix based on the multiple microseismic source approximate solutions. The third step involves clustering the standardized microseismic source approximation solution matrix using a clustering algorithm to obtain multiple clusters composed of microseismic source approximation solutions, along with the cluster center for each cluster. Then, basic cluster information is obtained for each cluster, followed by a comprehensive quality score for each cluster based on its basic information and cluster center. This comprehensive quality score and basic information are then used to screen each cluster for quality to retain preliminary effective clusters. Spatial anomaly screening is then applied to these preliminary effective clusters to obtain spatial anomaly clusters, followed by statistical anomaly screening to obtain statistical anomaly clusters. Finally, spatial anomaly clusters and statistical anomaly clusters are removed from the preliminary effective clusters to retain the remaining effective clusters. Step 4: First, obtain the weights and median center based on the effective clusters, and then obtain the three-dimensional location coordinates of the microseismic source based on the weights and median center, thus completing the location of the microseismic source; In the third step, obtaining the comprehensive quality score of each cluster based on the basic information of each cluster and the cluster center means: obtaining the point factor, compactness factor, density factor and shape factor based on the basic information of each cluster and the cluster center; then obtaining the comprehensive quality score of each cluster based on the point factor, compactness factor, density factor and shape factor; and finally, performing threshold screening on the clusters based on the comprehensive quality score and basic information of each cluster to retain the initially effective clusters.

2. The combined weighted microseismic source localization method based on principal component adaptive search according to claim 1, characterized in that: In the third step, the spatial anomaly screening of the preliminary effective clusters to obtain spatial anomaly clusters refers to: first obtaining the Euclidean distance matrix based on the preliminary effective clusters, then obtaining the mean and standard deviation based on the Euclidean distance matrix, then obtaining the spatial critical value based on the mean and standard deviation, and finally screening the preliminary effective clusters based on the spatial critical value to obtain spatial anomaly clusters. In the third step, the statistical anomaly screening of the preliminary effective clusters to obtain statistically abnormal clusters refers to: first obtaining a set of comprehensive quality scores based on the comprehensive quality scores of the clusters, then obtaining a statistical threshold based on the set of comprehensive quality scores, and finally screening the preliminary effective clusters based on the statistical threshold to obtain statistically abnormal clusters.

3. The combined weighted microseismic source localization method based on principal component adaptive search according to claim 2, characterized in that: In the fourth step, obtaining the weights and median center based on the effective clusters, and then obtaining the three-dimensional location coordinates of the microseismic source based on the weights and median center, means: first, obtaining the normalized weight of each effective cluster based on the comprehensive mass score of the clusters; then, obtaining the mass score weighted center of the effective clusters based on the normalized weight of each cluster; then, obtaining the median center based on the effective clusters; then, obtaining the preliminary solution based on the median center and the mass score weighted center; and finally, restoring the coordinates of the preliminary solution to obtain the three-dimensional location coordinates of the microseismic source.

4. The combined weighted microseismic source localization method based on principal component adaptive search according to claim 1, characterized in that: In the first step, the principal component analysis of the monitoring data to obtain the global search domain refers to: first, obtaining a decentralized array coordinate matrix based on the monitoring data; then, performing principal component analysis on the decentralized array coordinate matrix to obtain the principal component coordinate system; then, converting the three-dimensional position coordinate data of multiple stations in the monitoring data into the principal component coordinate system to obtain the principal component coordinates of multiple stations; then, determining the basic range based on the principal component coordinates of multiple stations; then, expanding the basic range to obtain the extended range; then, generating a cube based on the extended range, the cube including eight vertex coordinates; then, converting the eight vertex coordinates into eight one-to-one corresponding three-dimensional position coordinates; then, obtaining the minimum and maximum coordinate values ​​among the eight three-dimensional position coordinates; and finally, obtaining the global search domain based on the minimum and maximum coordinate values.

5. The combined weighted microseismic source localization method based on principal component adaptive search according to claim 4, characterized in that: In the first step, performing principal component analysis on the decentralized array coordinate matrix and then obtaining the principal component coordinate system means: firstly, performing covariance matrix calculation and eigenvalue decomposition on the decentralized array coordinate matrix in sequence, and then obtaining the principal component coordinate system. The principal component coordinate system includes a first principal component, a second principal component, and a third principal component. The first principal component, the second principal component, and the third principal component are mutually perpendicular to each other. The first principal component is the direction with the longest extension of the station distribution, the second principal component is the direction with the second longest extension of the station distribution, and the third principal component is the direction with the shortest extension of the station distribution.

6. The combined weighted microseismic source localization method based on principal component adaptive search according to claim 5, characterized in that: In the first step, determining the basic range based on the principal component coordinates of multiple stations means: obtaining the three minimum and three maximum coordinate values ​​corresponding to the first, second, and third principal components based on the principal component coordinates of multiple stations, and then determining the basic range based on the three sets of minimum and maximum coordinate values. The process of expanding the basic range to obtain the extended range refers to: first, obtaining the extended distance based on the maximum arrival time observation value, average wave velocity, and safety factor in the monitoring data; then, subtracting the extended distance from the three minimum coordinate values ​​to obtain the three minimum boundary values; and finally, increasing the extended distance from the three maximum coordinate values ​​to obtain the maximum boundary values. The range between the three maximum boundary values ​​and the three minimum boundary values ​​is the extended range.

7. The combined weighted microseismic source localization method based on principal component adaptive search according to claim 1, characterized in that: In the third step, the clustering algorithm refers to the DBSCAN algorithm, which optimizes the neighborhood radius and minimum point threshold using the CTCM algorithm.

8. The combined weighted microseismic source localization method based on principal component adaptive search according to claim 1, characterized in that: In the second step, the solution algorithm refers to a hybrid optimization algorithm that combines the Grey Wolf Optimization Algorithm and the Genetic Algorithm.

9. The combined weighted microseismic source localization method based on principal component adaptive search according to claim 1, characterized in that: In the second step, the fitness function refers to: first, obtaining the time difference mathematical model according to the specified formula, then obtaining the optimal model according to the time difference mathematical model, and then constructing a ternary quadratic nonlinear fitness function with four-by-four combinations and equal weights through arithmetic summation based on the optimal model.

Images