Method for identifying defects of underground energy storage space structure

Abstract

The invention discloses a method for identifying defects of an underground energy storage space structure, and the method comprises the steps: obtaining the point cloud data of the underground energy storage space structure, and constructing a three-dimensional point cloud model based on the point cloud data; performing feature extraction on the three-dimensional point cloud model to obtain a crack region; calculating the crack area to obtain occurrence information and a group system of the crack; carrying out refinement processing on the crack region through a three-dimensional morphological analysis method to obtain a crack skeleton line, and carrying out length calculation on the skeleton line to obtain a crack continuity parameter; according to the crack skeleton line, carrying out width calculation on the crack region to obtain width distribution characteristics; obtaining filling corrosion characteristics according to the point cloud reflection intensity information of the crack area; evaluation is carried out according to occurrence information and groups of the cracks, crack continuity parameters, width distribution characteristics, filling dissolution characteristics and density information, a comprehensive score is obtained, and an underground energy storage space structural body defect degree identification result is obtained according to the comprehensive score.

Description

Technical Field

[0001] The present invention belongs to the technical field of structural defect identification, and in particular relates to a method for identifying structural defects in underground energy storage spaces. Background Art

[0002] Various types of crack defects exist within underground energy storage structures, necessitating accurate identification and characterization of their geometric parameters and properties. Conventional methods currently struggle to achieve high-resolution characterization of crack organization, width, density, occurrence, filling and dissolution conditions, and continuity. The incompleteness and inaccuracy of crack parameters hinder subsequent defect analysis and stability assessment. A technical approach that can accurately characterize and quantitatively characterize crack defects is urgently needed to provide a reliable basis for the safe operation of underground energy storage systems. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes a method for identifying defects in underground energy storage space structures to solve the problems existing in the above-mentioned prior art.

[0004] To achieve the above objectives, the present invention provides a method for identifying defects in underground energy storage space structures, comprising:

[0005] Acquire point cloud data of the underground energy storage space structure, and construct a three-dimensional point cloud model based on the point cloud data;

[0006] Performing feature extraction on the three-dimensional point cloud model to obtain a crack area;

[0007] Calculating the fracture area to obtain fracture occurrence information and organization;

[0008] The crack region is refined by a three-dimensional morphological analysis method to obtain a crack skeleton line, and the length of the skeleton line is calculated to obtain a crack continuity parameter;

[0009] Calculating the width of the crack region according to the crack skeleton line to obtain a width distribution feature;

[0010] Obtaining filling and dissolution characteristics based on point cloud reflection intensity information of the fracture area;

[0011] An evaluation is performed based on the occurrence information and group system of the cracks, crack continuity parameters, width distribution characteristics, filling and dissolution characteristics, and density information to obtain a comprehensive score, and an identification result of the degree of defects in the underground energy storage space structure is obtained based on the comprehensive score.

[0012] Optionally, the process of constructing a 3D point cloud model includes:

[0013] Preprocessing the point cloud data, wherein the preprocessing process includes denoising;

[0014] The pre-processed point cloud data is registered by the point cloud data registration algorithm, and the registered point cloud data is fused. The surface of the fused point cloud data is reconstructed by the triangulated meshing algorithm to obtain a three-dimensional point cloud model.

[0015] Optionally, the process of obtaining the crack area includes:

[0016] The three-dimensional point cloud model is processed and analyzed using a semantic segmentation algorithm based on deep learning to obtain a crack area, wherein the crack area includes different point cloud data containing crack categories, and the location information and range information of the crack area are obtained according to the point cloud distribution of the crack area.

[0017] Optionally, the process of obtaining the occurrence information and system information of the fractures includes:

[0018] The crack area is preprocessed, and the point cloud in the crack area is segmented by a region growing algorithm to obtain a crack surface point cloud subset. The normal vector of the crack surface point cloud subset and each crack area in the crack surface point cloud subset is calculated by a principal component analysis method. The inclination and dip of the crack surface are obtained according to the direction and angle of the normal vector. Cluster analysis is performed based on the inclination and dip of the crack surface to obtain group information, wherein the occurrence information includes the inclination and dip of the crack surface.

[0019] Optionally, morphological filtering is performed on the crack area using a three-dimensional morphological analysis method, and skeleton extraction and skeleton refinement are performed on the filtered crack area in sequence using a skeleton extraction algorithm and a skeleton refinement algorithm to obtain a crack skeleton line. A threshold judgment of the number of pixel points is performed on the crack skeleton line to obtain a continuity parameter, and the length parameter of the crack skeleton line is calculated based on the continuity parameter.

[0020] Optionally, the process of obtaining the width distribution feature includes:

[0021] Local point cloud data of the crack is obtained according to the crack skeleton line, and plane fitting is performed on the local point cloud data of the crack by the least squares method to obtain a fitting plane. Distance information is obtained according to the points on the fitting plane and the crack skeleton line. The distance field distribution of the crack direction is obtained according to the distance information. The width change trend is obtained according to the distance field distribution. The width change trend is judged. Based on the judgment result, high-density point cloud data of the abnormal area is obtained by adopting a local encrypted sampling method. The high-density point cloud data of the abnormal area is calculated to obtain width change information, i.e., width distribution characteristics.

[0022] Optionally, the process of acquiring the filling and dissolution characteristics includes:

[0023] The point cloud reflection intensity information of the crack area is extracted, and the point cloud reflection intensity information is split and identified using a support vector machine algorithm to obtain the type of filling material inside the crack. Based on the changing trend of the point cloud reflection intensity, the degree of dissolution of the filling material is obtained. Based on the type and degree of dissolution of the filling material inside the crack, the filling dissolution characteristics are obtained.

[0024] Optionally, the process of obtaining the comprehensive score includes:

[0025] The occurrence information and group system, fracture continuity parameters, width distribution characteristics, filling and dissolution characteristics and density information are weighted and calculated to obtain a comprehensive score, and the comprehensive score is graded and evaluated to obtain the defect degree identification result of the underground energy storage space structure.

[0026] On the other hand, the present invention also provides a system for identifying defects in underground energy storage space structures, which is used to execute the above method.

[0027] Compared with the prior art, the present invention has the following advantages and technical effects:

[0028] The present invention discloses a method for identifying defects in underground energy storage space structures. The method first uses high-precision three-dimensional laser scanning to obtain point cloud data inside the underground energy storage space structure, and obtains a high-quality three-dimensional point cloud model through preprocessing. Then, the semantic segmentation algorithm of deep learning is used to automatically identify and extract the crack area to obtain the precise position and range of the crack. On this basis, the crack occurrence is judged by point cloud density analysis and normal vector calculation, and the crack skeleton line is extracted and the continuity parameter is calculated by three-dimensional morphological analysis. The crack width distribution characteristics are obtained by combining local fitting of the point cloud. Finally, the type of crack filling material and the degree of dissolution are analyzed based on the point cloud reflection intensity information. The present invention summarizes the crack parameters obtained by the above analysis, constructs a three-dimensional parameterized model of the crack defects in the underground energy storage space, realizes the fine characterization and quantitative characterization of the crack defects, and provides an important basis for the safety assessment and optimization design of the underground energy storage space. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0030] Figure 1 This is a flow chart of a method for identifying defects in underground energy storage space structures according to an embodiment of the present invention;

[0031] Figure 2 This is a structural diagram of a system for identifying defects in underground energy storage space structures according to an embodiment of the present invention. DETAILED DESCRIPTION

[0032] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0033] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0034] like Figure 1 In this embodiment, a method and system for identifying structural defects in underground energy storage spaces may specifically include:

[0035] S101. Use high-precision 3D laser scanning technology to obtain 3D point cloud data inside the underground energy storage space structure. Preprocess the point cloud data to remove noise points and obtain a high-quality 3D point cloud model.

[0036] Acquire high-precision three-dimensional point cloud data from the interior of the underground energy storage space structure and pre-process the acquired raw point cloud data. Remove noise points from the point cloud data using a point cloud data filtering algorithm to obtain point cloud data after noise point removal. Based on the point cloud data after noise point removal, use a point cloud data registration algorithm to achieve registration of multiple sets of point cloud data obtained by scanning from different perspectives. Fuse the multiple sets of registered point cloud data to obtain complete three-dimensional point cloud model data. Perform surface reconstruction on the three-dimensional point cloud model data and generate a smooth and continuous three-dimensional model using a triangulated meshing algorithm. Based on the reconstructed three-dimensional model, extract the geometric parameter information of the underground energy storage space structure, including key parameters such as volume and surface area. Visualize the extracted geometric parameter information to generate an intuitive and interactive three-dimensional model for easy analysis and application by engineers.

[0037] For example, obtaining high-precision 3D point cloud data of the interior of underground energy storage structures is a crucial step in geological engineering. This is typically accomplished using equipment such as laser scanners or depth cameras. For example, in a certain underground gas storage project, engineers used a high-precision laser scanner to scan multiple locations, acquiring millions of 3D coordinate points. Raw point cloud data often contains noise and requires preprocessing. Common filtering algorithms include statistical filtering and radius filtering. For example, statistical filtering can be used to set a threshold to remove outliers. In the aforementioned gas storage project, engineers set a threshold of 2 times the standard deviation, successfully removing over 99% of noise points and retaining approximately 5 million valid points. Point cloud registration is the process of aligning point cloud data obtained from different perspectives. A common algorithm is the Iterative Closest Point (ICP) algorithm. In practical applications, coarse registration can be performed followed by fine registration. For example, in a certain underground heat storage project, engineers first used feature point matching for coarse registration, followed by fine registration using the ICP algorithm. Ultimately, five sets of point cloud data were precisely aligned, with an error level within millimeter levels. Point cloud fusion is the process of combining multiple sets of registered point cloud data into a complete model. Common methods include voxel grid fusion and KD tree fusion. In an underground hydrogen storage project, engineers used voxel grid fusion to fuse eight sets of point cloud data into a complete model containing approximately 20 million points, accurately recreating the complex cave structure. Surface reconstruction of a 3D point cloud model converts discrete points into a continuous surface. Common algorithms include Poisson reconstruction and greedy projection triangulation. In an underground energy storage project, engineers used an improved Poisson reconstruction algorithm to successfully convert point cloud data into a smooth, continuous surface model containing approximately 1 million triangles, accurately reflecting the geometry of the energy storage space. Extracting geometric parameters from the reconstructed 3D model is a crucial basis for engineering design. Common methods include voxelization and surface integration. Visualization of 3D models is crucial for engineering analysis. Common technologies include OpenGL rendering and WebGL online display. In a certain underground energy storage project, engineers developed a WebGL-based 3D visualization platform that supports interactive operations such as model rotation, scaling, and sectioning, greatly facilitating collaborative design among multiple departments.

[0038] S102. Based on the preprocessed three-dimensional point cloud model, a semantic segmentation algorithm based on deep learning is used to automatically identify and extract crack areas in the point cloud model to obtain location and range information of the crack areas.

[0039] Based on the preprocessed 3D point cloud model data, a semantic segmentation algorithm based on deep learning is used to process and analyze the point cloud model. The semantic segmentation algorithm automatically identifies crack areas in the point cloud model, determines the category of each point in the point cloud, and distinguishes between crack and non-crack areas. For the identified crack area point cloud, its 3D spatial coordinate information is further extracted to obtain the precise location of the crack area. Based on the spatial distribution of the crack area point cloud, the boundary and range of the crack are determined, and complete crack area range information is obtained. The location and range information of the crack area is visualized to generate a 3D model of the crack area, which intuitively displays the spatial distribution of the crack. The extracted crack area information is quantitatively analyzed to calculate the geometric parameters of the crack, such as length, width, and depth, and comprehensively assess the severity of the crack. The analysis results of the crack area are fused with the point cloud model to generate a 3D model containing crack information, providing data support for subsequent crack detection and assessment.

[0040] For example, deep learning-based semantic segmentation algorithms are key technologies in point cloud processing, automatically identifying and classifying different regions within a point cloud. In crack detection within underground energy storage structures, this algorithm can effectively distinguish between cracked and non-cracked areas. For example, an improved three-dimensional convolutional neural network model is used to extract local and global features of the point cloud through multi-layer convolution and pooling operations. Finally, a fully connected layer and softmax classifier are used to predict the category of each point. In practical applications, the model can be trained using a manually annotated dataset and then applied to new point cloud data.

[0041] For identified crack regions, their precise 3D spatial coordinates need to be extracted. This can be achieved by calculating the centroid coordinates of the crack region's point cloud. For example, if a crack approximately 2 meters long is identified, it can be divided into several small segments, with the centroid coordinates calculated for each segment, thus obtaining a spatial distribution curve for the crack.

[0042] Determining the boundaries and extent of cracks is a crucial step in assessing crack severity. A region growing algorithm can be used, starting with a seed point in the crack region and gradually expanding to neighboring points that meet the criteria until further expansion is impossible. This method yields a complete outline of the crack region. For example, for a crack with varying width, different thresholds can be set to capture variations in crack width. Visualizing cracks is crucial for engineers to intuitively understand crack conditions. 3D graphics libraries such as OpenGL can be used to render 3D models of the crack region. For example, different colors can be used to represent crack depth, with red representing deeper areas and green representing shallower areas. This provides a visual representation of the crack's spatial distribution. Quantitative analysis of cracks involves calculating their geometric parameters. Length can be calculated by accumulating distance from the crack centerline, width can be measured perpendicular to the centerline, and depth can be estimated by the height difference between the crack region point cloud and the surrounding non-cracked areas. For example, for a crack 1 meter long, with an average width of 5 cm and a maximum depth of 2 cm, the width and depth can be calculated at 10 cm intervals along its length, thereby obtaining detailed crack geometry. Finally, the crack analysis results are integrated with the original point cloud model to generate a complete 3D model containing crack information. This model not only captures the overall structure of the underground energy storage space but also identifies the location, extent, and severity of the cracks. For example, cracked areas can be marked with a specific texture or color within the model, along with information about the crack's geometric parameters. This comprehensive model provides comprehensive data support for engineers, helping them develop appropriate repair and reinforcement plans to ensure the safe operation of the underground energy storage space.

[0043] The process of further calculating the geometric parameters of the identified crack area point cloud includes: extracting its three-dimensional spatial coordinate information and obtaining the precise location of the crack area.

[0044] For the identified crack area point cloud, a subset of candidate crack area point clouds is obtained; for each candidate crack area point cloud subset, its three-dimensional spatial coordinate information is extracted to obtain the three-dimensional bounding box of the subset point cloud; based on the three-dimensional bounding box of each subset point cloud, its three-dimensional spatial center point coordinates are calculated as the precise position of the crack area; if there are multiple crack areas, the center point coordinates of each crack area are summarized to obtain complete crack location distribution information; based on the crack location distribution information, a spatial clustering algorithm is used to cluster and group the crack positions to obtain category identifications of different cracks; for each crack category, its three-dimensional size features, including length, width, depth, etc., are extracted for subsequent crack quantitative analysis; the precise position, category identification and size features of each crack are associated and stored to construct a three-dimensional information database of crack areas.

[0045] For example, obtaining a 3D bounding box for a candidate crack region is essential for determining crack location and size. Taking the minimum bounding box algorithm as an example, this method calculates the smallest cube that encompasses all points. For a crack measuring 10 cm long, 0.5 cm wide, and 2 cm deep, its bounding box dimensions might be 11 × 1 × 3 cm, slightly larger than the actual size to encompass all points. Calculating the center point of the bounding box as the precise crack location is crucial. For the crack described above, the center point coordinates might be (5.5, 0.5, 1.5) cm. This representation is concise and convenient, facilitating subsequent analysis and visualization. When multiple cracks are present, summarizing the coordinates of all center points can intuitively reflect the spatial distribution pattern of the cracks. Spatial clustering algorithms are used to distinguish between different cracks. For example, K-means clustering can classify crack center points into K clusters. Suppose 10 crack regions are detected on a wall. Clustering might categorize them into three types: vertical, horizontal, and diagonal. This classification helps analyze the causes and development trends of cracks. Extracting 3D dimensional features of cracks is crucial for assessing their severity. The length can be estimated by calculating the diagonal length of the bounding box, while the width and depth can be directly calculated using the corresponding dimensions of the bounding box. For example, the characteristics of the crack described above can be expressed as: length 10.5 cm, width 1 cm, depth 3 cm. This data can be used to determine whether the crack exceeds the safety threshold. Building a three-dimensional information database of crack areas is the foundation for intelligent monitoring. The database can contain information such as each crack's ID, location coordinates, category identifier, and dimensional characteristics. For example, a crack with ID 001 might have the following record: location (5.5, 0.5, 1.5), category "vertical crack", length 10.5 cm, width 1 cm, depth 3 cm. This structured storage facilitates subsequent data analysis and visualization, providing a basis for formulating maintenance strategies.

[0046] Based on the above geometric parameters, the following specific method is proposed: quantitative analysis is performed on the extracted crack area information, geometric parameters such as crack length, width, and depth are calculated, and the severity of the crack is comprehensively evaluated.

[0047] The input crack image is processed using an image segmentation algorithm to extract pixel coordinates of the crack area, generating a binary mask of the crack region. Based on the binary mask of the crack region, a contour extraction algorithm is used to obtain the contour coordinates of the crack region. The perimeter of the contour is calculated to obtain the crack length parameter. A skeleton extraction algorithm is used on the binary mask of the crack region to obtain the skeleton lines of the crack region and the average width of the skeleton lines is calculated to obtain the crack width parameter. Based on the binary mask of the crack region, a deep learning semantic segmentation model is used to estimate the depth of the crack region at the pixel level, generating a depth map of the crack region. Statistical analysis of the depth map is performed to calculate the average and maximum depths of the crack region to obtain the crack depth parameter. Based on the preset crack severity assessment criteria, the crack length, width, and depth parameters are integrated to determine the crack severity and obtain a quantitative crack assessment result. The crack geometric parameters and severity assessment results are visualized to generate a crack analysis report, providing a reference for subsequent crack treatment and repair decisions.

[0048] For example, image segmentation is a key step in crack detection. By segmenting an image into crack and background regions, crack locations can be precisely located. Common segmentation algorithms include threshold segmentation, edge detection, and region growing. For example, using threshold segmentation, an appropriate threshold is selected based on the image's grayscale histogram. Pixels with grayscale values ​​below the threshold are identified as crack regions, generating a binary mask. Contour extraction is an important method for capturing crack geometric features. Based on the binary mask, a boundary tracing algorithm can be used to extract the crack contour. Starting from a boundary point, the algorithm gradually traces along the crack edge until it returns to the starting point, resulting in a complete contour. By calculating and accumulating the Euclidean distances between contour points, the crack length can be determined. For example, if a crack contour consists of 100 points, accumulating the distances between adjacent points yields a total length of 500 mm. Skeleton extraction can be used to determine the crack centerline for width measurement. Common skeleton extraction algorithms include distance transform and thinning algorithms. For example, the thinning algorithm repeatedly removes boundary points until no further deletions are possible, resulting in a single-pixel-wide skeleton line. The crack width is measured perpendicular to the skeleton line, and the average value is used as the crack width parameter. For example, a crack skeleton has a length of 400 mm and an average width of 2 mm. Depth estimation is key to obtaining three-dimensional crack information. Deep learning-based semantic segmentation models, such as fully convolutional networks, can achieve pixel-level depth prediction. The model inputs a crack image and outputs a depth value for each pixel, generating a depth map. Statistical analysis of the depth map can determine the average and maximum depth of the cracks. For example, the average depth of a cracked area is 5 mm, and the maximum depth is 10 mm. Crack severity assessment requires comprehensive consideration of parameters such as length, width, and depth. Assessment criteria can be established, such as categorizing cracks into three levels: minor, moderate, and severe. For example, in concrete structures, cracks with a length of less than 100 mm, a width of less than 0.3 mm, and a depth of less than 3 mm are considered minor; cracks between 100 and 300 mm in length, between 0.3 and 0.5 mm in width, and between 3 and 5 mm in depth are considered moderate; and cracks outside these ranges are considered severe. Visualization and report generation provide intuitive presentation of analysis results. Pseudo-color images can be used to display crack depth distribution, with different colors representing different depth ranges. At the same time, an analysis report containing crack geometry parameters, depth information, and severity assessment results can be generated. This information provides an important basis for structural repair and reinforcement, helping to formulate reasonable treatment plans and improve structural safety and service life.

[0049] S103. In the fracture area, determine the fracture occurrence information, including the inclination and dip of the fracture surface, and determine the spatial distribution characteristics of the fracture system through point cloud density analysis and normal vector calculation.

[0050] Acquire three-dimensional point cloud data within the fracture area and preprocess the point cloud data to remove noise points and outliers to improve point cloud quality. Based on changes in point cloud density, segment the point cloud using a region growing algorithm to extract a subset of the fracture surface point cloud. For the fracture surface point cloud subset, calculate the point cloud normal vector using principal component analysis to obtain spatial posture information of the fracture surface. Determine the inclination and dip of the fracture surface based on the direction and angle of the normal vector, and store the inclination and dip data in an attribute table. Perform cluster analysis on the inclination and dip data of the fracture surface, and use the K-means algorithm to divide the fracture surface into different groups. Analyze the spatial distribution characteristics of each fracture group, calculate the average inclination and dip of the fracture surface within the group, and determine the dominant direction of the group. Fuse the spatial distribution information of the fracture group with the three-dimensional geological model to generate a three-dimensional visualization model of the fracture group, visually displaying the spatial distribution patterns of the fracture group.

[0051] For example, after obtaining the three-dimensional point cloud data in the crack area, the point cloud needs to be preprocessed to improve the quality. Preprocessing includes removing noise points and outliers, and methods such as statistical outlier filtering can be used. Next, the region growing algorithm is used to segment the point cloud and extract the subset of the crack surface point cloud. The region growing algorithm starts from the seed point and gradually adds adjacent points that meet the growth conditions to the region until it can no longer grow. For the crack surface, growth conditions such as the normal vector angle and curvature can be set. In this way, a complex crack network can be divided into multiple independent crack surfaces. For the extracted subset of the crack surface point cloud, the principal component analysis method is used to calculate the normal vector of the point cloud to obtain the spatial posture information of the crack surface. Principal component analysis determines the direction of the normal vector by calculating the eigenvector of the point cloud covariance matrix. For example, for a crack surface point cloud that is approximately planar, the direction of its third principal component is the direction of the normal vector. According to the direction and angle of the normal vector, the inclination and dip of the crack surface can be determined. Dip refers to the projection direction of the normal vector onto the horizontal plane, usually expressed as 0° to 360°; the inclination refers to the angle between the normal vector and the vertical direction, ranging from 0° to 90°. For example, if the normal vector direction of a fracture surface is (0.5, 0.866, 0.5), it can be calculated that its dip is approximately 60° and its inclination is approximately 60°. These data are stored in the attribute table for subsequent analysis. Cluster analysis of the dip and inclination data of the fracture surface can be performed using the K-means algorithm to divide the fracture surface into different groups. The K-means algorithm divides data points into the nearest cluster center through iterative optimization. For example, for a data set containing 1,000 fracture surfaces, 3 to 5 major fracture groups may be identified. For each fracture group, its spatial distribution characteristics are analyzed, the average dip and inclination of the fracture surfaces within the group are calculated, and the dominant direction of the group is determined. For example, the average dip of a certain fracture system is 235° and the average inclination is 65°, indicating that the fracture plane of this system is generally southwest-dipping and has a steep inclination. This analysis helps understand the regional stress field and tectonic evolution history. Finally, the spatial distribution information of the fracture system is integrated with the 3D geological model to generate a 3D visualization model of the fracture system. This visualization model can intuitively display the spatial distribution patterns of the fracture system, helping geological engineers assess rock mass stability and predict groundwater flow paths.

[0052] S104: Using a three-dimensional morphological analysis method, the extracted crack region is refined to obtain a crack skeleton line, and the length of the crack skeleton line is calculated to obtain a crack continuity parameter.

[0053] A three-dimensional morphological analysis method is used to perform morphological filtering on the obtained crack area to remove image noise. A skeleton extraction algorithm is used to extract the skeleton of the filtered crack area to obtain the skeleton line of the crack. The extracted crack skeleton line is refined according to the skeleton refinement algorithm to obtain a refined crack skeleton line. If the number of pixels in the crack skeleton line is less than a preset threshold, the crack is judged to be a discontinuous crack, otherwise it is judged to be a continuous crack. The number of pixels in the refined crack skeleton line is calculated using a length calculation algorithm to obtain the length parameter of the crack. The crack length parameter is combined with the continuity parameter as the basis for crack detection, and the final crack detection result is output.

[0054] For example, a skeleton line is extracted from the filtered crack region using a skeleton extraction algorithm. Skeleton extraction is a technique that refines a binary image into lines with a single pixel width, which can reflect the topological structure of the target. For example, a skeleton extraction algorithm based on distance transformation can be used. This algorithm first calculates the distance from each pixel in the crack region to the background, and then gradually refines the crack region based on the distance value until a skeleton line with a single pixel width is obtained. The purpose of this step is to simplify the crack region into a single line, which facilitates subsequent length calculation and continuity judgment. For example, for a crack with a width of 5 pixels, a skeleton extraction algorithm can obtain a skeleton line with a single pixel width, which can accurately reflect the direction and length of the crack. Next, the extracted crack skeleton line is refined using a skeleton refinement algorithm. The purpose of this step is to remove burrs and short branches in the skeleton line to obtain a more refined crack skeleton line. For example, an algorithm based on morphological refinement can be used to iteratively remove pixels at the endpoints of the skeleton line until the skeleton line no longer changes. Suppose the skeleton line obtained in step 3 contains some short branches of one or two pixels in length. These short branches may be caused by noise or irregularities at the crack edge. A thinning algorithm can remove these short branches, resulting in a smoother and more accurate skeleton line. Next, the continuity of the crack is determined based on the number of pixels in the skeleton line. If the number of pixels is less than a preset threshold, the crack is considered discontinuous; otherwise, it is considered continuous. For example, a threshold of 10 pixels can be set. If a crack's skeleton line has only 5 pixels, the crack is considered discontinuous; if it has 20 pixels, the crack is considered continuous. The purpose of this step is to distinguish the severity of the crack; continuous cracks are generally more dangerous than discontinuous cracks. Subsequently, a length calculation algorithm is used to calculate the number of pixels in the refined crack skeleton line to obtain the crack length parameter. For example, the number of pixels in the skeleton line can be directly counted, or the length of the skeleton line can be calculated based on the distance between pixels. Suppose a crack's skeleton line consists of 50 pixels, and the distance between each pixel is 1 unit length. Therefore, the length of the crack is 50 units length. Finally, the length parameter of the crack is combined with the continuity parameter as the basis for crack detection, and the final crack detection result is output. For example, cracks that are longer than 100 units and continuous can be marked as severe cracks, and cracks that are less than 50 units and non-continuous can be marked as minor cracks. The purpose of this step is to comprehensively consider the length and continuity of the cracks, evaluate the severity of the cracks, and provide a basis for subsequent maintenance and repair. For example, during the inspection of a bridge, a crack with a length of 150 units and continuous is detected. It can be marked as a severe crack and needs to be treated first; while another crack with a length of 30 units and non-continuous can be marked as a minor crack and can be temporarily ignored.These judgments are based on the analysis of a large amount of actual data and experience summary, which is very necessary.

[0055] In the above content, the extracted crack skeleton line is refined according to the skeleton refinement algorithm to obtain a refined crack skeleton line.

[0056] Based on the crack distribution information, an edge extraction algorithm is used to detect the edges of the crack area and obtain the crack contour edge data. Skeleton extraction is performed on the extracted crack contour edge data, and a skeleton line representation of the crack is obtained using the skeleton extraction algorithm. The extracted crack skeleton line is refined using a skeleton refinement algorithm to eliminate redundant pixels in the skeleton line and obtain a refined crack skeleton line. Morphological analysis is performed on the refined crack skeleton line to calculate geometric characteristic parameters such as the skeleton line's length, direction, and connectivity. Based on the geometric characteristic parameters of the crack skeleton line, the severity and distribution of the crack are determined, and the crack damage level of the concrete surface is determined. The crack damage level is compared with a preset threshold. If the threshold is exceeded, the concrete surface is considered to have severe cracks and require repair and reinforcement. Otherwise, the crack is considered to be within an acceptable range and regular monitoring continues.

[0057] For example, after obtaining a preliminary skeleton, refinement is required to eliminate redundant pixels. The core of the skeleton refinement algorithm is to iteratively remove boundary points beyond the endpoints until no further deletion is possible. The Zhang-Suen refinement algorithm is a classic parallel refinement algorithm that achieves rapid skeleton refinement through two iterative scans. This algorithm effectively removes bifurcations and burrs, resulting in a smooth and continuous crack skeleton line. Morphological analysis of the refined skeleton line can be performed to extract the geometric characteristic parameters of the crack. The length parameter is obtained by summing the number of pixels along the skeleton line. The direction parameter can be obtained using principal component analysis to calculate the covariance matrix of the skeleton line pixels. The eigenvector corresponding to the maximum eigenvalue is the main direction of the crack. Connectivity parameters can be obtained by analyzing the topological structure of the skeleton line, such as the number of endpoints and bifurcations. Based on the extracted geometric features, a quantitative assessment of the crack can be performed. For example, the crack length can be compared with a preset threshold. Cracks longer than 10 cm may affect the safety of the structure. When the crack direction aligns with the force direction, it often indicates that the structure is experiencing excessive tensile stress. When multiple cracks interconnect to form a network, the structure may have severe internal damage. For example, consider a crack measuring 15 cm in length and 0.2 mm in width. This length exceeds the 10 cm warning line, posing a significant risk. Analysis of the crack's orientation reveals that it is parallel to the beam's main reinforcement, suggesting a flexural crack likely caused by overloading. Further investigation of the crack's connectivity reveals that the crack is connected to several other smaller cracks, forming a "dendritic" pattern. This morphological characteristic often indicates a potential for further crack growth. Taking into account factors such as crack length, width, direction, and connectivity, the damage level of the crack is assessed as "moderate." While not yet reaching the "serious" level, it is beyond acceptable limits. Repair measures such as grouting and enhanced follow-up monitoring are recommended. Furthermore, the structure's bearing capacity should be reviewed, and reinforcement measures implemented if necessary. This crack detection method, based on image processing and morphological analysis, offers the advantages of high automation and precision. By extracting geometric characteristic parameters, quantitative crack assessment can be achieved, providing a scientific basis for structural health assessment and repair decisions. However, this method also has some limitations. For example, severe surface contamination or poor lighting conditions may affect the accuracy of crack identification. Furthermore, relying solely on surface crack characteristics makes it difficult to fully assess the internal damage of a structure. Therefore, in practical applications, it is often necessary to combine it with other nondestructive testing methods, such as ultrasonic testing and infrared thermal imaging, to obtain more comprehensive and reliable results.

[0058] S105. Based on the crack skeleton line, obtain the width variation information along the crack direction through local point cloud fitting and distance field calculation, and determine the distribution characteristics and statistical parameters of the crack width.

[0059] Based on the crack skeleton line, the local point cloud data of the crack is obtained, and the least squares method is used to perform plane fitting on the local point cloud to obtain the fitting plane equation. For each point on the fitting plane, the distance from the crack skeleton line is calculated to obtain the distance field distribution along the crack direction. By analyzing the distance field distribution, the width change trend along the crack direction is determined, and it is judged whether there is a sudden change or anomaly in the width change. If there is a sudden change or anomaly in the width change, a local encrypted sampling method is used to obtain high-density point cloud data in the abnormal area. Based on the high-density point cloud data in the abnormal area, local plane fitting and distance field calculation are re-performed to obtain refined width change information. The width change information along the crack direction is statistically analyzed to obtain the distribution characteristics of the crack width, including parameters such as average width, maximum width and minimum width. Based on the distribution characteristics and statistical parameters of the crack width, the severity of the crack is judged, providing a basis for subsequent crack treatment and repair.

[0060] For example, the crack skeleton is an important feature of crack morphology. By acquiring its local point cloud data, the geometric characteristics of the crack can be deeply analyzed. The least squares method is a commonly used plane fitting method that can effectively fit the local plane where the crack is located. For example, for a concrete surface crack approximately 2 meters long, a set of point cloud data can be collected every 10 centimeters along the crack skeleton line, with each set containing 100 points within a 5-centimeter radius. Fitting these point clouds using the least squares method yields a plane equation of the form Ax+By+Cz+D=0, where A, B, C, and D are the fitted coefficients. The distance field distribution reflects the variation in crack width. By calculating the perpendicular distance from each point on the fitted plane to the crack skeleton line, the distance field along the crack direction can be obtained. For example, for the 2-meter-long crack mentioned above, the distance can be calculated every 1 millimeter along the skeleton line, resulting in approximately 2,000 distance values. The changing trend of these distance values ​​directly reflects the variation in crack width. Sudden or abnormal changes in width may indicate local expansion or bifurcation of the crack. For example, if the maximum distance field value suddenly increases from 2 mm to 5 mm within a 10-centimeter interval, this may indicate crack expansion at that location. For such anomaly areas, a more intensive sampling approach can be employed, such as collecting a high-density point cloud with 500 points every 1 cm, to obtain more detailed geometric information. Re-performing the local plane fitting and distance field calculations can yield refined width variation information for the anomaly area. For example, for the aforementioned 10-centimeter interval, approximately 1,000 high-precision distance values ​​can be obtained. These values ​​more accurately reflect the geometric characteristics of the crack at that location, such as the presence of bifurcations and small step-like structures. Statistical analysis of the distribution of crack widths can provide a comprehensive assessment of the overall crack condition. For example, for the aforementioned 2-meter-long crack, the following statistical results might be obtained: average width 2.5 mm, maximum width 5 mm, minimum width 1 mm, and standard deviation 0.8 mm. These parameters comprehensively reflect the overall morphological characteristics of the crack. Based on these statistical parameters, a preliminary assessment of the crack's severity can be made. For example, if the maximum width exceeds 5 mm or the standard deviation exceeds 1 mm, it may indicate that the crack has reached a level requiring urgent attention. These assessments provide a crucial basis for subsequent crack treatment and repair, helping to develop appropriate repair plans, such as selecting appropriate grouting materials and pressure. This refined analysis method not only accurately describes the geometric characteristics of cracks but also promptly identifies potential anomalies, providing reliable technical support for structural safety assessment and maintenance. The advantage of this method lies in its ability to capture subtle changes that traditional visual inspections might overlook, enabling earlier detection of potential structural problems and improving the efficiency and accuracy of structural maintenance.

[0061] S106. Analyze the type and degree of dissolution of the filling material inside the crack based on the point cloud reflection intensity information of the crack area, and determine the qualitative characteristics of the crack filling dissolution.

[0062] Obtain point cloud data of the fracture area and extract its reflection intensity information. Based on this reflection intensity information, establish a filling material type identification model. Use a support vector machine algorithm to classify and identify the filling material type. Based on the classification results, determine the specific type of filling material within the fracture. Analyze the changing trends in the reflection intensity information to determine the degree of dissolution of the filling material. Dramatic changes in reflection intensity indicate a high degree of dissolution; gentle changes indicate a low degree of dissolution. By combining the analysis results of filling material type and dissolution degree, qualitative characteristics of fracture filling dissolution are derived.

[0063] For example, obtaining point cloud data of the crack area is the basis for crack filling and dissolution analysis. Point cloud data contains the geometry and reflection intensity information of the crack surface, reflecting the characteristics of the filling material inside the crack. For example, for a crack with a length of 5 meters, high-density point cloud data of about 100,000 points can be collected, and each point contains three-dimensional coordinates and reflection intensity values. Extracting point cloud reflection intensity information is a key step in the analysis. Reflection intensity is affected by factors such as material properties and surface roughness, and can be used to determine the type of filling material. For example, the reflection intensity of cement slurry is usually between 50-70, while the reflection intensity of lime slurry is between 30-50. By analyzing the distribution characteristics of reflection intensity, the type of filling material can be preliminarily determined. Establishing a filling material type judgment model is an important step in achieving automated identification. The support vector machine (SVM) algorithm is an effective classification method that can classify filling materials into different categories based on reflection intensity characteristics. For example, statistical features such as the mean, standard deviation, and skewness of reflection intensity can be selected as inputs to the SVM to train a multi-category classifier. In practical applications, model training can be performed using samples of known filling materials to improve classification accuracy. Classifying and identifying the filling material type is the application process of the judgment model. By inputting the extracted reflection intensity features into a trained SVM model, the specific filling material type can be determined. For example, if the reflection intensity of a certain crack segment has a mean of 60 and a standard deviation of 5, it can be classified as cement slurry after SVM classification. This automated identification method can quickly process large amounts of fracture data, improving analysis efficiency. Determining the specific type of filling material within the crack is a key analysis result. Different filling materials have different physical and chemical properties, which significantly affect the stability and durability of the crack. For example, cracks filled with cement slurry generally have better strength and durability, while cracks filled with lime slurry may be more susceptible to dissolution. Analyzing the changing trends of reflection intensity information can determine the degree of dissolution of the filling material. Dissolution causes the surface of the filling material to become rough, resulting in changes in reflection intensity. For example, for a 5-meter-long crack, reflection intensity values ​​can be collected every 10 centimeters and a curve plotted along the crack's direction. If the curve shows dramatic fluctuations, with fluctuations exceeding 20%, it may indicate that the filling material has undergone severe dissolution. Drastic changes in reflection intensity generally indicate a high degree of dissolution. For example, if the reflection intensity of a certain fracture section fluctuates rapidly between 50-70, and the intensity difference between adjacent sampling points can reach over 30%, it can be determined that the filling material has undergone significant dissolution, and there may be localized detachment or voids. This dissolution state requires high attention and may require prompt repair. Gentle changes in reflection intensity indicate a low degree of dissolution. For example, if the reflection intensity of a certain fracture section remains stable between 55-65, and the intensity difference between adjacent sampling points does not exceed 10%, it can be assumed that the filling material is well maintained and the degree of dissolution is low.Cracks in this state usually do not require immediate treatment, but still require regular monitoring. By combining the analysis results of the filling material type and the degree of dissolution, the qualitative characteristics of the crack filling and dissolution can be obtained. These characteristics can be used to assess the overall condition and potential risks of the cracks. For example, for a crack filled with cement slurry and with a low degree of dissolution, it can be judged that its condition is good and does not require special treatment in the short term. For a crack filled with lime slurry and with a high degree of dissolution, a repair plan needs to be formulated to prevent further expansion of the crack. Through this point cloud data-based crack filling and dissolution analysis method, the condition of a large number of cracks can be quickly and accurately evaluated, providing an important basis for engineering maintenance and risk management. The advantage of this method is that crack information can be obtained non-contact and non-destructively, and it is suitable for crack detection and evaluation in various complex environments.

[0064] S107. Summarize the parameters such as the fracture group, width, density, occurrence, filling and dissolution conditions, and continuity obtained from the above analysis to construct a three-dimensional parametric model of fracture defects in the underground energy storage space, thereby achieving a detailed description and quantitative characterization of the fracture defects.

[0065] A three-dimensional geometric model of fractures is constructed based on parameters such as fracture organization, width, density, occurrence, filling and dissolution, and continuity, enabling visualization of the spatial distribution of fractures. A three-dimensional spatial interpolation algorithm is used to grid the fracture parameters, generating a regular three-dimensional parameter field for continuous characterization of the fracture parameters. Based on the three-dimensional parameter field, key parameters such as fracture width, density, and continuity are extracted to construct a fracture defect evaluation index system to quantify the severity of fracture defects. Based on the fracture filling and dissolution, an image segmentation algorithm is used to extract the three-dimensional spatial distribution of fracture fillings and dissolution cavities and assess their impact on fracture conductivity. A comprehensive fracture defect evaluation model is constructed based on parameters such as fracture width, density, continuity, and filling and dissolution to determine the severity of fracture defects. The fracture defect evaluation results are correlated with the three-dimensional fracture model to generate a three-dimensional distribution map of the fracture defects, visually displaying the spatial distribution characteristics of the defects. Based on the fracture defect evaluation results and the three-dimensional distribution characteristics, the safety and stability of the underground energy storage space are determined, providing a decision-making basis for project construction and operation management.

[0066] For example, the construction of a three-dimensional geometric model of cracks is a key step in the safety assessment of underground energy storage space. First, it is necessary to collect various parameters of the cracks, such as group system, width, density, etc. Taking a certain karst area as an example, geological surveys have found that there are mainly three groups of cracks: NE-directed, NW-directed, and nearly horizontal. The average width of NE-directed cracks is 0.5-2 cm, and the density is 3-5 cracks / m; the width of NW-directed cracks is 0.3-1 cm, and the density is 2-4 cracks / m; the width of nearly horizontal cracks is 0.2-0.8 cm, and the density is 1-3 cracks / m. These parameters are gridded using a three-dimensional spatial interpolation algorithm to generate a continuous three-dimensional parameter field. Commonly used interpolation methods include Kriging and Inverse Distance Weighting. Taking Kriging as an example, the sampling point data can be extended to the entire study area to obtain a continuous distribution of parameters such as crack width and density. Based on the three-dimensional parameter field, key parameters are extracted to construct a crack defect evaluation index system. For example, crack widths can be categorized as tiny (<0.5 cm), medium (0.5-2 cm), and large (>2 cm), while density can be categorized as low (<2 cracks / m), medium (2-5 cracks / m), and high (>5 cracks / m). Continuity can be characterized by the length of the crack traced, with <1 m being considered low continuity, 1-5 m being medium continuity, and >5 m being high continuity. Image segmentation algorithms are used to extract the three-dimensional spatial distribution of fracture fillings and dissolution. For example, the region growing method can be used, starting with a known seed point of fillings or dissolution cavities and gradually expanding to similar regions, ultimately obtaining the three-dimensional distribution of fillings and dissolution cavities. This helps assess their impact on fracture conductivity. For example, a high degree of filling reduces conductivity, while the presence of dissolution cavities may enhance it. By integrating these parameters, a comprehensive evaluation model for fracture defects can be constructed. A weighted summation method can be used, such as setting a weight of 0.3 for width, 0.2 for density, 0.3 for continuity, and 0.2 for filling and dissolution. Each parameter is scored (1-5) to calculate a comprehensive score. For example, if a region has large crack width (4 points), medium density (3 points), high continuity (5 points), and low filling (4 points), the comprehensive score is 4 × 0.3 + 3 × 0.2 + 5 × 0.3 + 4 × 0.2 = 4.1, indicating a high degree of crack defects in this region. The evaluation results are correlated with the three-dimensional crack model to generate a three-dimensional distribution map of crack defects. A color gradient can be used to indicate the degree of defects, such as green for minor defects, yellow for moderate defects, and red for severe defects. This visualization method intuitively displays the spatial distribution characteristics of defects and helps identify high-risk areas. Finally, based on the evaluation results and distribution characteristics, the safety and stability of the underground energy storage space are determined. For example, if more than 80% of the area around a storage cavity has a low degree of crack defects (comprehensive score < 3), the cavity can be judged to have good safety and stability. Conversely, if there are large areas with high defects (comprehensive score > 4), reinforcement measures or adjustments to the energy storage solution are necessary.These analysis results provide important decision-making basis for engineering construction and operation management, helping to ensure the long-term safe operation of underground energy storage projects.

[0067] like Figure 2 As shown, the present invention provides a system for identifying defects in underground energy storage space structures, which mainly includes:

[0068] The point cloud data acquisition module is used to obtain 3D point cloud data of the interior of the underground energy storage space structure using high-precision 3D laser scanning technology. By preprocessing the point cloud data and removing noise points, a high-quality 3D point cloud model is obtained.

[0069] The crack area recognition module is used to automatically identify and extract crack areas from the pre-processed 3D point cloud model using a deep learning-based semantic segmentation algorithm to obtain the precise location and range information of the crack areas.

[0070] The fracture occurrence analysis module is used to determine the fracture occurrence information, including the inclination and dip of the fracture surface, and the spatial distribution characteristics of the fracture system through point cloud density analysis and normal vector calculation within the fracture area;

[0071] The crack continuity analysis module is used to refine the extracted crack area using a three-dimensional morphological analysis method, obtain the crack skeleton line, calculate the length of the crack skeleton line, and obtain the crack continuity parameter;

[0072] The crack width analysis module is used to obtain the width variation information along the crack direction based on the crack skeleton line through local point cloud fitting and distance field calculation, and determine the distribution characteristics and statistical parameters of the crack width;

[0073] The fracture filling and dissolution analysis module is used to analyze the type and degree of dissolution of the filling material inside the fracture based on the point cloud reflection intensity information of the fracture area, and to determine the qualitative characteristics of the fracture filling and dissolution;

[0074] The fracture parametric modeling module is used to summarize the parameters such as fracture group, width, density, occurrence, filling and dissolution conditions, and continuity obtained from the above analysis, to construct a three-dimensional parametric model of fracture defects in underground energy storage space, and to achieve detailed characterization and quantitative representation of fracture defects.

[0075] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for identifying defects in underground energy storage space structures, characterized in that: include: Acquire point cloud data of the underground energy storage space structure, and construct a three-dimensional point cloud model based on the point cloud data; Performing feature extraction on the three-dimensional point cloud model to obtain a crack area; Calculating the fracture area to obtain fracture occurrence information and organization; The crack region is refined by a three-dimensional morphological analysis method to obtain a crack skeleton line, and the length of the skeleton line is calculated to obtain a crack continuity parameter; Calculating the width of the crack region according to the crack skeleton line to obtain a width distribution feature; Obtaining filling and dissolution characteristics based on point cloud reflection intensity information of the fracture area; An evaluation is performed based on the occurrence information and group system of the cracks, crack continuity parameters, width distribution characteristics, filling and dissolution characteristics, and density information to obtain a comprehensive score, and an identification result of the degree of defects in the underground energy storage space structure is obtained based on the comprehensive score.

2. The method according to claim 1, characterized in that The process of building a 3D point cloud model includes: Preprocessing the point cloud data, wherein the preprocessing process includes denoising; The pre-processed point cloud data is registered by the point cloud data registration algorithm, and the registered point cloud data is fused. The surface of the fused point cloud data is reconstructed by the triangulated meshing algorithm to obtain a three-dimensional point cloud model.

3. The method according to claim 1, characterized in that The process of obtaining the crack area includes: The three-dimensional point cloud model is processed and analyzed using a semantic segmentation algorithm based on deep learning to obtain a crack area, wherein the crack area includes different point cloud data containing crack categories, and the location information and range information of the crack area are obtained according to the point cloud distribution of the crack area.

4. The method according to claim 1, wherein The process of obtaining the occurrence information and system information of the fracture includes: The crack area is preprocessed, and the point cloud in the crack area is segmented by a region growing algorithm to obtain a crack surface point cloud subset. The normal vector of the crack surface point cloud subset and each crack area in the crack surface point cloud subset is calculated by a principal component analysis method. The inclination and dip of the crack surface are obtained according to the direction and angle of the normal vector. Cluster analysis is performed based on the inclination and dip of the crack surface to obtain group information, wherein the occurrence information includes the inclination and dip of the crack surface.

5. The method according to claim 1, wherein The crack area is morphologically filtered using a three-dimensional morphological analysis method, and the filtered crack area is sequentially subjected to skeleton extraction and skeleton refinement using a skeleton extraction algorithm and a skeleton refinement algorithm to obtain a crack skeleton line. A threshold judgment of the number of pixel points of the crack skeleton line is performed to obtain a continuity parameter, and based on the continuity parameter, the length parameter of the crack skeleton line is calculated.

6. The method according to claim 1, characterized in that The process of obtaining width distribution characteristics includes: Local point cloud data of the crack is obtained according to the crack skeleton line, and plane fitting is performed on the local point cloud data of the crack by the least squares method to obtain a fitting plane. Distance information is obtained according to the points on the fitting plane and the crack skeleton line. The distance field distribution of the crack direction is obtained according to the distance information. The width change trend is obtained according to the distance field distribution. The width change trend is judged. Based on the judgment result, high-density point cloud data of the abnormal area is obtained by adopting a local encrypted sampling method. The high-density point cloud data of the abnormal area is calculated to obtain width change information, i.e., width distribution characteristics.

7. The method according to claim 1, characterized in that The process of obtaining the filling and dissolution characteristics includes: The point cloud reflection intensity information of the crack area is extracted, and the point cloud reflection intensity information is split and identified using a support vector machine algorithm to obtain the type of filling material inside the crack. Based on the changing trend of the point cloud reflection intensity, the degree of dissolution of the filling material is obtained. Based on the type and degree of dissolution of the filling material inside the crack, the filling dissolution characteristics are obtained.

8. The method according to claim 1, characterized in that The process of obtaining the comprehensive score includes: The occurrence information and group system, fracture continuity parameters, width distribution characteristics, filling and dissolution characteristics and density information are weighted and calculated to obtain a comprehensive score, and the comprehensive score is graded and evaluated to obtain the defect degree identification result of the underground energy storage space structure.

9. A system for identifying defects in underground energy storage space structures, characterized in that: Used to perform the method according to any one of claims 1 to 8.

Images