Micro-seismic monitoring and early warning method based on data purification and space-time weighted Markov clustering

Abstract

The invention provides a micro-seismic monitoring and early warning method based on data purification and space-time weighted Markov clustering, and the method comprises the steps: carrying out the spatial boundary screening and statistical outlier elimination processing of original micro-seismic monitoring data, and obtaining an effective micro-seismic data set; key features are extracted from the effective micro-seismic data, a spatial component and a time component are separated, and statistics of the spatial component and the time component are calculated respectively; calculating a space-time weighted mahalanobis distance by combining space and time variability and correlation, and constructing a dissimilarity matrix between microseismic events based on the space-time weighted mahalanobis distance; carrying out clustering analysis by adopting a Ward hierarchical clustering algorithm based on the dissimilarity degree matrix, and dividing the microseismic events into different clusters; and mapping a clustering result back to an original data space, and generating and outputting various visual charts. According to the method, fine cleaning and high-precision spatial-temporal clustering of the micro-seismic data can be realized, and the data quality and rule recognition capability of mine micro-seismic monitoring can be remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of coal mine safety monitoring and intelligent geological data analysis technology, specifically a microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering. Background Technology

[0002] Microseismic monitoring is a crucial preventative measure for early warning of disasters such as rockbursts and rock bursts in underground engineering projects like coal mines and tunnels. During monitoring, the collected microseismic data often contains significant noise, invalid events outside the working face, and outliers caused by measurement errors. Furthermore, microseismic events exhibit spatial clustering and temporal evolution. Existing technologies, such as traditional Euclidean distance clustering methods (e.g., K-means), assume that features in each dimension are independent and homoscedastic, neglecting the correlation of microseismic data in the X, Y, and Z spatial directions (e.g., due to the influence of geological structure trends, the distribution range in one direction is much larger than in others). They also struggle to effectively handle the inconsistency between time units (days / seconds) and spatial units (meters), resulting in clustering results that fail to accurately reflect the spatiotemporal patterns of microseismic activity and affecting the reliability of early warning models. Therefore, there is an urgent need to provide a microseismic monitoring and early warning method based on data purification and spatiotemporally weighted Markov clustering. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides a microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering. This method can effectively solve the problems of inconsistent spatiotemporal dimensions and neglect of spatial correlation in traditional methods. It can achieve fine cleaning and high-precision spatiotemporal clustering of microseismic data, and can significantly improve the data quality and pattern recognition capability of mine microseismic monitoring.

[0004] To achieve the above objectives, this invention provides a microseismic monitoring and early warning method based on data cleansing and spatiotemporal weighted Markov clustering, comprising the following steps: S1: Data Cleansing; Spatial boundary screening and statistical outlier removal were performed on the original microseismic monitoring data to obtain an effective microseismic dataset; S2: Feature construction and parameter calculation; Key features are extracted from effective microseismic data, and spatial and temporal components are separated. Then, the statistics of spatial and temporal components are calculated separately. S3: Weighted Markov clustering; S31: Combine spatial and temporal variability and correlation to calculate the spatiotemporal weighted Mahalanobis distance, and construct the dissimilarity matrix between microseismic events based on the spatiotemporal weighted Mahalanobis distance; S32: Based on the dissimilarity matrix, the Ward hierarchical clustering algorithm is used for cluster analysis to divide the microseismic events into different clusters; S4: Visual output; The clustering results are mapped back to the original data space to generate and output various visualization charts.

[0005] Furthermore, in order to provide a high-quality data foundation for subsequent microseismic feature analysis and cluster identification, the process of spatial boundary screening and statistical outlier removal of the original microseismic monitoring data in S1 is as follows: S11: Read the original microseismic monitoring data from the microseismic monitoring system, construct a polygonal path for the working face boundary, and use the ray method or bounding box algorithm to determine whether the coordinates of the microseismic event are within the path, and remove external interference data; S12: For the selected microseismic events, calculate the mean values ​​of the X, Y, and Z axis coordinates respectively. with standard deviation And calculate the Z-score value of each data point according to formula (1). ; (1); In the formula, These are the observations of microseismic events on the X, Y, or Z axes; S13: Set threshold Remove any dimension Spatial outliers were identified to obtain an effective microseismic dataset.

[0006] Furthermore, in order to effectively remove extreme noise data while preserving as much far-field key information with geological early warning value as possible, and to achieve the optimal balance between data quality and information integrity, in S13 of S1, based on the long-tail distribution characteristics and the requirement for geological event fidelity, the threshold is set... The value is 5.0.

[0007] Furthermore, in order to provide statistically reliable basic parameters for subsequent spatiotemporal weighted distance calculations, the calculation process for the statistics of the spatial and temporal components in S2 is as follows: S21: Extracting spatial coordinate vectors from effective microseismic data and normalized time features Construct feature vectors ; S22: Define two microseismic events and The feature vectors are respectively and And the feature vectors are separated into spatial components. and time components ; S23: Calculate the covariance matrix of spatial dimension data And introduce a small regularization parameter. Calculate its inverse matrix ; S24: Calculate the variance of the time dimension data And calculate its reciprocal. .

[0008] Furthermore, in order to achieve scientific quantification of the multidimensional characteristics of microseismic events and more accurately reflect the spatiotemporal correlation characteristics of microseismic activity, in S31 of S3, the spatiotemporal weighted Mahalanobis distance is calculated according to formula (2). ; (2); In the formula, The square of the spatial distance. ; It is the square of the time distance. ; This is the time weighting coefficient, used to adjust the influence of the time dimension on the clustering results.

[0009] Furthermore, in order to effectively identify microseismic activity areas with independent physical significance and achieve optimal fusion of spatiotemporal features while avoiding chain effects, the Ward hierarchical clustering algorithm is used in S32 of S3 for cluster analysis. The process of dividing microseismic events into different clusters is as follows: S32-1: Calculate the distances between all initial clusters using the Ward minimum variance method; S32-2: For each pair of clusters and Calculate the sum of squared errors after merging according to formula (3). ; (3); Secondly , Clusters ,cluster Event data; For clusters and cluster The spacetime weighted Mahalanobis distance between the centroids; S32-3: Select from all cluster pairs Merge the two smallest clusters and update the cluster set; S32-4: Repeat S32-2 and S32-3 until the preset number of clusters is reached; S32-5: Record each merge operation and generate a clustering dendrogram; S32-6: Automatically cut the dendrogram according to the preset range of cluster numbers to obtain the final cluster labels.

[0010] Furthermore, in order to fully reveal the spatiotemporal distribution characteristics and clustering structure of microseismic data and significantly improve the intuitiveness of data analysis and the efficiency of interactive decision-making, S4 includes a variety of visualization charts, including monthly 2D planar distribution maps, 3D spatial distribution scatter plots, clustering statistics charts, and interactive 3D HTML charts.

[0011] This invention provides a microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering. Firstly, during data purification, spatial boundary filtering accurately defines the analysis area, initially eliminating irrelevant external interference signals. Simultaneously, boundaries can be defined based on actual geological structures, improving the focus of the analysis and reducing the processing of invalid data, thus enhancing the performance of subsequent algorithms. Secondly, statistical outlier removal effectively separates noise generated by equipment failure or human factors, significantly improving data quality and providing clean data based on real microseismic activity patterns for subsequent clustering analysis. Thirdly, in feature construction and parameter calculation, the separation of spatial and temporal components allows for independent examination of spatial focus and temporal series patterns. Furthermore, the calculation of spatial and temporal component statistics provides an independent statistical basis for subsequent spatiotemporal weighting. The spatial component statistics capture the statistical distribution characteristics of the spatial coordinates of microseismic events, helping to reveal the spatial correspondence between microseismic activity and fault structure, thus ensuring accurate fault identification and classification. The statistical measures of the time component provide a mathematical foundation for the temporal analysis of microseismic data, ensuring that subsequent clustering algorithms can accurately identify and distinguish microseismic activity patterns with different temporal characteristics. Subsequently, by combining spatial and temporal variability and correlation, the calculation of spatiotemporally weighted Mahalanobis distance can effectively eliminate the linear correlation between spatial coordinates and capture the intrinsic correlation between features through the spatial component statistics. At the same time, the statistical measures of the time component can eliminate the influence of differences in different feature units and magnitudes, and the adjustment of the spatiotemporal coupling degree can ensure that the subsequent clustering results are both spatially compact and temporally continuous. This allows the subsequent clustering analysis results to conform to the physical laws of the grouped occurrence of microseismic activity, significantly improving the data quality and pattern identification capability of mine microseismic monitoring. Building upon this foundation, the application of the Ward hierarchical clustering algorithm enables high-precision automated clustering of microseismic events. This ensures that the clustering results accurately reflect the spatiotemporal patterns of microseismic activity, effectively overcoming the common 'chain reaction' in microseismic data. Specifically, it avoids misclassifying loosely connected events belonging to different faults or different periods as the same category, thus more accurately delineating microseismic activity areas with independent physical significance. Simultaneously, it ensures high similarity among similar events and visually displays the clustering relationships of microseismic activity through dendrograms. Finally, through the output of various visualization charts, the abstract clustering results can be restored to the actual geographic space, intuitively revealing the temporal development patterns of microseismic activity and comprehensively showcasing the interactive influence of spatiotemporal characteristics.

[0012] This method, through an improved spatiotemporal weighted Mahalanobis distance algorithm combined with the Ward hierarchical clustering algorithm, effectively solves the problems of inconsistent spatiotemporal dimensions and neglect of spatial correlation in traditional methods. It achieves refined cleaning and high-precision spatiotemporal clustering of microseismic data, significantly improving the data quality and pattern recognition capabilities of mine microseismic monitoring. Attached Figure Description

[0013] Figure 1 This is a flowchart of the present invention; Figure 2 This is a comparison diagram of the spatial distribution of data before and after purification in an embodiment of the present invention; Figure 3 This is a comparison chart of clustering effects under different time weights in the embodiments of the present invention; Figure 4 This is a schematic diagram of the 3D interactive clustering results in an embodiment of the present invention. Detailed Implementation

[0014] The invention will now be further described with reference to the accompanying drawings.

[0015] like Figure 1 As shown, this invention provides a microseismic monitoring and early warning method based on data cleansing and spatiotemporal weighted Markov clustering, comprising the following steps: S1: Data Cleansing; Spatial boundary screening and statistical outlier removal were performed on the original microseismic monitoring data to obtain an effective microseismic dataset; To provide a high-quality data foundation for subsequent microseismic feature analysis and cluster identification, the process of spatial boundary screening and statistical outlier removal of the original microseismic monitoring data is as follows: S11: Read the original microseismic monitoring data from the microseismic monitoring system, construct a polygonal path for the working face boundary, and use the ray method or bounding box algorithm to determine whether the coordinates of the microseismic event are within the path, and remove external interference data; Ray method determines the relationship between points and polygons by calculating the parity of the intersection points of rays and boundaries; bounding box algorithm quickly eliminates obvious external points by utilizing coordinate range, thereby effectively removing interference data from surrounding rock fractures or sensor noise.

[0016] As a preferred approach, a polygonal path is constructed based on the coordinates of the four corners of the working face, and then the Path.contains_points algorithm is used to determine whether the microseismic event is located within the path. S12: For the selected microseismic events, calculate the mean values ​​of the X, Y, and Z axis coordinates respectively. with standard deviation And calculate the Z-score value of each data point according to formula (1). This process transforms the data into a standard normal distribution, eliminating the influence of dimensions and facilitating subsequent outlier detection. (1); In the formula, These are the observations of microseismic events on the X, Y, or Z axes; S13: Set threshold Remove any dimension By identifying and removing spatial outliers that are far from the normal data distribution, the quality of data for subsequent analysis is effectively ensured.

[0017] To effectively remove extreme noise data while preserving the most valuable far-field information for geological early warning, and to achieve the optimal balance between data quality and information integrity, a threshold was set based on long-tail distribution characteristics and the requirement for faithful geological event preservation. The value is 5.0. In specific operations, outliers with an absolute value greater than 5 are removed. Because microseismic monitoring data exhibits a non-standard normal long-tailed distribution in space, traditional... The criterion (i.e., a threshold of 3) easily leads to the loss of far-field events with early warning value. Meanwhile, key geological events such as high-level rock strata fracturing often occur in areas far from the cluster center, requiring the preservation of far-field information. Through extensive experimental verification, the threshold was adjusted... A value of 5 can effectively balance noise removal and information fidelity, filtering out extreme location drift data while retaining more than 99.9% of the effective microseismic geological information.

[0018] S2: Feature construction and parameter calculation; Key features are extracted from effective microseismic data, and spatial and temporal components are separated. Then, the statistics of spatial and temporal components are calculated separately. In order to provide statistically reliable basic parameters for subsequent spatiotemporal weighted distance calculations, the calculation process of the statistics for the spatial and temporal components is as follows: S21: Extracting spatial coordinate vectors from effective microseismic data and normalized time features Construct feature vectors ; S22: Define two microseismic events and The feature vectors are respectively and And the feature vectors are separated into spatial components. and time components ; S23: Calculate the covariance matrix of spatial dimension data To measure the distribution and relationships of the dataset, and to introduce a small regularization parameter. To ensure the matrix is ​​invertible, calculate its inverse matrix. This is used to eliminate the correlation of spatial dimensions; as a preferred option, The value is 10 -6 To avoid numerical instability; S24: Calculate the variance of the time dimension data This is used to measure volatility over time and to calculate its reciprocal. This is used to eliminate differences in spatiotemporal dimensions and provide a basis for distance calculation; S3: Weighted Markov clustering; S31: Combine spatial and temporal variability and correlation to calculate the spatiotemporal weighted Mahalanobis distance, so as to more accurately measure the dissimilarity between microseismic events, and construct the dissimilarity matrix between microseismic events based on the spatiotemporal weighted Mahalanobis distance; In order to achieve scientific quantification of the multidimensional characteristics of microseismic events and more accurately reflect the spatiotemporal correlation characteristics of microseismic activity, the spatiotemporal weighted Mahalanobis distance is calculated according to formula (2). ; (2); In the formula, The square of the spatial distance reflects the difference in spatial location between two microseismic events, and is weighted using the inverse of the covariance matrix. ; This is the square of the time distance, reflecting the difference in the time of occurrence of the two micro-earthquakes, and is weighted by the reciprocal of the time variance. ; The time weighting coefficient is used to adjust the influence of the time dimension on the clustering results, and is considered an optimal choice. ; S32: Based on the dissimilarity matrix, the Ward hierarchical clustering algorithm is used for cluster analysis to divide the microseismic events into different clusters; To effectively identify microseismic activity areas with independent physical significance and achieve optimal fusion of spatiotemporal features while avoiding chain effects, the Ward hierarchical clustering algorithm is used for cluster analysis. The process of dividing microseismic events into different clusters is as follows: S32-1: Calculate the distances between all initial clusters using the Ward minimum variance method; S32-2: For each pair of clusters and Calculate the sum of squared errors after merging according to formula (3). ; (3); Secondly , Clusters ,cluster Event data; For clusters and cluster The spacetime weighted Mahalanobis distance between the centroids; S32-3: Select from all cluster pairs The two smallest clusters are merged, and the cluster set is updated; this ensures that the generated microseismic clusters have the highest internal cohesion in the spatiotemporal feature space. S32-4: Repeat S32-2 and S32-3 until the preset number of clusters is reached; S32-5: Record each merge operation and generate a clustering dendrogram; S32-6: Automatically cut the dendrogram according to the preset range of cluster numbers to obtain the final cluster labels.

[0019] This method is a cohesive hierarchical clustering approach, the core of which lies in minimizing the variance within clusters. Compared to the Single Linkage method, the Ward method effectively overcomes the 'chain effect' commonly found in microseismic data, that is, it avoids misclassifying loosely connected events belonging to different faults or different periods as the same category, thereby more accurately delineating microseismic activity areas with independent physical significance.

[0020] S4: Visual output; The clustering results are mapped back to the original data space to generate and output a variety of visualization charts, enabling users to intuitively explore and analyze the spatiotemporal distribution and clustering characteristics of microseismic activity.

[0021] In order to fully reveal the spatiotemporal distribution characteristics and clustering structure of microseismic data and significantly improve the intuitiveness of data analysis and the efficiency of interactive decision-making, S4 provides a variety of visualization charts, including monthly 2D planar distribution maps, 3D spatial distribution scatter plots, clustering statistics charts, and interactive 3D HTML charts. Among them, the monthly 2D planar distribution map is used to display the planar distribution evolution trend of microseismic events each month in conjunction with the working face boundary; in the 3D spatial distribution scatter plot, different colors are used to distinguish clusters, and the microseismic energy is mapped by the size of the points to show the distribution of microseismic events in three-dimensional space; the clustering statistics chart is used to display the statistical information of the clustering results; the interactive 3D HTML chart is an interactive three-dimensional visualization interface that supports zooming, rotation, and switching between multiple perspectives such as front view and top view.

[0022] Example: Taking the microseismic monitoring data of a coal mine as an example, a dataset containing 3493 microseismic events was selected.

[0023] (1) Data purification: Set the boundary coordinates of the working surface and remove 428 data points outside the boundary. Calculate the Z-score value for the remaining data and remove the rest. There are 15 outliers with a value of 0; only valid data points are retained. For example... Figure 2 As shown, the purified microseismic data is more concentrated in the mining area.

[0024] (2) Distance Calculation: The spatial covariance matrix of the effective data was calculated, and it was found that the variance in the X direction was much greater than that in the Y direction, indicating significant anisotropy. If Euclidean distance is used, the clustering results will be stretched by the X-axis. The spatiotemporal weighted Mahalanobis distance in this invention effectively eliminates this correlation. At the same time, the time unit (day) and the spatial unit (meter) are normalized by variance, making them comparable in distance calculation.

[0025] (3) Cluster analysis: setting time weights Hierarchical clustering was performed, automatically dividing the data into 5 clusters. For example... Figure 3 As shown, compared with the traditional K-means method, the clustering results of this invention have better spatiotemporal continuity and can clearly identify the shift of microseismic activity centers as mining progresses.

[0026] (4) Visualization: The generated 3D interactive diagrams are as follows Figure 4 As shown, different colors represent different clusters, and the size of the dots represents the energy level, clearly demonstrating the spatial distribution of microseismic events around the working face.

[0027] This invention provides a microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering. Firstly, during data purification, spatial boundary filtering accurately defines the analysis area, initially eliminating irrelevant external interference signals. Simultaneously, boundaries can be defined based on actual geological structures, improving the focus of the analysis and reducing the processing of invalid data, thus enhancing the performance of subsequent algorithms. Secondly, statistical outlier removal effectively separates noise generated by equipment failure or human factors, significantly improving data quality and providing clean data based on real microseismic activity patterns for subsequent clustering analysis. Thirdly, in feature construction and parameter calculation, the separation of spatial and temporal components allows for independent examination of spatial focus and temporal series patterns. Furthermore, the calculation of spatial and temporal component statistics provides an independent statistical basis for subsequent spatiotemporal weighting. The spatial component statistics capture the statistical distribution characteristics of the spatial coordinates of microseismic events, helping to reveal the spatial correspondence between microseismic activity and fault structure, thus ensuring accurate fault identification and classification. The statistical measures of the time component provide a mathematical foundation for the temporal analysis of microseismic data, ensuring that subsequent clustering algorithms can accurately identify and distinguish microseismic activity patterns with different temporal characteristics. Subsequently, by combining spatial and temporal variability and correlation, the calculation of spatiotemporally weighted Mahalanobis distance can effectively eliminate the linear correlation between spatial coordinates and capture the intrinsic correlation between features through the spatial component statistics. At the same time, the statistical measures of the time component can eliminate the influence of differences in different feature units and magnitudes, and the adjustment of the spatiotemporal coupling degree can ensure that the subsequent clustering results are both spatially compact and temporally continuous. This allows the subsequent clustering analysis results to conform to the physical laws of the grouped occurrence of microseismic activity, significantly improving the data quality and pattern identification capability of mine microseismic monitoring. Building upon this foundation, the application of the Ward hierarchical clustering algorithm enables high-precision automated clustering of microseismic events. This ensures that the clustering results accurately reflect the spatiotemporal patterns of microseismic activity, effectively overcoming the common 'chain reaction' in microseismic data. Specifically, it avoids misclassifying loosely connected events belonging to different faults or different periods as the same category, thus more accurately delineating microseismic activity areas with independent physical significance. Simultaneously, it ensures high similarity among similar events and visually displays the clustering relationships of microseismic activity through dendrograms. Finally, through the output of various visualization charts, the abstract clustering results can be restored to the actual geographic space, intuitively revealing the temporal development patterns of microseismic activity and comprehensively showcasing the interactive influence of spatiotemporal characteristics.

[0028] This method, through an improved spatiotemporal weighted Mahalanobis distance algorithm combined with the Ward hierarchical clustering algorithm, effectively solves the problems of inconsistent spatiotemporal dimensions and neglect of spatial correlation in traditional methods. It achieves refined cleaning and high-precision spatiotemporal clustering of microseismic data, significantly improving the data quality and pattern recognition capabilities of mine microseismic monitoring.

Claims

1. A microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering, characterized in that, Includes the following steps: S1: Data Cleansing; Spatial boundary screening and statistical outlier removal were performed on the original microseismic monitoring data to obtain an effective microseismic dataset; S2: Feature construction and parameter calculation; Key features are extracted from effective microseismic data, and spatial and temporal components are separated. Then, the statistics of spatial and temporal components are calculated separately. S3: Weighted Markov clustering; S31: Combine spatial and temporal variability and correlation to calculate the spatiotemporal weighted Mahalanobis distance, and construct the dissimilarity matrix between microseismic events based on the spatiotemporal weighted Mahalanobis distance; S32: Based on the dissimilarity matrix, the Ward hierarchical clustering algorithm is used for cluster analysis to divide the microseismic events into different clusters; S4: Visual output; The clustering results are mapped back to the original data space to generate and output various visualization charts.

2. The microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering according to claim 1, characterized in that, In S1, the process of spatial boundary screening and statistical outlier removal of the raw microseismic monitoring data is as follows: S11: Read the original microseismic monitoring data from the microseismic monitoring system, construct a polygonal path for the working face boundary, and use the ray method or bounding box algorithm to determine whether the coordinates of the microseismic event are within the path, and remove external interference data; S12: For the selected microseismic events, calculate the mean values ​​of the X, Y, and Z axis coordinates respectively. with standard deviation And calculate the Z-score value of each data point according to formula (1). ; （1）； In the formula, These are the observations of microseismic events on the X, Y, or Z axes; S13: Set threshold Remove any dimension Spatial outliers were identified to obtain an effective microseismic dataset.

3. The microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering according to claim 1, characterized in that, In S13 of S1, based on the long-tail distribution characteristics and the requirement for fidelity in geological events, the threshold is... The value is 5.

0.

4. The microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering according to claim 1, characterized in that, In S2, the calculation process for the statistics of the spatial and temporal components is as follows: S21: Extracting spatial coordinate vectors from effective microseismic data and normalized time features Construct feature vectors ; S22: Define two microseismic events and The feature vectors are respectively and And the feature vectors are separated into spatial components. and time components ; S23: Calculate the covariance matrix of spatial dimension data And introduce a small regularization parameter. Calculate its inverse matrix ; S24: Calculate the variance of the time dimension data And calculate its reciprocal. .

5. The microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering according to claim 1, characterized in that, In S31 of S3, the spatiotemporal weighted Mahalanobis distance is calculated according to formula (2). ; （2）； In the formula, The square of the spatial distance. ; It is the square of the time distance. ; This is the time weighting coefficient, used to adjust the influence of the time dimension on the clustering results.

6. The microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering according to claim 1, characterized in that, In S32 of S3, the Ward hierarchical clustering algorithm is used for cluster analysis. The process of dividing the microseismic events into different clusters is as follows: S32-1: Calculate the distances between all initial clusters using the Ward minimum variance method; S32-2: For each pair of clusters and Calculate the sum of squared errors after merging according to formula (3). ; （3）； Secondly , Clusters ,cluster Event data; For clusters and cluster The spacetime weighted Mahalanobis distance between the centroids; S32-3: Select from all cluster pairs Merge the two smallest clusters and update the cluster set; S32-4: Repeat S32-2 and S32-3 until the preset number of clusters is reached; S32-5: Record each merge operation and generate a clustering dendrogram; S32-6: Automatically cut the dendrogram according to the preset range of cluster numbers to obtain the final cluster labels.

7. The microseismic monitoring and early warning method based on data purification and spatiotemporal weighted Markov clustering according to claim 1, characterized in that, In S4, a variety of visualization charts are available, including monthly 2D planar distribution charts, 3D spatial distribution scatter plots, cluster statistics charts, and interactive 3D HTML charts.

Images