A crack face reconstruction method and device based on a dynamic graph neural network

Abstract

The application discloses a crack surface reconstruction method and device based on a dynamic graph neural network, and the reconstruction method comprises the following steps: acquiring a coal rock rupture source core parameter; adopting a standardization processing method to pre-process the rupture source core parameter, constructing a crack space-time topology analysis model based on a DGCNN graph neural network, and taking the crack as a graph node; utilizing an edge convolution mechanism, constructing an adjacency graph through the space-time relationship between nodes, aggregating neighborhood features to update the node representation, capturing local structure and global topology; calculating crack feature similarity and clustering according to a density threshold by using a DBSCAN algorithm; and combining crack magnitude to optimize the clustering result, identifying the main crack surface crack with high expansion potential; adopting a B-spline surface fitting algorithm to reconstruct the macroscopic crack surface; selecting a control point to determine the crack main direction through principal component analysis, constructing a B-spline curve to generate a boundary contour based on the crack point distribution, and accurately fitting the crack surface by using a B-spline algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fracture reconstruction, and particularly relates to a fracture surface reconstruction method and device based on a dynamic graph neural network. BACKGROUND

[0002] With the increase of the depth of coal resource exploitation, the micro-cracks in rock mass gradually expand and form through macro-fracture surfaces, which seriously threaten the stability and safety of mine engineering. The formation of through fractures is not only a typical multi-scale dynamic evolution phenomenon of rock mass instability and failure, but also significantly increases the risk of disaster failure, causing economic losses and safety hazards. Therefore, fracture surface reconstruction is of great significance to reveal the instability evolution mechanism of rock mass and improve the stability of deep mining. In addition, the accurate reconstruction of fracture network is also crucial in resource development and environmental management technologies such as hydraulic fracturing and carbon dioxide mineralization storage. Through fracture surface reconstruction, the spatial distribution, geometric morphology and through characteristics of the cracks can be accurately reproduced, providing a scientific basis for evaluating the effect of artificial fracture system and optimizing reservoir reconstruction design. Therefore, exploring the fracture surface reconstruction method and its application in the dynamic evolution of through cracks has important value for disaster prevention and control and energy technology development.

[0003] The formation and penetration of fracture surfaces are complex "black box" evolution processes, which are difficult to reveal the mechanism by traditional methods. Microseismic monitoring, as a non-destructive detection technology, provides an effective means to open this "black box" by capturing the elastic wave signals associated with the loading of coal and rock. Through microseismic inversion, not only can the spatial position of the fracture source be accurately located, but also the spatio-temporal dynamic process of crack propagation and the fracture mechanism can be quantitatively deduced, which provides scientific support for studying the dynamic evolution law of internal cracks in rock mass, the penetration characteristics and the mechanism of induced disasters. Patent CN114966849A provides a rock fracture characterization method based on microseismic or acoustic emission and focal mechanism constraints, which realizes fracture characterization by clustering and fitting fracture surfaces based on microseismic point cloud data. Patent CN115077437A discloses a rock hydraulic fracturing fracture morphology characterization method based on acoustic emission positioning constraints, which generates point clouds by acoustic emission positioning, segments and optimizes facets to determine the orientation and geometric parameters of the fractures. Patent CN118884524A proposes a fracture network model construction and seepage channel identification method based on microseismic monitoring, which uses microseismic moment tensors and focal parameters to construct a seepage model and identify the main channel with the minimum flow time. Although these fracture characterization methods can reveal the spatial distribution characteristics of fractures to some extent and realize three-dimensional description from the perspective of point cloud, they mostly rely on distance or energy clustering of point cloud, and fail to fully consider the complex spatio-temporal topological relationship between fracture points. Moreover, they lack comprehensive analysis of the multi-dimensional characteristics of fractures, which limits the in-depth understanding and characterization of the global characteristics of fracture network. Under this background, deep learning methods, especially graph neural network technology, become the key to breaking through traditional methods. Graph neural networks have the advantages of capturing the spatio-temporal relationship between nodes, modeling the physical connectivity between nodes, and high flexibility and efficiency in handling complex topological relationships, which enable them to efficiently process complex fracture point cloud data and provide a new approach and tool for fracture surface reconstruction. Therefore, a method for intelligent reconstruction of fracture surfaces using a graph neural network model is proposed to realize automatic inversion and reconstruction of fracture surfaces in the time and space dimensions, fully describe the complex spatio-temporal topological relationship between fracture points and the penetration mechanism, and provide strong support for deep understanding of the multi-scale dynamic evolution process of rock mass instability and fracture. SUMMARY

[0004] The present scheme proposes a fracture surface reconstruction method and device based on a dynamic graph neural network to address the problems and needs mentioned above. The technical features adopted enable the achievement of the above technical purposes and other technical effects.

[0005] One object of the present application is to propose a fracture surface reconstruction method based on a dynamic graph neural network, comprising the following steps:

[0006] S10: Obtain original acoustic wave data through a microseismic monitoring system, and perform positioning calculation and focal mechanism inversion on the original acoustic wave data to obtain coal rock rupture source core parameters directly related to rupture; wherein the coal rock rupture source core parameters include: crack spatial coordinates, crack volume, crack energy, crack spatial orientation and crack magnitude;

[0007] S20: Preprocess the rupture source core parameters by using a standardization processing method, construct a crack spatio-temporal topology analysis model based on a DGCNN graph neural network, and take cracks as graph nodes; utilize an edge convolution mechanism to construct an adjacency graph through the spatio-temporal relationship between nodes, aggregate neighborhood features to update node representation, and capture local structure and global topology;

[0008] S30: Calculate crack feature similarity based on Euclidean distance, cluster according to a density threshold by using a DBSCAN algorithm, and group crack points with similar features into the same cluster, each cluster representing a crack surface and being assigned a unique number; and combine crack magnitude to optimize the clustering results and identify main crack surfaces with high expansion potential;

[0009] S40: Reconstruct the main crack surface by using a B-spline surface fitting algorithm to generate a smooth and continuous macroscopic crack surface; select control points to determine the main direction of the crack by using a principal component analysis method, construct a B-spline curve based on crack point distribution to generate a boundary contour, triangulate the outer contour line and divide the grid, and combine crack points and triangular grids to accurately fit the crack surface by using a B-spline algorithm.

[0010] In one example of the present application, in the step S10, the original acoustic wave data is subjected to positioning calculation and focal mechanism inversion to obtain focal mechanism core parameters directly related to rupture, including the following steps:

[0011] S11: Use the acoustic first arrival time data at different position sensors of the microseismic monitoring system to calculate the crack spatial coordinates by using a simplex positioning robust algorithm for spatial positioning calculation of the acoustic emission source;

[0012] S12: Use the acoustic first arrival time data at different position sensors of the microseismic monitoring system to calculate the crack spatial coordinates by using a simplex positioning robust algorithm for spatial positioning calculation of the acoustic emission source;

[0013] S13: Based on the constraint conditions of the tensile-shear rupture focal mechanism model, solve the six components M pq of the rupture moment tensor from the acoustic first arrival amplitude data collected by different position sensors.

[0014] S14: Quantitatively invert and calculate the coal rock rupture source core parameters according to the calculated moment tensor components, wherein the coal rock rupture source core parameters include crack spatial coordinates, crack volume, crack energy and crack spatial orientation.

[0015] In one example of the present application, in the step S14, the coal rock fracture source core parameter is calculated according to the following formula:

[0016] Fracture volume:

[0017]

[0018] In the formula, ΔV is the fracture volume of the fracture point, M1 and M3 are eigenvalues of the moment tensor M pq , μ is the Lame constant;

[0019] Fracture spatial orientation:

[0020]

[0021] n=(cosα, 0, ±sinα)

[0022]

[0023] In the formula, α is the angle between the motion direction of the fracture surface of the source and the normal direction, M1, M2 and M3 are eigenvalues of the moment tensor M pq , μ and λ are Lame constants, n is the spatial orientation, and b is the motion direction.

[0024] Fracture energy:

[0025]

[0026] In the formula, σ t is the tensile strength, σ s is the shear strength, ΔA is the fracture area of the fracture point, and σ is the material strength.

[0027] Fracture magnitude:

[0028]

[0029] In the formula, M1, M2 and M3 are eigenvalues of the moment tensor M pq .

[0030] In one example of the present application, in the step S20, the fracture source core parameter is preprocessed by using a standardization processing method; specifically including the following:

[0031] Suppose that the preprocessed fracture source core parameter is x1, x2, …, x n-1 , x n , which is mapped to the interval [0, 1] by the minimum-maximum normalization method, and the standardization formula is as follows:

[0032]

[0033] In the formula, a minimum value of a pre-processing parameter, a maximum value of a pre-processing parameter, y i ∈ [0, 1].

[0034] In an example of the present application, in the step S20, the crack spatio-temporal topology analysis model comprises a spatial transformation module, an edge convolution module, a maximum pooling layer and a fully connected layer,

[0035] The spatial transformation module is configured to normalize and align the input point cloud data, and an affine transformation matrix is obtained through learning, which is used to convert the input point cloud data into a standard coordinate system; this helps to eliminate the rotation and translation differences of the point cloud data, thereby improving the robustness and generalization ability of the network.

[0036] The edge convolution module is configured to perform convolution operations for feature extraction and modeling of local structure information on the point cloud data, dynamically learning and extracting local structure features;

[0037] The maximum pooling layer is configured to aggregate the features of the crack points, and extract the most representative local information;

[0038] The fully connected layer is configured to map the global features to a higher dimensional space, so as to capture complex nonlinear patterns and potential relationships in the data;

[0039] In an example of the present application, the convolution operation for feature extraction and modeling of local structure information on the point cloud data dynamically learns and extracts local structure features, which specifically includes the following steps:

[0040] (a) Point cloud data input: assuming that the input point cloud data contains n crack points, the point cloud data X is represented as:

[0041] X = {x1, x2,.... xn} ∈ R n}∈R D

[0042] In the formula: x i = {px i , py i , pz i , t i , nx i , ny i , nz i , V i , E i} is the feature of the i-th crack point; D is the feature dimension; {px i , py i , pz i} is the spatial coordinates of the i-th crack point; t i is the timestamp of the i-th crack point; {nxi ,ny i ,nz i}is the spatial orientation of the i-th fracture point; V i is the fracture volume of the i-th fracture point; E i is the release energy of the i-th fracture point;

[0043] (b) establishing an adjacency relationship: the adjacency relationship is used to establish a graph structure between the point clouds, each fracture point as a node of the graph, and the adjacency relationship between the nodes needs to be dynamically constructed according to a predefined rule; wherein, whether two fracture points establish a connection is judged based on time adjacency, spatial topological relationship, spatial orientation similarity and energy similarity;

[0044] (c) constructing a graph structure: defining a directed graph G used to represent the local point cloud structure, G composed of vertices V and directed edges E, expressed as:

[0045] G=(V,E),E∈V×V

[0046] In the formula: V i ={x1,x2,....x n}is the set of all positioning points, represents the set of directed edges of x i and its adjacent k vertices;

[0047] (d) edge feature extraction: calculating in the manner of obtaining local features with the difference between the positioning point and its adjacent points, and extracting edge features by using a multi-layer perception f Θ , assuming that the edge feature between the positioning point x i and its j-th adjacent point is:

[0048]

[0049] In the formula: f Θ is a series of nonlinear functions parameterized by a set of learnable parameters Θ;

[0050] (e) feature aggregation: after the edge feature extraction of each center point, first, linear transformation is performed through a full connection layer, and then normalization processing is performed; then, the features of the center point and the features of the points in its local neighborhood are aggregated through an aggregation function to obtain the aggregated features, the input center point features and the aggregated center point features are added to realize residual connection; finally, the ReLU activation function is used to realize nonlinear mapping to obtain the output of the edge convolutional network.

[0051] In an example of the present application, in step (b) of establishing an adjacency relationship, whether two fracture points establish a connection needs to meet the following conditions simultaneously:

[0052] Temporal proximity: the difference between the two localization points in time should be less than a time threshold, indicating that they occur close in time and possibly belong to the same crack propagation process:

[0053] |t i -t j |≤Δt

[0054] where t i , t j are the fracture times of crack points i and j respectively, and Δt is a predefined time threshold;

[0055] Spatial topological relation: by calculating the geometric relation between crack points, three topological relations are defined: intersection, adjacency and disjunction; only when two crack points are intersected or adjacent can an adjacency relation be established, indicating that they are close enough in space and can be penetrated, belonging to the same crack;

[0056] where the spatial distance d ij between crack point i and other crack point j is calculated:

[0057]

[0058] where (x i , y i , z i ), (x j , y j , z j ) are the localization spatial coordinates of crack points i and j respectively;

[0059] Spatial orientation similarity: the spatial orientation vectors of two localization points need to be similar within a certain error range, then they are considered to belong to the same crack, and the similarity of the direction is judged by calculating the angle between the direction vectors, where the expression of the angle between the direction vectors is:

[0060]

[0061] where n i = (nx i , ny i , nz i ) and n j = (nx j , ny j , nz j ) are the spatial orientation vectors of two crack points, and Δθ is a predefined angle error range;

[0062] Energy similarity: the energy difference of two positioning points should be less than a certain threshold, indicating that they have similar energy release patterns and can be physically connected or penetrated, and its judgment expression is:

[0063] |E i -E j |≤ΔE

[0064] wherein: E i , E j are the release energy values of the fracture points i and j respectively, and ΔE is a predefined exponential threshold.

[0065] In an example of the present application, in the step S30, the DBSCAN algorithm is used for density threshold clustering, including the following steps:

[0066] S31: Feature vector reading and Euclidean distance calculation: read the high-dimensional feature vector of each fracture point obtained by the fracture space-time topology analysis model, calculate the Euclidean distance between the fracture points according to the high-dimensional feature vector, and measure the similarity between the fracture points; wherein, for each pair of fracture points P i =(x1,x2,x3....x n ) and P j =(y1,y2,y3....y n ), the Euclidean distance calculation formula is as follows:

[0067]

[0068] wherein: n is the dimension of the feature vector, x k and y k are the values of the kth dimension in the high-dimensional feature vectors of the two fracture points, respectively;

[0069] S32: Determine the neighborhood distance and the minimum number of points: select an appropriate neighborhood distance ε and the minimum number of points MinPts; based on the Euclidean distance and the neighborhood distance ∈, calculate the neighborhood of each fracture point, if the number of points contained in the neighborhood of a fracture point is greater than or equal to MinPts, then the fracture point is a core point; otherwise, it is a boundary point or a noise point;

[0070] S33: Cluster expansion and core point processing: select one from the unmarked core points as the initial point, mark it as the starting point of a new cluster, and assign a cluster number to it; expand the neighborhood of the core point, and put all the points in the neighborhood into the same cluster; check whether the newly added point in the expanded cluster is a core point; if it is a core point, continue to expand the cluster; if it is a boundary point, keep its cluster label unchanged; continue to select new points from the unmarked core points for cluster expansion until all core points are processed;

[0071] S34: Cluster number and cluster result check: according to the result of DBSCAN clustering in step S33, a unique crack surface number is assigned to each cluster, and each cluster represents an independent crack surface; if a certain fracture point belongs to the neighborhood of multiple core points at the same time, the point will be assigned to the crack surface number of the first discovered cluster; then check the clustering result to ensure that each cluster is assigned a unique crack surface number, and the number of noise points is-1; if it is found that the cluster number assignment is wrong or missing, necessary adjustments are made to ensure that the final crack surface number is accurate;

[0072] S35: Cluster result optimization based on magnitude: after obtaining the preliminary clustering result, the clustering result is optimized by combining the magnitude of each fracture point; by visualizing the magnitude distribution of the fracture points in the crack surface, the fracture points with larger magnitude are identified, which belong to the main crack surface; the points with small magnitude represent the secondary crack surface or the crack surface with smaller expansion potential.

[0073] In one example of the present application, in step S40, B-spline surface fitting algorithm is used to reconstruct the fracture points of the main crack surface to generate smooth and continuous macroscopic crack surface, including the following steps:

[0074] S41: From the results of clustering analysis, read all effective fracture points classified as the same crack surface; wherein each fracture point includes fracture space coordinates, fracture volume, fracture space orientation, fracture energy and fracture magnitude;

[0075] S42: Combine energy and normal direction to comprehensively describe the spatial characteristics and fracture influence of the fracture point;

[0076] Assuming that the normal direction vector of a certain fracture point is n = (n x ,n y ,n z ), the formula for calculating the length of the vector is as follows:

[0077]

[0078] The normal vector is normalized to ensure that the normalized normal vector only represents the direction, and the normalization process calculation formula is as follows:

[0079]

[0080] For each fracture point, the comprehensive feature weight is calculated according to its fracture energy and normal direction; assuming that the feature weight W i of fracture point i is the combination of fracture energy and normal direction, then the calculation formula of the comprehensive feature weight of each fracture point is as follows:

[0081]

[0082] wherein, is the normalized length of the normal direction, usually 1; E i is the fracture energy of the i-th crack point; and alpha and beta are weight coefficients;

[0083] S43: Based on the feature weight and three-dimensional spatial position information of the same cluster crack point, the crack points with higher feature weight are extracted as key control points for crack surface reconstruction through principal component analysis, so as to determine the main direction of the crack surface;

[0084] S44: After determining the main direction of the crack surface, the boundary of the same cluster crack point is fitted by using a B-spline curve fitting method to form a closed curve as the final crack surface contour line;

[0085] S45: The crack surface contour line formed is triangulated, and the internal part of the final crack surface is triangulated;

[0086] S46: Read all the same cluster crack points, and combine the established triangular grid to finally fit the final crack surface to be generated, to form the final macroscopic crack surface of the same cluster crack point.

[0087] Another object of the present application is to provide a crack surface intelligent reconstruction device based on a dynamic graph neural network, comprising:

[0088] The parameter acquisition module is configured to acquire original acoustic wave data by a microseismic monitoring system, perform positioning calculation and source mechanism inversion on the original acoustic wave data, and obtain coal rock fracture source core parameters directly related to the fracture;

[0089] The model construction module is configured to pre-process the fracture source core parameters by using a standardization processing method, construct a crack spatio-temporal topology analysis model based on a DGCNN graph neural network, and take the cracks as graph nodes; utilize an edge convolution mechanism to construct an adjacency graph through the spatio-temporal relationship between nodes, aggregate neighborhood features to update node representations, capture local structures and global topologies; the crack surface clustering module is configured to calculate crack feature similarity based on Euclidean distance, cluster according to a density threshold by using a DBSCAN algorithm, and assign the same cluster to crack points with similar features, each cluster representing a crack surface and being assigned a unique number; and combine the crack magnitude to optimize the clustering results and identify main crack surfaces with high expansion potential;

[0090] The crack surface reconstruction module is configured to reconstruct the main crack surface by using a B-spline surface fitting algorithm to generate a smooth and continuous macroscopic crack surface; determine the crack main direction by selecting control points through a principal component analysis method, construct a B-spline curve based on crack point distribution to generate a boundary contour, triangulate the outer contour line and divide the grid, and combine the crack points and the triangular grid to accurately fit the crack surface by using a B-spline algorithm.

[0091] The present application has the following beneficial effects relative to the prior art:

[0092] 1、The present application can directly reflect the source characteristics of the fracture point by obtaining the core parameters of the seismic source mechanism through microseismic monitoring inversion, and has high reliability and accuracy. Compared with the traditional representation method relying only on spatial positioning points, the core parameters provided by the seismic source mechanism inversion as data input can more accurately capture the spatio-temporal correlation and physical connectivity between fractures, significantly improve the scientificity and reliability of intelligent reconstruction of the fracture surface, and are closer to the real evolution law of the internal fracture of the rock mass, thereby laying a solid data foundation for analyzing the fracture propagation and penetration mechanism.

[0093] 2、The present application constructs a fracture spatio-temporal topology analysis model based on a dynamic graph neural network, captures the spatio-temporal correlation and global topology characteristics of the fracture point through edge convolution, and uses the adjacency relationship to deeply mine the spatio-temporal correlation, physical connectivity and topology structure between the fracture points, thereby breaking through the limitations of traditional methods in dealing with complex fracture networks, significantly improving the accuracy and topology representation ability of fracture network analysis, and providing strong technical support for intelligent reconstruction of the fracture surface.

[0094] 3、To overcome the shortcomings of the traditional fracture representation method in analyzing the spatio-temporal topology correlation and penetration mechanism, the present application improves the adjacency relationship in the graph neural network, further integrates the spatial orientation and energy dissipation characteristics of the fracture point on the basis of considering the spatial position and time sequence relative relationship of the fracture point, deeply mines the physical connectivity and topology structure between the fracture points, and more effectively captures and analyzes the complex spatio-temporal topology relationship between the fracture points.

[0095] 4、The present application can reconstruct the geometric shape and dynamic evolution characteristics of the fracture surface with high precision through the combination of graph neural network modeling, DBSCAN clustering and B-spline fitting, directly display the propagation process, time sequence evolution and spatial distribution of the macroscopic fracture surface, greatly improve the accuracy and reliability of fracture surface inversion, provide advanced technical support for in-depth analysis of the rock mass failure evolution mechanism, and provide a more scientific basis for rock mass engineering stability evaluation and disaster warning.

[0096] The optimal embodiments of the present application will be described in more detail below with reference to the accompanying drawings, so that the features and advantages of the present application can be easily understood. BRIEF DESCRIPTION OF DRAWINGS

[0097] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application will be briefly introduced below. The drawings only serve to show some embodiments of the present application, and do not limit the entire embodiments of the present application to this.

[0098] Figure 1A flowchart of a dynamic graph convolutional neural network based fracture surface reconstruction method according to an embodiment of the present application;

[0099] Figure 2 A DGCNN schematic diagram according to an embodiment of the present application;

[0100] Figure 3 A convolutional structure schematic diagram according to an embodiment of the present application;

[0101] Figure 4 An edge convolution flowchart according to an embodiment of the present application;

[0102] Figure 5 A DBSCAN clustering flowchart according to an embodiment of the present application.

[0103] Figure 6 A B-spline surface fitting fracture surface flowchart according to an embodiment of the present application. DETAILED DESCRIPTION

[0104] In order to make the purpose, technical scheme and advantages of the technical scheme of the present application more clear, the technical scheme of the embodiments of the present application will be described clearly and completely below in combination with the drawings of the embodiments of the present application. The same reference signs in the drawings represent the same components. It should be noted that the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without any creative effort fall within the scope of protection of the present application.

[0105] Unless otherwise defined, technical terms or scientific terms used herein should be understood as having the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terms "first", "second", and similar terms used in the description and the claims of the present application do not necessarily mean any order, number, or importance, but are only used to distinguish different components. Similarly, the terms "one" or "a" or similar terms do not necessarily mean a quantity limitation. The terms "include" or "contain" or similar terms mean that the elements or objects before the terms cover the elements or objects listed after the terms and their equivalents, without excluding other elements or objects. The terms "connected" or "connected" or similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "up", "down", "left", "right", and the like only represent relative positional relationships, which may

[0106] A dynamic graph convolutional neural network based fracture surface reconstruction method according to the first aspect of the present application, as shown in Figure 1 includes the following steps:

[0107] S10: Obtain original acoustic wave data through a microseismic monitoring system, perform positioning calculation and source mechanism inversion on the original acoustic wave data, and obtain coal rock fracture source core parameters directly related to the fracture; wherein the coal rock fracture source core parameters include: fracture spatial coordinates, fracture volume, fracture energy, fracture spatial orientation, and fracture magnitude.

[0108] S20: Preprocess the fracture source core parameters using a standardization processing method, construct a fracture spatio-temporal topology analysis model based on a DGCNN graph neural network, take the fracture as a graph node, and the node features include time, spatial coordinates, volume, energy, and spatial orientation; use edge convolution mechanism to construct an adjacency graph through the spatio-temporal relationship between nodes, and aggregate neighborhood features to update node representation, capture local structure and global topology; wherein, to ensure the quality and applicability of the input data, the data obtained by inversion calculation is preprocessed to provide standardized data for the input of the subsequent graph neural network model, avoiding interference of inconsistent data on model training, thereby improving the accuracy and generalization ability of the model; wherein, the model outputs a high-dimensional feature vector by optimizing the parameters by minimizing the loss function, which comprehensively reflects the spatio-temporal correlation, physical connectivity, and topological structure;

[0109] S30: Calculate the similarity of fracture features based on the Euclidean distance, use the DBSCAN algorithm to cluster according to the density threshold, and group the fracture points with similar features into the same cluster, each cluster represents a crack surface, and a unique number is assigned, and noise points are excluded to reduce interference; and combine the fracture magnitude to optimize the clustering results and identify the main crack surface fracture with high expansion potential;

[0110] S40: Reconstruct the main crack surface using B-spline surface fitting algorithm to generate a smooth and continuous macroscopic crack surface; select control points to determine the main direction of the crack by principal component analysis, construct B-spline curves based on the distribution of fracture points to generate the boundary contour, triangulate the outer contour line and divide the grid, and combine the fracture points and triangular grid to accurately fit the fracture surface using B-spline algorithm.

[0111] The method can obtain source mechanism core parameters through microseismic monitoring inversion, which can directly reflect the source characteristics of the fracture points, and has high reliability and accuracy. Compared with the traditional representation method relying only on spatial positioning points, the core parameters provided by the source mechanism inversion as data input can more accurately capture the spatio-temporal correlation and physical connectivity between fractures, significantly improve the scientificity and reliability of intelligent reconstruction of the fracture surface, and be closer to the real evolution law of the internal fracture of the rock mass, laying a solid data foundation for analyzing the fracture expansion and penetration mechanism.

[0112] The method constructs a fissure space-time topology analysis model based on a dynamic graph neural network, captures the space-time correlation and global topology characteristics of fissure points through edge convolution, deeply mines the space-time correlation, physical connectivity and topology structure between fissure points by using adjacency relationship, breaks through the limitations of traditional methods in processing complex fissure networks, and significantly improves the accuracy and topology characterization capability of fissure network analysis, thereby providing strong technical support for intelligent reconstruction of fissure surfaces.

[0113] In order to overcome the deficiency of traditional fissure characterization methods in analyzing space-time topology correlation and through mechanism, the method improves the adjacency relationship in the graph neural network, further fuses the spatial orientation and energy dissipation characteristics of the fissure points on the basis of considering the spatial position and time sequence relative relationship of the fissure points, deeply mines the physical connectivity and topology structure between the fissure points, and more effectively captures and analyzes the complex space-time topology relationship between the fissure points.

[0114] The method can reconstruct the geometric shape and dynamic evolution characteristics of the fissure surface with high precision by combining the graph neural network modeling, DBSCAN clustering and B-spline fitting, directly display the expansion process, time sequence evolution and spatial distribution of the macro fissure surface, greatly improve the accuracy and reliability of the fissure surface inversion, provide advanced technical support for in-depth analysis of the rock mass failure evolution mechanism, and provide a more scientific basis for rock mass engineering stability evaluation and disaster warning.

[0115] In one example of the present application, in the step S10, the original acoustic wave data is subjected to positioning calculation and source mechanism inversion to obtain source mechanism core parameters directly related to the rupture, including the following steps:

[0116] S11: The original acoustic wave data collected by the microseismic monitoring system is subjected to post-processing, and the Akaike information criterion method is used to extract the first arrival time and amplitude information;

[0117] S12: The first arrival time data of the acoustic wave at different positions of the sensor of the microseismic monitoring system is used, and the simplex positioning robust algorithm is used to calculate the spatial positioning of the acoustic emission source to obtain the fissure spatial coordinates, wherein the positioning control equation is as follows:

[0118]

[0119] In the formula, x (x, y, z) and x i (x i ,y i ,z i ) are the spatial coordinates of the acoustic emission source and the i th sensor respectively, c p is the P wave velocity, t i is the P wave arrival time at the i th sensor, and t is the earthquake time of the acoustic emission source;

[0120] S13: solving 6 components of the rupture moment tensor M of the seismic source based on the amplitude data of the first wave of the sound wave collected by the sensors at different positions under the constraint condition of the tensile-shear rupture source model pq ;

[0121] S14: quantitatively inverting and calculating the core parameters of the coal rock rupture source according to the calculated moment tensor components, wherein the core parameters of the coal rock rupture source include a crack volume, a crack energy, a crack spatial orientation and a crack magnitude.

[0122] In an example of the present application, in the step S14, the core parameters of the coal rock rupture source are specifically calculated according to the following formula:

[0123] Crack volume:

[0124]

[0125] In the formula, ΔV is the rupture volume of the crack point, M1 and M3 are eigenvalues of the moment tensor M pq , μ is the Lame constant;

[0126] Crack spatial orientation:

[0127]

[0128] n=(cosα, 0, ±sinα)

[0129]

[0130] In the formula, α is the included angle between the movement direction and the normal direction of the rupture surface of the seismic source, M1, M2 and M3 are eigenvalues of the moment tensor M pq , μ and λ are Lame constants, n is the spatial orientation, and b is the movement direction.

[0131] Crack energy:

[0132]

[0133] In the formula: σ t is the tensile strength, σ s is the shear strength, ΔA is the rupture area of the crack point, and σ is the material strength.

[0134] Crack magnitude:

[0135]

[0136] In the formula, M1, M2 and M3 are eigenvalues of the moment tensor M pq .

[0137] In one example of the present application, in the step S20, the rupture source core parameters are preprocessed by a standardization processing method to ensure that the parameter values are in the same scale range and to avoid interference of parameters with different dimensions on subsequent analysis; specifically including the following:

[0138] Suppose the preprocessed rupture source core parameters are x1, x2, …, x n-1 , x n , which are mapped to the interval [0, 1] by the min-max normalization method, and the standardization formula is as follows:

[0139]

[0140] In the formula, is the minimum value of the preprocessed parameter, is the maximum value of the preprocessed parameter, y i ∈[0,1].

[0141] In one example of the present application, in the step S20, as shown in Figure 2 and Figure 3 , the crack spatiotemporal topology analysis model includes a spatial transformation module, an edge convolution module, a maximum pooling layer, and a full connection layer,

[0142] The spatial transformation module is configured to normalize and align the input point cloud data, and an affine transformation matrix is obtained through learning to convert the input point cloud data into a standard coordinate system; this helps to eliminate the rotation and translation differences of the point cloud data, thereby improving the robustness and generalization ability of the network.

[0143] The edge convolution module is configured to perform convolution operations on the point cloud data to extract features and model local structure information, dynamically learning and extracting local structure features. Each point in the point cloud data is regarded as a vertex of a graph. For each point, a set of points related to the point can be determined by defining an adjacency relationship, thereby obtaining a directed graph with weighted edges. Then, the features in these local neighborhoods are aggregated to obtain the local feature representation of the point. In order to update the feature representation of each point, a multi-layer perceptron or a convolution kernel and feature dot product operation can be used. Through these operations, a higher level of feature representation of each point can be extracted. Finally, the edge convolution uses maximum pooling to summarize the features of all points to obtain the global feature representation of the entire point cloud data.

[0144] The maximum pooling layer is configured as a global pooling method for aggregating the features of the crack points and extracting the most representative local information. In the neighborhood of each crack point, the maximum pooling selects the maximum value, thereby focusing on the most significant features, which helps to improve the robustness and stability of the model. Specifically, the maximum pooling gives the model translational invariance by reducing the impact of spatial transformation, that is, the pooled features remain consistent regardless of the position or spatial configuration of the crack points. This feature enables the DGCNN to effectively process unordered acoustic emission crack point cloud data, identify common patterns and structural features of crack propagation, and capture changes in spatio-temporal relationships. Finally, the global features aggregated by the maximum pooling layer are passed as input to the fully connected layer to provide high-level feature representation for further crack clustering and path analysis.

[0145] The fully connected layer is configured to further process and optimize the pooled global features. In the DGCNN, the maximum pooling layer aggregates the local features of the point cloud into global features, while the fully connected layer is responsible for mapping these global features to a higher dimensional space to capture complex nonlinear patterns and potential relationships in the data. The role of the fully connected layer is to enhance the expressive power of the features through layer-by-layer linear transformation and nonlinear activation. After multiple layers of fully connected processing, the network outputs a high-dimensional feature vector that integrates the global information and nonlinear relationships of the input point cloud data, effectively representing the features of the entire point cloud.

[0146] In one example of the present application, as shown in FIG. 1, the convolution operation is performed on the point cloud data to extract features and model local structural information, dynamically learning and extracting local structural features, which includes the following steps: Figure 4

[0147] (a) Point cloud data input: assuming that the input point cloud data contains n crack points, the point cloud data X is represented as:

[0148] X = {x1, x2,.... xn} ∈ R n D

[0149] where: x i = {px i , py i , pz i , t i , nx i , ny i , nz i , V i , E i is the feature of the i-th crack point; D is the feature dimension; {px i , py i , pz​​i} is the spatial coordinate of the i-th fracture point; t i is the time stamp of the i-th fracture point; {nx i ,ny i ,nz i} is the spatial orientation of the i-th fracture point; V i is the fracture volume of the i-th fracture point; E i is the released energy of the i-th fracture point;

[0150] (b) Establishing adjacency relationship: the adjacency relationship is used to establish the graph structure between the point cloud, each fracture point as a node of the graph, and the adjacency relationship (edge) between the nodes needs to be dynamically constructed according to the pre-defined rules; wherein, whether two fracture points establish connection is judged based on time adjacency, spatial topological relationship (including spatial distance and volume), spatial orientation similarity and energy similarity; that is, if the above four conditions are met at the same time, the next step is executed, otherwise, the next fracture point is judged;

[0151] (c) Constructing graph structure: define a directed graph G used to represent the local point cloud structure, G is composed of vertices V and directed edges E, and its expression is:

[0152] G = (V, E), E ∈ V × V

[0153] In the formula: V i = {x1, x2,....x n} is the set of all positioning points, represents the set of directed edges of x i and its adjacent k vertices;

[0154] (d) Edge feature extraction: calculate in the manner of obtaining local features with the difference between the positioning point and its adjacent points, and extract the edge features by using a multi-layer perception f Θ , suppose that the edge feature between the positioning point x i and its j-th adjacent point is:

[0155]

[0156] In the formula: f Θ is a series of nonlinear functions parameterized by a set of learnable parameters Θ;

[0157] (e) Feature aggregation: After the edge feature extraction of each center point, first, a linear transformation is performed through a fully connected layer, and then a normalization process is performed to ensure that the mean value of the data is 0 and the variance is 1; then, the features of the center point and the features of the points in the local neighborhood of the center point are aggregated through an aggregation function to obtain the aggregated features, in order to prevent the gradient disappearance phenomenon in the training process, the input center point features and the aggregated center point features are added to realize the residual connection; finally, the ReLU activation function is used to realize the nonlinear mapping to obtain the output of the edge convolutional network.

[0158] In one example of the present application, in the step (b) of establishing the adjacency relationship, it is necessary to determine whether two crack points are connected by satisfying the following conditions simultaneously:

[0159] Temporal proximity: the difference between the two positioning points in time should be less than a time threshold, indicating that they occur at close times and may belong to the same crack propagation process:

[0160] |t i -t j |≤Δt

[0161] In the formula: t i , t j are the fracture times of crack points i and j respectively, and Δt is a predefined time threshold;

[0162] Spatial topological relationship: by calculating the geometric relationship between crack points, three topological relationships are defined: intersection, adjacency and separation; only when two crack points intersect or are adjacent can an adjacency relationship be established, indicating that they are close enough in space and can be connected, belonging to the same crack;

[0163] wherein the spatial distance d ij between crack point i and other crack point j is calculated:

[0164]

[0165] In the formula: (x i , y i , z i ), (x j , y j , z j ) are the positioning spatial coordinates of crack points i and j respectively;

[0166] In consideration of the calculation error of the crack geometric parameters, the radius of the crack is increased by 5% and decreased by 5% respectively as the inner and outer diameters of the crack, and the spatial topological relationship between the cracks is established:

[0167] Assuming that the inner and outer diameters of crack points x i and x j are ri,in r j,in and r i,out r j,ou ;

[0168] Intersection: If the inner diameters of two fracture points xi and xj intersect, i.e., d ij ≤(r i,in +r j,in This indicates that the crack points directly overlap in space, may belong to the same crack, or at least are physically connected to each other and can establish an adjacency relationship;

[0169] Adjacent: If the outer diameters of two fracture points xi and xj intersect but their inner diameters do not intersect, i.e., d ij ≤(r i,out +r j,out And d ij >(r i,in +r j,in ), indicating that the crack points partially overlap, may be on different expansion paths, and can establish an adjacency relationship, but the degree of connection is weak;

[0170] Separated: If two fracture points x i and x j The inner and outer diameters do not intersect, and the distance between them is greater than the sum of the outer diameters, i.e., d. ij >(r i,out +r j,out This indicates that there is no physical connection between the two fracture points, and they are not adjacent to each other.

[0171] Spatial orientation similarity: If the spatial orientation vectors of two positioning points are similar within a certain error range, they are considered to belong to the same crack. The similarity of directions is determined by calculating the angle between the direction vectors. The expression for the angle between the direction vectors is as follows:

[0172]

[0173] In the formula, n i =(nx i ,ny i ,nz i ) and n j =(nx j ,ny j ,nz j ) are the spatial orientation vectors of the two fracture points, and Δθ is the predefined range of the included angle error;

[0174] Energy similarity: The energy difference between two locations should be less than a certain threshold, indicating that they have similar energy release patterns and can be physically connected or interoperable. The judgment expression is:

[0175] |Ei -E j |≤ΔE

[0176] wherein: E i , E j are the release energy values of the fracture points i and j respectively, and ΔE is a predefined exponential threshold value.

[0177] In one example of the present application, in the step S30, as shown in Figure 5 , the DBSCAN algorithm is used to cluster by density threshold, including the following steps:

[0178] S31: Feature vector reading and Euclidean distance calculation: read the high-dimensional feature vector of each fracture point obtained by the fracture space-time topology analysis model, and calculate the Euclidean distance between the fracture points according to the high-dimensional feature vectors to measure the similarity between the fracture points; wherein, for each pair of fracture points P i = (x1, x2, x3...x n ) and P j = (y1, y2, y3...y n ), the Euclidean distance calculation formula is as follows:

[0179]

[0180] wherein: n is the dimension of the feature vector, x k and y k are the values of the kth dimension in the high-dimensional feature vectors of the two fracture points, respectively;

[0181] S32: Determine the neighborhood distance and the minimum number of points: select an appropriate neighborhood distance ε and the minimum number of points MinPts; ε determines the neighborhood radius of a fracture point in the Euclidean space, and MinPts is the minimum number of points required to be contained in the neighborhood, which is commonly used to judge whether a point is a core point. Based on the Euclidean distance and the neighborhood distance ε, the neighborhood of each fracture point is calculated, and if the number of points contained in the neighborhood of a fracture point is greater than or equal to MinPts, the fracture point is a core point; otherwise, it is a boundary point or a noise point;

[0182] S33: Cluster expansion and core point processing: select one from the unmarked core points as the initial point, mark it as the starting point of a new cluster, and assign a cluster number to it. Then, expand the neighborhood of the core point, and put all the points in the neighborhood (including core points and density-reachable boundary points) into the same cluster; for the newly added points in the expanded cluster, check whether they are core points; if they are core points, continue to expand the cluster; if they are boundary points, keep their cluster labels unchanged; continue to select new points from the unmarked core points for cluster expansion until all core points are processed;

[0183] That is, as shown inFigure 5 As shown, for the core point, it is necessary to determine whether each core point has a label and whether all core points are read, if the core point has no label, the neighborhood needs to be expanded and the label is set, otherwise the next core point is read, if it is determined that there is a point that has not been read among all core points, it is necessary to continue to determine whether the core point has a label, otherwise, all crack clusters are generated; wherein, after the core point is expanded and the label is set, it is necessary to determine whether the point newly added in the neighborhood is a core point, if it is a core point, it is necessary to continue to expand and set the same label as the initial core point, and it is necessary to continue to determine whether the point newly added in the neighborhood is a core point; otherwise, the label remains unchanged, and a new crack cluster is generated to form all crack clusters.

[0184] S34: Cluster number and cluster result check: according to the result of DBSCAN clustering in step S33, a unique crack surface number is assigned to each cluster, and each cluster represents an independent crack surface; if a certain fracture point belongs to the neighborhood of multiple core points (i.e. the point is a density reachable point), the point will be assigned to the crack surface number of the first discovered cluster; then the clustering result is checked to ensure that each cluster is assigned a unique crack surface number, and the number of noise points is -1; if it is found that the cluster number assignment is wrong or missing, necessary adjustments are made to ensure that the final crack surface number is accurate;

[0185] S35: Cluster result optimization based on magnitude: after obtaining the preliminary clustering result, the clustering result is optimized by combining the magnitude of each fracture point, the magnitude distribution of the fracture points in the crack surface is visualized, and the fracture points with larger magnitude are identified, i.e. belonging to the main crack surface; the points with small magnitude represent the secondary crack surface or the crack surface with smaller expansion potential; according to the distribution characteristics of the magnitude, the division of the fracture points in each cluster is re-evaluated, the fracture points with larger magnitude are preferentially marked as part of the main crack surface, and the fracture points with smaller magnitude are divided into secondary crack surfaces. For some fracture points with larger magnitude, if they are located in different clusters, merging or re-dividing the clusters can be considered to better reflect the expansion potential of the fractures. The optimized clustering result will be visualized in three-dimensional space, and the main crack surface and the secondary crack surface are distinguished by different colors or sizes, helping to analyze the crack expansion characteristics.

[0186] In one example of the present application, in the step S40, as shown, Figure 6 a B-spline surface fitting algorithm is used to reconstruct the fracture points of the main crack surface to generate a smooth and continuous macroscopic crack surface, including the following steps:

[0187] S41: read all effective fracture points classified into the same crack surface from the clustering analysis result; wherein, each fracture point includes its fracture space coordinates, fracture volume, fracture energy, fracture space orientation and fracture magnitude;

[0188] S42: Since the release energy of the crack point reflects its fracture strength, and the higher the release energy, the more likely it belongs to an area with faster expansion speed or greater expansion potential. Therefore, energy as a feature weight can more effectively guide the crack surface reconstruction process. At the same time, the normal direction of the crack point provides spatial orientation information of the crack surface, so the combination of energy and normal direction comprehensively describes the spatial characteristics and fracture influence of the crack point;

[0189] Suppose the normal direction vector of a certain crack point is n = (n x ,n y ,n z ), the length of this vector is calculated by the following formula:

[0190]

[0191] Since the size of the normal vector itself may be affected by factors such as the position and size of the crack point, the normal vector is normalized to ensure that the normalized normal vector only represents the direction. The normalization process is calculated as follows:

[0192]

[0193] For each crack point, the comprehensive feature weight is calculated according to its fracture energy and normal direction. Suppose the feature weight W i of crack point i is the combination of fracture energy and normal direction, the following weighted formula is used to calculate the comprehensive feature weight of each crack point:

[0194]

[0195] In the formula, is the length of the normalized normal direction, usually 1; E i is the fracture energy of the i-th crack point; α and β are weight coefficients to adjust the influence of normal direction and fracture energy on the final feature weight;

[0196] S43: Based on the feature weight and three-dimensional spatial position information of the crack points in the same cluster, the crack points with higher feature weight are extracted by principal component analysis (PCA) as the key control points for crack surface reconstruction, thereby determining the main direction of the crack surface. Specifically, the two crack points with the largest feature weight are selected as the initial control points for PCA, and the main direction of the final crack surface is defined and spanned by these points, providing a basic vector for subsequent fitting and morphology construction;

[0197] S44: After determining the main direction of the crack surface, the boundary of the crack points in the same cluster is fitted using the B-spline curve fitting method to form a closed curve as the outer contour line of the final crack surface;

[0198] S45: Triangulate the outer contour line of the crack surface that has been formed, and triangulate the interior of the final crack surface;

[0199] S46: Read all the fissure points of the same cluster, combine the established triangular grid to finally fit the final crack surface to be generated, and form the final macroscopic crack surface of the fissure points of the same cluster.

[0200] According to the second aspect of the application, an intelligent reconstruction device for a fissure surface based on a dynamic graph neural network comprises:

[0201] The parameter acquisition module is configured to acquire original acoustic wave data by a microseismic monitoring system, perform positioning calculation and source mechanism inversion on the original acoustic wave data, and obtain coal rock rupture source core parameters directly related to the rupture.

[0202] The model construction module is configured to preprocess the rupture source core parameters by using a standardization processing method, construct a fissure space-time topology analysis model based on a DGCNN graph neural network, take the fissure as a graph node, and take the node features as time, spatial coordinates, volume, energy and spatial orientation; an adjacency graph is constructed through the space-time relationship between nodes by using an edge convolution mechanism, and the node representation is updated by aggregating the neighborhood features to capture the local structure and global topology; the model outputs a high-dimensional feature vector by optimizing the parameters by minimizing the loss function, which comprehensively reflects the space-time correlation, physical connectivity and topology structure.

[0203] The crack surface clustering module is configured to calculate the fissure feature similarity based on the Euclidean distance, cluster according to the density threshold by using the DBSCAN algorithm, group the fissure points with similar features into the same cluster, and assign a unique number to each cluster representing a crack surface, and exclude noise points to reduce interference; and combine the fissure magnitude to optimize the clustering results and identify the main crack surface fissure with high expansion potential.

[0204] The crack surface reconstruction module is configured to reconstruct the main crack surface by using a B-spline surface fitting algorithm to generate a smooth and continuous macroscopic crack surface; the principal component analysis method is used to select control points to determine the main direction of the crack, a B-spline curve is constructed based on the distribution of the fissure points to generate the boundary contour, the outer contour line is triangulated and divided into a grid, and the B-spline algorithm is used to accurately fit the fissure surface in combination with the fissure points and the triangular grid.

[0205] The system obtains the core parameters of the focal mechanism by microseismic monitoring inversion, can directly reflect the source characteristics of the fracture points, and has high reliability and accuracy. Compared with the traditional representation method relying only on spatial positioning points, the core parameters provided by the focal mechanism inversion as data input can more accurately capture the spatio-temporal correlation and physical connectivity between fractures, significantly improve the scientificity and reliability of the intelligent reconstruction of the fracture surface, and are closer to the real evolution law of the internal fractures of the rock mass, thereby laying a solid data foundation for analyzing the fracture propagation and penetration mechanism.

[0206] The system constructs a fracture spatio-temporal topology analysis model based on a dynamic graph neural network, captures the spatio-temporal correlation and global topology characteristics of the fracture points through edge convolution, and uses adjacency relationships to deeply mine the spatio-temporal correlation, physical connectivity and topology structure between the fracture points, thereby breaking through the limitations of traditional methods in dealing with complex fracture networks and significantly improving the accuracy and topology representation capability of fracture network analysis, thereby providing strong technical support for the intelligent reconstruction of the fracture surface.

[0207] To overcome the deficiencies of traditional fracture representation methods in analyzing the spatio-temporal topology correlation and penetration mechanism, the system improves the adjacency relationship in the graph neural network, further integrates the spatial orientation and energy dissipation characteristics of the fracture points on the basis of considering the spatial position and time sequence relative relationship of the fracture points, deeply mines the physical connectivity and topology structure between the fracture points, and more effectively captures and analyzes the complex spatio-temporal topology relationship between the fracture points.

[0208] The system can reconstruct the geometric morphology and dynamic evolution characteristics of the fracture surface with high precision through the combination of graph neural network modeling, DBSCAN clustering and B-spline fitting, directly display the propagation process, time sequence evolution and spatial distribution of the macroscopic fracture surface, greatly improve the accuracy and reliability of the fracture surface inversion, provide advanced technical support for in-depth analysis of the rock mass failure evolution mechanism, and provide a more scientific basis for rock mass engineering stability evaluation and disaster warning.

[0209] The exemplary embodiments of the fracture surface reconstruction method and device based on the dynamic graph neural network proposed by the present application are described in detail above with reference to the preferred embodiments, however, those skilled in the art can understand that various modifications and improvements can be made to the above specific embodiments without departing from the concept of the present application, and various technical features and structures proposed by the present application can be combined without exceeding the protection scope of the present application, and the protection scope of the present application is determined by the appended claims.

Claims

1. A method for reconstructing a fracture surface based on a dynamic graph neural network, characterized in that, Includes the following steps: S10: Raw acoustic wave data is acquired through a microseismic monitoring system. The raw acoustic wave data is used for location calculation and focal mechanism inversion to obtain the core parameters of the coal and rock fracture source that are directly related to the fracture. Among them, the core parameters of the coal and rock fracture source include: fracture spatial coordinates, fracture volume, fracture energy, fracture spatial orientation, and fracture magnitude. S20: The core parameters of the fracture source are preprocessed using a standardized processing method. A spatiotemporal topology analysis model of the fracture is constructed based on the DGCNN graph neural network, with the fracture as a graph node. The adjacency graph is constructed through the spatiotemporal relationship between nodes using the edge convolution mechanism, and the node representation is updated by aggregating neighborhood features to capture the local structure and global topology. The standardized processing method is used to preprocess the core parameters of the rupture origin, specifically including the following: Assuming the core parameters of the pre-treated fracture origin are x 1, x 2, ..., x n-1 , x n It is mapped to an interval using the min-max normalization method. Its standardized formula is as follows: In the formula, This is the minimum value of the preprocessing parameters. This is the maximum value of the preprocessing parameters. y i ∈[0,1]; The spatiotemporal topology analysis model for the crack includes a spatial transformation module, an edge convolution module, a max pooling layer, and a fully connected layer. The spatial transformation module is configured to normalize and align the input point cloud data, and learns an affine transformation matrix to convert the input point cloud data into a normalized coordinate system. The edge convolution module is configured to perform convolution operations on point cloud data to extract features and model local structural information, and to dynamically learn and extract local structural features. The maximum pooling layer is configured to aggregate the features of the crack points and extract the most representative local information. The fully connected layer is configured to map global features to a higher-dimensional space in order to capture complex nonlinear patterns and potential relationships in the data; S30: Based on Euclidean distance, the similarity of fracture features is calculated. The DBSCAN algorithm is used to cluster fractures according to density thresholds, and fracture points with similar features are grouped into the same cluster. Each cluster represents a crack surface and is assigned a unique number. The clustering results are optimized by combining fracture magnitude to identify main crack surface fractures with high propagation potential. S40: The main crack surface is reconstructed using the B-spline surface fitting algorithm to generate a smooth and continuous macroscopic crack surface; the main direction of the crack is determined by selecting control points through principal component analysis; the boundary contour is generated by constructing B-spline curves based on the distribution of crack points; the outer contour line is triangulated and meshed; and the crack surface is accurately fitted using the B-spline algorithm by combining the crack points and the triangular mesh.

2. The crack surface reconstruction method based on dynamic graph neural network according to claim 1, characterized in that, In step S10, the original acoustic wave data is used for location calculation and focal mechanism inversion to obtain the core parameters of the focal mechanism directly related to the rupture, including the following steps: S11: Post-process the raw acoustic data acquired by the microseismic monitoring system, and use the red-delay information criterion method to extract the arrival time and amplitude information of the first wave; S12: Using the arrival time data of the first wave of acoustic waves at different locations of sensors in the microseismic monitoring system, the spatial coordinates of the fracture are obtained by spatially locating the acoustic emission source using a robust simplex localization algorithm. S13: Solving for the six components of the source rupture moment tensor based on the acoustic first-wave amplitude data collected by sensors at different locations under the constraint of the tension-shear rupture source model. M pq ; S14: Based on the calculated moment tensor components, quantitatively invert and calculate the core parameters of the coal and rock fracture source. Among them, the core parameters of the coal and rock fracture source include fracture volume, fracture energy, fracture spatial orientation, and fracture magnitude.

3. The crack surface reconstruction method based on dynamic graph neural network according to claim 2, characterized in that, In step S14, the specific calculation formula for the core parameters of the coal and rock fracturing source is as follows: Crack volume: In the formula, The fracture volume at the fracture point. M 1. M 3 is the moment tensor M pq eigenvalues, Let Lamé constant be denoted by . Spatial orientation of the fracture: In the formula, The angle between the direction of motion of the rupture surface at the earthquake source and the normal direction. M 1. M 2. M 3 is the moment tensor M pq eigenvalues, and Let Lamé constant be . n For spatial orientation, b In terms of direction of motion; fracture energy: In the formula: The fracture area at the crack point. For material strength, where For tensile strength, Shear strength; Fissure magnitude: In the formula, M 1. M 2. M 3 is the moment tensor M pq eigenvalues.

4. The crack surface reconstruction method based on dynamic graph neural network according to claim 1, characterized in that, Convolutional operations are performed on point cloud data to extract features and model local structural information, dynamically learning and extracting local structural features. The specific steps include: (a) Point cloud data input: Assuming the input point cloud data contains n crack points, then the point cloud data X is represented as: In the formula: For the first i Characteristics of each crack point; D For feature dimensions; For the first i Spatial coordinates of each fracture point; t i For the first i Timestamps of each crack point; For the first i Spatial orientation of each crack point; V i For the first i The fracture volume at each fracture point; E i For the first i Energy released at each crack point; (b) Establishing adjacency relationships: Adjacency relationships are used to establish the graph structure between point clouds. Each fracture point is a node in the graph, and the adjacency relationships between nodes are dynamically constructed according to predefined rules. Among them, the connection between two fracture points is determined based on temporal adjacency, spatial topological relationship, spatial orientation similarity, and energy similarity. (c) Constructing the graph structure: Define a directed graph G to represent the local point cloud structure. G consists of vertices V and directed edges E, and its expression is: In the formula: For the set of all positioning points, express x i Adjacent to k A set of directed edges with vertices; (d) Edge feature extraction: Local features are calculated by using the difference between the local point and its neighboring points, and a multilayer perceptron is employed. Edge features are extracted, assuming localization points. x i Its first j Neighboring points The edge features between them are: In the formula: It is a series of nonlinear functions parameterized by a set of learnable parameters Θ; (e) Feature aggregation: After extracting the edge features of each center point, the features are first linearly transformed through a fully connected layer and then normalized. Then, the features of the center point are aggregated with the features of the points in its local neighborhood through an aggregation function to obtain aggregated features. The input center point features are added to the aggregated center point features to achieve residual connections. Finally, the ReLU activation function is used to achieve nonlinear mapping to obtain the output of the edge convolutional network.

5. The crack surface reconstruction method based on dynamic graph neural network according to claim 4, characterized in that, In step (b), determining whether two fracture points should be connected requires that the following conditions be met simultaneously: Temporal contiguity: The temporal difference between two location points should be less than a time threshold, indicating that they occurred close in time and may belong to the same crack propagation process. In the formula: t i , t j These are the fracture points. i and j The time of rupture, For a predefined time threshold; Spatial topological relationships: By calculating the geometric relationships between crack points, three topological relationships are defined: intersecting, adjacent, and disjoint; an adjacency relationship can only be established when two crack points intersect or are adjacent, indicating that they are spatially close enough to be connected and belong to the same crack. Among them, the calculation of crack points i Other fracture points j Spatial distance between d ij : In the formula: , These are the fracture points. i and j The spatial coordinates of the location; Spatial orientation similarity: If the spatial orientation vectors of two positioning points are similar within a certain error range, they are considered to belong to the same crack. The similarity of directions is determined by calculating the angle between the direction vectors. The expression for the angle between the direction vectors is as follows: In the formula, and These are the spatial orientation vectors of the two fracture points. The included angle error range is defined. Energy similarity: The energy difference between two locations should be less than a certain threshold, indicating that they have similar energy release patterns and can be physically connected or interoperable. The judgment expression is: In the formula: E i , E j These are the fracture points. i and j The energy released, This is a predefined exponential threshold.

6. The crack surface reconstruction method based on dynamic graph neural network according to claim 1, characterized in that, In step S30, clustering is performed using the DBSCAN algorithm based on a density threshold, including the following steps: S31: Feature Vector Reading and Euclidean Distance Calculation: Read the high-dimensional feature vector of each fracture point obtained through the fracture spatiotemporal topology analysis model, and calculate the Euclidean distance between fracture points based on the high-dimensional feature vector to measure the similarity between fracture points; wherein, for each pair of fracture points... and The Euclidean distance is calculated using the following formula: In the formula: n The dimension of the feature vector. x k and y k These are the first two high-dimensional eigenvectors of the two fracture points. k The value of the dimension; S32: Determine the neighborhood distance and minimum number of points: Select an appropriate neighborhood distance ϵ and a minimum number of points MinPts; Based on the Euclidean distance and the neighborhood distance ϵ, calculate the neighborhood of each fracture point. If the number of points contained in the neighborhood of a fracture point is greater than or equal to MinPts, then the fracture point is a core point; otherwise, it is a boundary point or a noise point. S33: Cluster Expansion and Core Point Processing: Select an initial point from the unlabeled core points, mark it as the starting point of a new cluster, and assign a cluster number to the point; expand the neighborhood of the core point, grouping all points within the neighborhood into the same cluster; for newly added points in the expanded cluster, check if they are core points; if they are core points, continue expanding the cluster; if they are boundary points, keep their cluster labels unchanged; continue selecting new points from the unlabeled core points for cluster expansion until all core points have been processed; S34: Cluster Numbering and Clustering Result Check: Based on the DBSCAN clustering results in step S33, assign a unique crack surface number to each cluster, with each cluster representing an independent crack surface. If a crack point belongs to the neighborhood of multiple core points, that point will be assigned the crack surface number of the first discovered cluster. Then check the clustering results to ensure that each cluster is assigned a unique crack surface number, with noise points numbered as -1. If any errors or omissions in cluster numbering are found, make necessary adjustments to ensure that the final crack surface numbering is accurate. S35: Optimization of clustering results based on magnitude: After obtaining the preliminary clustering results, the clustering results are optimized by combining the magnitude of each crack point. By visualizing the magnitude distribution of crack points within the crack surface, crack points with larger magnitudes are identified, which belong to the main crack surface; points with smaller magnitudes represent secondary crack surfaces or crack surfaces with less propagation potential.

7. The crack surface reconstruction method based on dynamic graph neural network according to claim 1, characterized in that, In step S40, the crack points of the main crack surface are reconstructed using a B-spline surface fitting algorithm to generate a smooth and continuous macroscopic crack surface, including the following steps: S41: From the results of cluster analysis, read all valid fracture points that have been classified into the same crack surface; where each fracture point includes fracture spatial coordinates, fracture volume, fracture spatial orientation, fracture energy and fracture magnitude; S42: Combining energy and normal direction to comprehensively describe the spatial characteristics and fracture effects of fracture points; Suppose the normal direction vector at a certain crack point is The formula for calculating the magnitude of this vector is as follows: The normal vector is normalized to ensure that the normalized normal vector only represents direction. The normalization process is calculated using the following formula: For each fracture point, a comprehensive characteristic weight is calculated based on its fracture energy and normal direction; assuming the fracture point... i Feature weights W i Since it is a combination of fracture energy and normal direction, the formula for calculating the comprehensive characteristic weight of each fracture point is as follows: In the formula, It is the normalized magnitude of the normal direction; E i For the first i The rupture energy at each fracture point; and These are the weighting coefficients; S43: Based on the feature weights and three-dimensional spatial location information of fracture points in the same cluster, fracture points with high feature weights are extracted through principal component analysis as key control points for fracture surface reconstruction, thereby clarifying the main direction of the fracture surface; S44: After determining the main direction of the crack surface, the boundary of the same cluster of crack points is fitted using the B-spline curve fitting method to form a closed-loop curve as the outer contour line of the final crack surface. S45: Triangulate the outer contour of the formed crack surface and divide the interior of the final crack surface into triangular meshes. S46: Read all the crack points in the same cluster, and combine them with the established triangular mesh to perform the final fitting on the crack surface to be generated, forming the final macroscopic crack surface of the crack point in the same cluster.

8. A fracture surface reconstruction device based on a dynamic graph neural network, characterized in that, include: The parameter acquisition module is configured to acquire raw acoustic wave data through the microseismic monitoring system, perform location calculations and source mechanism inversion on the raw acoustic wave data, and obtain the core parameters of the coal and rock fracture source that are directly related to the fracture. The model building module is configured to preprocess the core parameters of the fracture source using a standardized processing method, construct a spatiotemporal topology analysis model of the fracture based on the DGCNN graph neural network, and treat the fracture as a graph node; using the edge convolution mechanism, an adjacency graph is constructed through the spatiotemporal relationship between nodes, and the node representation is updated by aggregating neighborhood features to capture the local structure and global topology; The standardized processing method is used to preprocess the core parameters of the rupture origin, specifically including the following: Assuming the core parameters of the pre-treated fracture origin are x 1, x 2, ..., x n-1 , x n It is mapped to an interval using the min-max normalization method. Its standardized formula is as follows: In the formula, This is the minimum value of the preprocessing parameters. This is the maximum value of the preprocessing parameters. y i ∈[0,1]; The spatiotemporal topology analysis model for the crack includes a spatial transformation module, an edge convolution module, a max pooling layer, and a fully connected layer. The spatial transformation module is configured to normalize and align the input point cloud data, and learns an affine transformation matrix to convert the input point cloud data into a normalized coordinate system. The edge convolution module is configured to perform convolution operations on point cloud data to extract features and model local structural information, and to dynamically learn and extract local structural features. The maximum pooling layer is configured to aggregate the features of the crack points and extract the most representative local information. The fully connected layer is configured to map global features to a higher-dimensional space in order to capture complex nonlinear patterns and potential relationships in the data; The crack surface clustering module is configured to calculate the similarity of crack features based on Euclidean distance, and use the DBSCAN algorithm to cluster cracks by density threshold, grouping crack points with similar features into the same cluster. Each cluster represents a crack surface and is assigned a unique number. The clustering results are optimized by combining crack magnitude to identify main crack surface cracks with high propagation potential. The crack surface reconstruction module is configured to reconstruct the main crack surface using a B-spline surface fitting algorithm to generate a smooth and continuous macroscopic crack surface. The main direction of the crack is determined by selecting control points through principal component analysis. Based on the distribution of crack points, a B-spline curve is constructed to generate the boundary contour. The outer contour line is triangulated and meshed. Combining the crack points and the triangular mesh, the B-spline algorithm is used to accurately fit the crack surface.

Images