A lining detachment detection system for thick tunnels with mud-filled broken zones in earthquake-affected areas

Abstract

The present invention discloses a lining loss detection system for mud-filled, crushed zones in thick, seismic-affected tunnels. The system comprises an acquisition module for scanning the tunnel surface using multiple laser radar arrays to acquire point cloud data; a reconstruction module for reconstructing the point cloud data using a point cloud density adaptive algorithm to generate a reconstructed topological map of the thick, seismic-affected tunnel zone; and a detection module for analyzing the mud-filled, crushed zones in the reconstructed topological map. By analyzing changes in the lining surface morphology of the mud-filled, crushed zones, the system identifies lining loss areas and generates a graded early warning report. Accurate lining loss detection can promptly identify potential safety hazards and enable measures to repair them, significantly improving tunnel safety.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel detection, and in particular to a lining shedding detection system for a mud-filled broken zone in a wide and thick earthquake-prone tunnel. Background Art

[0002] The safety of tunnel structures has always been a key research issue in the field of transportation infrastructure, particularly in tunnel lining and water leakage detection. Traditional tunnel detection methods mostly rely on image data or manual inspection. These methods are often affected by factors such as environmental complexity and insufficient lighting, resulting in inaccurate identification of water leakage areas and even difficulty capturing small cracks or damaged areas within the tunnel. While image processing technology performs well in visualization, it cannot provide depth information within the tunnel, thus having certain limitations in tunnel detection. Furthermore, image data poses significant difficulties in accurately locating and quantifying water leakage areas, making it impossible to achieve rapid and accurate detection of leaking areas.

[0003] Meanwhile, laser detection and ranging (LiDAR) technology has been gradually gaining application in tunnel inspection. LiDAR can obtain three-dimensional point cloud data of a tunnel by measuring the intensity of reflected light, providing relatively comprehensive and highly accurate spatial information. In tunnel water leakage detection, LiDAR can help achieve detailed reconstruction of the tunnel's internal structure. However, existing point cloud processing technologies present several challenges, particularly with regard to the complexity of tunnel structures and the density variations of point cloud data. Traditional processing algorithms (such as Alpha-Shapes) often struggle to effectively cope with the unevenness of point cloud density, resulting in an inability to accurately capture small cracks, damage, and water leakage within the tunnel. While the traditional Alpha-Shapes algorithm performs well in point cloud reconstruction and morphological recognition, it suffers from poor adaptability to point cloud density and slow processing speed in tunnel inspection. The Alpha-Shapes algorithm performs poorly when point cloud density varies significantly, making it particularly difficult to effectively capture detailed tunnel information in complex tunnel environments.

[0004] Therefore, a lining loss detection system for thick tunnels with mud-filled fractured zones in earthquake-affected areas is urgently needed. Summary of the Invention

[0005] The present invention provides a lining shedding detection system for a mud-filled broken zone in a wide and thick earthquake-prone tunnel, so as to solve the above-mentioned problems existing in the prior art.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A lining loss detection system for thick tunnels with mud-filled broken zones in earthquake-affected areas, comprising:

[0008] An acquisition module is used to scan the tunnel surface based on multiple sets of lidar arrays to obtain point cloud data;

[0009] The reconstruction module is used to reconstruct the point cloud data based on the point cloud density adaptive algorithm to generate a reconstructed topological map of the tunnel's wide and thick earthquake zone;

[0010] The detection module is used to analyze the mud-filled fractured zones in the reconstructed topological map, identify the lining detachment areas through the changes in the lining surface morphology of the mud-filled fractured zones, and generate a graded early warning report.

[0011] The acquisition module includes:

[0012] The array layout submodule is used to layout multiple groups of lidar arrays along the tunnel axis, obtain the scanning coverage of each array, and generate a full-segment scanning path based on the array spacing;

[0013] The scanning submodule is used to control multiple groups of lidar arrays to perform synchronous scanning based on the full scanning path, and obtain point cloud data of the corresponding scanning area of ​​each array.

[0014] The reconstruction module includes:

[0015] The density distribution submodule is used to calculate the local density distribution of point cloud data based on the VAlpha-Shapes algorithm, and divide the tunnel into high-density areas, medium-density areas, and low-density areas according to the density distribution;

[0016] The adaptive adjustment submodule is used to dynamically adjust the Alpha parameter threshold of the VAlpha-Shapes algorithm according to the density distribution characteristics of each area to generate a reconstructed topology map.

[0017] The detection module includes:

[0018] The data modeling submodule is used to extract the three-dimensional spatial data of the mud-filled fracture zone from the reconstructed topological map, identify the fracture zone boundary based on the region growing algorithm, and generate a spatial distribution model containing length, width, and volume parameters;

[0019] The feature quantification submodule is used to analyze the geometric characteristics of the spatial distribution model. It determines the major axis length and minor axis width through principal component analysis, calculates the cross-sectional area fluctuation coefficient and surface roughness parameters to generate geometric feature vectors, and simultaneously extracts the maximum deformation, deformation rate, and deformation gradient parameters of the mud-filled fracture zone to construct a deformation feature matrix.

[0020] The feature fusion submodule is used to perform multi-dimensional fusion of geometric feature vectors and deformation feature matrices to generate comprehensive assessment indicators for mud-filled fracture zones to classify risk levels and output a three-dimensional heat map containing level labels;

[0021] The early warning generation submodule scans the lining surface morphology in high-grade fracture zones, extracts curvature anomalies, crack density, and deformation gradient parameters, and combines coverage area, deformation acceleration, and crack expansion trend to generate a graded early warning report containing warning level, risk location, and measures for areas that meet the fall-off judgment conditions.

[0022] Among them, the warning generation submodule includes:

[0023] The broken zone scanning unit is used to preferentially perform distributed scanning of the lining surface morphology in high-grade broken zones, generate corresponding morphological change data sets, and extract curvature anomalies, crack distribution density, and local deformation gradients of the lining surface based on the morphological change data sets;

[0024] The risk marking unit is used to mark the area with the risk of lining falling off as an area with the risk of lining falling off, based on the preset falling off judgment conditions, if the curvature anomaly value exceeds the dynamic threshold, the crack distribution density is higher than the regional average, and the local deformation gradient continues to increase;

[0025] The early warning information generation unit is used to count the coverage area, deformation acceleration and crack growth trend parameters of each lining loss risk area, and generate a graded early warning signal for the lining loss risk area based on the ratio of the coverage area to the preset area threshold, the deviation of the deformation acceleration from the preset safety value, and the correlation model between the crack growth trend and the structural stability;

[0026] The graded warning report output unit is used to integrate all graded warning signals and the corresponding spatial distribution data of risk areas, and output a graded warning report containing warning levels, risk locations and recommended measures.

[0027] Among them, the density distribution submodule includes:

[0028] A decreasing Alpha value sequence is used for multi-scale boundary extraction in high-density areas, a fixed Alpha value is used for contour fitting in medium-density areas, and an increasing Alpha value is used for redundant filtering in low-density areas. The upper limit of the decreasing sequence is less than the fixed Alpha value and greater than the lower limit of the increasing sequence.

[0029] The adaptive adjustment submodule includes:

[0030] The topology subgraph optimization unit is used to configure the strategy based on the Alpha parameter of each density area, perform multi-layer boundary fusion calculation on the high-density area to retain complex structural details, perform contour smoothing optimization processing on the medium-density area, and implement topology connection compensation on the low-density area to fill the breakpoints in the sparse area, thereby generating optimized topology subgraphs for each partition;

[0031] The topology reconstruction unit is used to input the optimized topology subgraphs of each partition into the spatial superposition submodule, eliminate the topological faults of adjacent areas through the density weighted interpolation algorithm, and perform geometric consistency correction on the cross nodes in combination with the curvature continuity constraint to generate a preliminary reconstructed topology map of the entire domain;

[0032] The final reconstructed topology map unit is used to verify the topological connectivity based on the preliminary reconstructed topology map of the entire domain. If an unclosed boundary or isolated node is detected, the secondary adjustment mechanism of the Alpha parameter of the corresponding density area is triggered. The boundary extraction and topology optimization are iteratively performed until the preset integrity threshold is met to generate the final reconstructed topology map.

[0033] Among them, the feature quantization submodule includes:

[0034] The displacement vector unit is used to establish a three-dimensional coordinate system in the spatial distribution model and calculate the displacement vector of each node in the mud-filled fracture zone relative to the benchmark model;

[0035] The high-frequency laser scanning unit is used to perform spatiotemporal interpolation of the displacement vector to generate the deformation rate field and deformation gradient field. When the deformation rate is detected to exceed the preset safety threshold, the local monitoring instruction is triggered and the high-frequency laser scanning is started.

[0036] Among them, the feature fusion submodule includes:

[0037] Based on the preset fracture zone classification conditions, the mud-fill fracture zone is divided into multiple graded areas;

[0038] The conditions for classifying the broken zones include: areas with porosity higher than the first preset threshold, permeability higher than the second preset threshold, and deformation rate higher than the third preset threshold are classified as high-grade broken zones; areas with decreasing porosity, permeability, and deformation rate are respectively classified as medium-grade broken zones and low-grade broken zones.

[0039] Among them, the risk marking unit includes:

[0040] The dynamic threshold is adjusted dynamically based on the grade of the broken zone. The dynamic threshold corresponding to the high-grade broken zone is lower than that of the medium and low-grade broken zones.

[0041] The crack distribution density is obtained by calculating the ratio of the total length of cracks per unit area to the area of ​​the region;

[0042] The local deformation gradient is calculated by differentiating the deformation rates of continuous time series, and the size of the gradient calculation window is negatively correlated with the grade of the fracture zone.

[0043] Compared with the prior art, the present invention has the following advantages:

[0044] A lining failure detection system for thick, seismically affected mud-filled, crushed zones in tunnels includes an acquisition module that scans the tunnel surface using multiple LiDAR arrays to acquire point cloud data; a reconstruction module that reconstructs the point cloud data using a point cloud density adaptive algorithm to generate a reconstructed topological map of the thick, seismically affected tunnel; and a detection module that analyzes the crushed zones in the reconstructed topological map, identifies lining failure areas based on surface morphology changes within the zones, and generates a graded early warning report. Accurate lining failure detection can promptly identify potential safety hazards and enable remedial measures to be taken, significantly improving tunnel safety.

[0045] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the present invention.

[0046] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0048] Figure 1 This is a structural diagram of a lining loss detection system for a thick tunnel with muddy filling and broken zone in an earthquake zone according to an embodiment of the present invention;

[0049] Figure 2 A structural diagram of an acquisition module in an embodiment of the present invention;

[0050] Figure 3 This is a structural diagram of the reconstruction module in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0052] The embodiment of the present invention provides Figures 1 to 3 The lining loss detection system for a thick tunnel with mud-filled broken zone in an earthquake zone shown in the figure includes:

[0053] An acquisition module is used to scan the tunnel surface based on multiple sets of lidar arrays to obtain point cloud data;

[0054] The reconstruction module is used to reconstruct the point cloud data based on the point cloud density adaptive algorithm to generate a reconstructed topological map of the tunnel's wide and thick earthquake zone;

[0055] The detection module is used to analyze the mud-filled fractured zones in the reconstructed topological map, identify the lining detachment areas through the changes in the lining surface morphology of the mud-filled fractured zones, and generate a graded early warning report.

[0056] The working principle of the above technical solution is as follows: the acquisition module uses multiple sets of laser radar arrays to scan the tunnel surface. The laser radar emits a laser beam and measures the distance between the reflected laser light spot and the sensor, thereby obtaining the position of each point on the tunnel surface and generating a set of three-dimensional point cloud data; a point cloud is a data set composed of a large number of three-dimensional coordinates, and each point represents the specific location of an object surface.

[0057] The acquired point cloud data is passed to the reconstruction module, which uses the density-adaptive VAlpha-Shapes algorithm to reconstruct the point cloud data. VAlpha-Shapes is a geometric algorithm used to construct the shape (i.e., surface) of an object from point cloud data. Density adaptation automatically adjusts the processing accuracy and method based on the distribution of the point cloud (e.g., dense points in some areas and sparse points in others). This process generates a topological map of the tunnel's wide and thick seismic zone, accurately depicting the tunnel surface shape and revealing potential areas of damage or deformation.

[0058] After obtaining a tunnel topology map, the detection module analyzes "mud-fill fracture zones." Mud-fill fracture zones are loose or cracked areas of the tunnel wall caused by the intrusion of soil or other materials. The system assesses the severity of damage in these areas based on changes in tunnel morphology and deformation. Deformation refers to the magnitude and direction of changes in the tunnel surface, such as whether a section of the wall has exhibited significant deformation or cracks. Based on these changes, the system assigns a grade to the fracture zone, indicating the severity of the damage. Within the mud-fill fracture zone, the system monitors changes in the lining surface morphology to identify areas of lining failure. The lining is a reinforcement layer used to strengthen the tunnel structure. If the lining fails or cracks, it can lead to structural instability. By tracking surface morphological changes in these areas, the system promptly detects signs of lining failure and generates graded warning reports. These reports pinpoint potential areas of damage and provide associated severity ratings for tunnel maintenance personnel to ensure tunnel safety.

[0059] The beneficial effects of the above technical solution are as follows: High-precision LiDAR scanning and point cloud data processing enable the system to accurately identify dangerous areas in the tunnel, particularly mud fill fracture zones and lining debonding areas. Accurate fracture zone classification and lining debonding detection enable timely detection of potential safety hazards and the implementation of remedial measures, significantly improving tunnel safety. The point cloud density-adaptive VAlpha-Shapes algorithm automatically processes point cloud data of varying densities, making the reconstruction process more accurate and efficient. Regardless of the point cloud density of the tunnel surface, the algorithm adapts to ensure a more complete reconstruction of the tunnel shape, providing high-quality data support for subsequent analysis. Through deformation analysis, the system accurately assesses the severity of mud fill fracture zones based on the degree of tunnel surface deformation and categorizes them. This quantitative assessment method is more objective and accurate than traditional manual inspection, providing tunnel maintenance personnel with a reliable basis for decision-making and avoiding excessive or insufficient repair work.

[0060] In another embodiment, the acquisition module includes:

[0061] The array layout submodule is used to layout multiple groups of lidar arrays along the tunnel axis, obtain the scanning coverage of each array, and generate a full-segment scanning path based on the array spacing;

[0062] The scanning submodule is used to control multiple groups of lidar arrays to perform synchronous scanning based on the full scanning path, and obtain point cloud data of the corresponding scanning area of ​​each array.

[0063] The working principle of the above technical solution is as follows: The array placement submodule's primary task is to determine the layout of multiple LiDAR arrays within the tunnel. Multiple LiDAR arrays are arranged along the tunnel's axis (i.e., the tunnel's long axis). Each LiDAR array is a group of sensors that scan the tunnel surface using lasers to acquire point cloud data. During the placement process, the system determines the spacing between arrays based on the scanning range of each LiDAR array. LiDAR scanning coverage is limited; it can only scan up to a certain distance, leaving areas beyond its range unreached. Therefore, the array placement submodule optimally plans the array spacing based on the degree of overlap between these scanning ranges to ensure that every section of the tunnel is fully scanned, avoiding any omissions. After the array placement is complete, the system generates a full-segment scanning path based on the array positions and scanning ranges. This full-segment scanning path refers to the scanning route from one array to another, ensuring that the LiDAR arrays can effectively scan the entire tunnel area, covering every section.

[0064] The scanning submodule controls the synchronized scanning of multiple LiDAR arrays based on the full scanning path generated by the array placement submodule. Multiple LiDAR arrays simultaneously scan the tunnel along a predetermined path, acquiring point cloud data within each scanned area. Point cloud data is the 3D coordinates of reflection points measured by the LiDAR arrays as they scan the tunnel surface. This point cloud data is used for further tunnel analysis and reconstruction. Because multiple LiDAR arrays operate synchronously, they can quickly and comprehensively cover every section of the tunnel, ensuring comprehensive and accurate scanning.

[0065] The beneficial effects of the above technical solution are: through reasonable array layout, the laser radar array can ensure that every part of the tunnel is scanned, avoiding blind spots or missed areas. The synchronous scanning of multiple laser radar arrays greatly improves the scanning efficiency. Compared with the gradual scanning of a single laser radar, multiple arrays working simultaneously can complete a comprehensive scan of the tunnel in a shorter time. This greatly facilitates real-time monitoring and rapid detection of tunnels, especially in cases where the tunnel is long or requires frequent inspections, which can significantly reduce the time cost of manual inspections. The array layout submodule can ensure that there are no omissions or excessive overlaps in the scanning area by precisely controlling the spacing of the laser radar array. This precise layout and scanning ensures the high quality of the point cloud data, and subsequent analysis and processing can be based on these precise raw data.

[0066] In another embodiment, the reconstruction module includes:

[0067] The density distribution submodule is used to calculate the local density distribution of point cloud data based on the VAlpha-Shapes algorithm, and divide the tunnel into high-density areas, medium-density areas, and low-density areas according to the density distribution;

[0068] The adaptive adjustment submodule is used to dynamically adjust the Alpha parameter threshold of the VAlpha-Shapes algorithm according to the density distribution characteristics of each area to generate a reconstructed topology map.

[0069] The working principle of this technical solution is as follows: The density distribution submodule analyzes point cloud data using the VAlpha-Shapes algorithm, calculating the local density distribution at different locations in the tunnel, thereby identifying areas of varying density within the tunnel. The VAlpha-Shapes algorithm is a geometric analysis algorithm based on point cloud data. It describes the distribution of the point cloud by constructing geometric shapes associated with the point cloud data. These geometric shapes help calculate the local density around each point. In other words, the VAlpha-Shapes algorithm can determine where the point cloud data is dense and where it is sparse.

[0070] Based on this algorithm, the density distribution submodule analyzes the tunnel's point cloud data and calculates the point cloud density of each area. Based on the density values, the submodule then divides the tunnel into three types of areas: High-density areas: Point cloud data is relatively concentrated, indicating that the tunnel surface in this area has more features or certain problems. Medium-density areas: Point cloud data is moderate, representing a regular part of the tunnel. Low-density areas: Point cloud data is relatively scattered, indicating that the tunnel surface in this area may not have obvious features, or the scanned area is relatively empty.

[0071] The adaptive adjustment submodule dynamically adjusts the Alpha parameter threshold of the VAlpha-Shapes algorithm based on the density distribution characteristics of different areas in the tunnel. This adjustment process allows the algorithm to better adapt to the characteristics of different density areas, thereby accurately generating a tunnel topology map.

[0072] The Alpha parameter controls how geometric shapes are constructed relative to point cloud data in the VAlpha-Shapes algorithm. Simply put, the Alpha parameter determines which points are connected and which are excluded. If the Alpha value is too large, distant points will be connected; if the Alpha value is too small, important points will be ignored. Therefore, adjusting this parameter is crucial for the results.

[0073] The adaptive adjustment submodule dynamically selects different alpha thresholds based on the density distribution. For example, in high-density areas, the algorithm requires a smaller alpha value to capture more precise details; in low-density areas, a larger alpha value is needed to connect more distant points. In this way, the algorithm optimizes based on the characteristics of the region, ultimately generating an accurate tunnel topology map. A topology map is a graphical representation of the tunnel's internal geometry, clearly demonstrating its structure and morphology and providing a basis for subsequent analysis.

[0074] The beneficial effects of the above technical solution are as follows: Through the local density calculation of the VAlpha-Shapes algorithm, the density distribution submodule can divide the tunnel into high-density, low-density, and medium-density areas. This division can help engineers more accurately understand the characteristics of different tunnel areas, facilitating subsequent analysis, inspection, and maintenance. The adaptive adjustment submodule dynamically adjusts the Alpha parameter based on the density characteristics of different areas to ensure the accuracy of the VAlpha-Shapes algorithm. In high-density areas, more details can be obtained; in low-density areas, the interference of excessive irrelevant points can be avoided. This adaptive optimization allows the algorithm to maintain high-precision calculation results throughout the entire tunnel, improving the reliability of the results. Dynamic adjustment of the Alpha parameter can generate a more accurate tunnel topology map. The topology map can describe the geometric structure of the tunnel, helping engineers better understand the tunnel shape and identify possible structural problems or safety hazards. Therefore, this process is of great significance in tunnel monitoring and maintenance.

[0075] In another embodiment, the detection module includes:

[0076] The data modeling submodule is used to extract the three-dimensional spatial data of the mud-filled fracture zone from the reconstructed topological map, identify the fracture zone boundary based on the region growing algorithm, and generate a spatial distribution model containing length, width, and volume parameters;

[0077] The feature quantification submodule is used to analyze the geometric characteristics of the spatial distribution model. It determines the major axis length and minor axis width through principal component analysis, calculates the cross-sectional area fluctuation coefficient and surface roughness parameters to generate geometric feature vectors, and simultaneously extracts the maximum deformation, deformation rate, and deformation gradient parameters of the mud-filled fracture zone to construct a deformation feature matrix.

[0078] The feature fusion submodule is used to perform multi-dimensional fusion of geometric feature vectors and deformation feature matrices to generate comprehensive assessment indicators for mud-filled fracture zones to classify risk levels and output a three-dimensional heat map containing level labels;

[0079] The early warning generation submodule scans the lining surface morphology in high-grade fracture zones, extracts curvature anomalies, crack density, and deformation gradient parameters, and combines coverage area, deformation acceleration, and crack expansion trend to generate a graded early warning report containing warning level, risk location, and measures for areas that meet the fall-off judgment conditions.

[0080] The spatial distribution model including length, width and volume parameters is generated, including:

[0081] The three-dimensional topological map includes point cloud data, which includes position coordinates, reflection intensity and color information to form a first data set describing the spatial structure of the tunnel;

[0082] Extracting 3D spatial data of mud-filled fractured zones from the 3D topological map. Specifically, based on the first data set, areas with a reflection intensity below 40DN and a darker color are selected as initial candidate areas for mud-filled fractured zones, thereby generating a second data set reflecting the potential distribution of the fractured zones.

[0083] When the second data set is obtained, boundary recognition is performed on each initial candidate area, and a three-dimensional boundary model of the broken zone is generated based on the region growing algorithm. Specifically, starting from the geometric center point of each candidate area, the growth conditions are set as follows: the difference in density of adjacent point clouds is less than 15% and the curvature change is less than 0.2 radians / meter, and similar points are gradually diffused outward to merge until a boundary point with a reflection intensity greater than 60 DN or a curvature greater than 0.5 radians / meter is encountered, forming a closed three-dimensional model as the third data set, where DN represents a normalized unit, which is a unit of reflection intensity measurement in point cloud data processing and is expressed in the range of 0-255 DN;

[0084] The third data set is displayed to generate a spatial distribution model containing length, width and volume parameters to enable the risk of lining failure to be assessed. Specifically, the dimensions of the circumscribed cube of each broken zone are calculated, including length, width and volume. The volume parameters in the spatial distribution model are calculated using a more precise method, which is different from the simplified estimation method used in the initial candidate area extraction stage. The principal component analysis is used to extract the main axis direction and length, and the width of the secondary axis in the vertical direction. The cross-section is cut every 0.5 meters along the main axis to calculate the area fluctuation coefficient. If the fluctuation coefficient exceeds 1.0, it is marked as a high-risk failure area.

[0085] The working principle of the above technical solution is: in the reconstructed three-dimensional topological map, the areas with reflection intensity lower than 40DN and dark color are automatically screened as the initial candidate areas of the mud-filled fracture zone, and the regional growing algorithm is started from the geometric center point of each candidate area. The density difference of adjacent point clouds is set to be less than 15% and the curvature change is less than 0.2 radians / meter as the growth conditions. Similar points are gradually diffused outward and merged until a boundary point with a sudden increase in reflection intensity (such as >60DN) or a sudden change in curvature (>0.5 radians / meter) is encountered, forming a closed three-dimensional model and calculating the size of its circumscribed cube (for example, 12 meters long, 2.3 meters wide, and 28 cubic meters in volume). Then, principal component analysis is performed on each fracture zone. By calculating the covariance matrix of all point coordinates, the direction of the maximum eigenvector is extracted as the principal axis (for example, extending longitudinally along the tunnel), the distance between the two farthest points along the principal axis is measured as the principal axis length (12 meters), and the maximum span in the vertical direction is taken as the secondary axis width (2.3 meters). At the same time, cross sections are cut every 0.5 meters along the principal axis, and the maximum fluctuation amplitude of each cross-sectional area is counted (for example, the area of ​​a section suddenly increases from 1.2㎡ to 4.5㎡, and the fluctuation coefficient = (4.5-1.2) / average area 2.8≈1.18). If the fluctuation coefficient exceeds the threshold of 1.0, it is marked as an irregular shape.

[0086] The system uses the Delaunay triangulation algorithm to connect the surface point cloud of the broken zone into a triangular mesh (with the side length of a single triangle limited to 0.1-0.3 meters). It then calculates the angle between the normal vectors of each triangle (for example, a 25° angle between the normal vectors of adjacent triangles is considered high curvature). The percentage of high-curvature patches is calculated as a roughness parameter (for example, if 30% of the patches exceed the curvature limit, the roughness is 0.3). The system also compares the current and historical scan data, calculating the three-dimensional displacement point by point (for example, a point shifts 0.05 meters), extracting the maximum value as the maximum deformation, and calculating the deformation rate based on the time interval (for example, 30 days) (0.05 meters / 30 days ≈ 0.0017 meters / day). The broken zone is divided into 1-meter segments along the main axis, and the difference in deformation between adjacent segments is calculated as the deformation gradient (for example, the second segment has a displacement of 0.03 meters, the third segment has a displacement of 0.06 meters, and the gradient is 0.03 meters / meter).

[0087] The geometric feature vector consists of length, width, fluctuation coefficient, and roughness, while the deformation feature matrix includes the maximum deformation, rate, and gradient. After normalizing the two types of data, the system sets three classification thresholds based on porosity (30%), permeability (1×10⁻¹²m²), and deformation rate (5mm / day): If a fracture zone has a porosity of 35%, a permeability of 1.2×10⁻¹²m², and a deformation rate of 6mm / day, all three items exceed the first threshold and are marked as red high risk; if the porosity is 25%, the permeability is 6×10⁻¹³m², and the deformation rate is 3mm / day, it is classified as yellow medium risk; if the porosity is 10%, the permeability is 8×10⁻¹, and the deformation rate is 8×10⁻¹, it is classified as yellow medium risk. 4 m² and a deformation rate of 1 mm / day are classified as green, low risk. Debris zones with a comprehensive score exceeding 80 points (out of 100) trigger red thermal map annotation and are highlighted in the 3D model.

[0088] For the red high-risk areas, the system starts high-frequency laser scanning (once per minute) to detect sudden changes in the curvature of the lining surface (such as the local curvature suddenly increasing from 0.3 to 0.6 radians / meter), crack density (the total length of cracks within 1㎡ increases from 2 meters to 3.5 meters) and deformation gradient acceleration (such as the gradient increases from 0.03 meters / meter to 0.05 meters / meter). When the curvature is greater than 0.5 radians / meter, the crack density is greater than 3 meters / square meter, and the gradient increases for three consecutive times, the risk area is calculated (for example, 2.5 square meters exceeds the threshold of 2 square meters), the deformation acceleration (rate from 0.0017 to 0.002 to 0.003 meters / day to acceleration of 0.0003 meters / day²), and the crack extension trend (the ARIMA model predicts that the crack will extend by 0.15 meters after three days) are calculated. A red alert report is generated with the annotation "The risk of lining failure within 72 hours in the 2.5 square meter area of ​​K25+300 is extremely high. Closure and disposal are recommended within 24 hours." The 3D coordinate bounding box (X: 100-105m, Y: 3-5m, Z: 0-2m) is also output for engineers to quickly locate the area.

[0089] The beneficial effects of the above technical solution are: it helps to accurately identify and evaluate the spatial distribution and deformation characteristics of mud-filled crushed zones. Combined with multi-dimensional analysis and early warning systems, it provides strong data support and decision-making basis for stability analysis, risk assessment and engineering decision-making in crushed zones, thereby improving safety and protection effects.

[0090] In another embodiment, the warning generation submodule includes:

[0091] The broken zone scanning unit is used to preferentially perform distributed scanning of the lining surface morphology in high-grade broken zones, generate corresponding morphological change data sets, and extract curvature anomalies, crack distribution density, and local deformation gradients of the lining surface based on the morphological change data sets;

[0092] The risk marking unit is used to mark the area with the risk of lining falling off as an area with the risk of lining falling off, based on the preset falling off judgment conditions, if the curvature anomaly value exceeds the dynamic threshold, the crack distribution density is higher than the regional average, and the local deformation gradient continues to increase;

[0093] The early warning information generation unit is used to count the coverage area, deformation acceleration and crack growth trend parameters of each lining loss risk area, and generate a graded early warning signal for the lining loss risk area based on the ratio of the coverage area to the preset area threshold, the deviation of the deformation acceleration from the preset safety value, and the correlation model between the crack growth trend and the structural stability;

[0094] The graded warning report output unit is used to integrate all graded warning signals and the corresponding spatial distribution data of risk areas, and output a graded warning report containing warning levels, risk locations and recommended measures.

[0095] The working principle of the above technical solution is: the system first starts high-precision laser scanning of the broken zone areas marked as high-risk levels in the three-dimensional thermal map, controls multiple groups of laser radars to scan these areas three times in a row with an interval of 10 minutes at a resolution of 0.1 mm, and records the changes in surface concave and convexity to form a dynamic data set. The curvature calculation algorithm is used to analyze the degree of curvature of each point on the lining surface. By comparing historical data, areas where the curvature suddenly increases and exceeds the dynamic threshold are screened out (for example, the curvature value of a certain point suddenly increases from 0.1 to 0.5, while the historical average curvature of the area is 0.2±0.05, and the dynamic threshold is set to 0.3). At the same time, linear areas with sudden color changes or point cloud breaks in the scanned data are identified, and the total length of cracks per square meter is counted as a density indicator (for example, there are 2.5 meters of cracks within 1 square meter in a certain area, and the density is marked as 2.5m / ㎡). The surface is divided into 0.1m×0.1m grid modules, and the rate of change of the deformation variable over time in each module is calculated (for example, the deformation variables of a grid module in three scans are 0.1mm, 0.3mm, and 0.6mm respectively, and the deformation gradient is calculated to be 0.25mm / min²). When an area meets the conditions of curvature exceeding the dynamic threshold, crack density being 1.5 times higher than the average value of the fracture zone, and deformation gradient increasing three times in a row, the system will mark it as a fall-off risk area. For example, if the curvature of an area is 0.35, the crack density is 3.0m / ㎡ (regional average is 2.0m / ㎡), and the deformation gradient increases from 0.1 to 0.3mm / min², it will be judged as high risk.

[0096] The system uses an image segmentation algorithm to calculate the projected area of ​​the risk zone within the 3D model (for example, a projected area of ​​2.3 square meters). It then performs a quadratic difference on the deformation rate to calculate the deformation acceleration (for example, three scans with deformation rates of 0.2, 0.4, and 0.7 mm / min yield an acceleration of 0.25 mm / min²). It then predicts crack growth trends based on time series analysis (for example, a crack has grown by 0.12 meters over the past 24 hours, with a fitted slope of 0.005 m / h). The covered area is compared with a preset safety threshold of 1.5 square meters (2.3 / 1.5 = 1.53 times). The deviation of the deformation acceleration from the safety value of 0.1 mm / min² is calculated ((0.25 - 0.1) / 0.1 × 100% = 150%). The crack growth trend is then input into the structural stability model to assess risk (for example, a slope of 0.005 m / h predicts that the crack will exceed the support structure's tolerance limit in three days). If the area ratio is ≥1.5, the deviation is ≥100%, and the model prediction expiration time is less than 7 days, a red alert is issued. If the area ratio is 1 ≤ <1.5, the deviation is 50% ≤ <100%, and the expiration time is 7-15 days, a yellow alert is issued. All other alerts are marked green. Ultimately, the alert level (color), 3D coordinates, and recommended actions are integrated into a report, and the risk distribution is displayed through a 3D heat map overlay.

[0097] The beneficial effects of this technical solution include early identification of areas at risk of lining failure and timely alerts, providing a scientific basis for project safety management. This method effectively prevents accidents caused by lining failure. Furthermore, through dynamic monitoring and data analysis, risk assessments can be adjusted in real time for precise prevention and control. The generation of graded warning signals allows for tailored response measures to each risk area, avoiding excessive intervention and waste of resources.

[0098] In another embodiment, the density distribution submodule includes:

[0099] A decreasing Alpha value sequence is used for multi-scale boundary extraction in high-density areas, a fixed Alpha value is used for contour fitting in medium-density areas, and an increasing Alpha value is used for redundant filtering in low-density areas. The upper limit of the decreasing sequence is less than the fixed Alpha value and greater than the lower limit of the increasing sequence.

[0100] The working principle of the above technical solution is as follows: the system first divides the area into three categories: high, medium, and low according to the point cloud density, and adopts differentiated Alpha value adjustment strategies for areas with different densities. For high-density areas, starting from the initial Alpha value of 0.8, it gradually decreases to 0.3 in steps of 0.1. After each adjustment, the Alpha shape algorithm is run to extract the boundary contour. For example, 0.8 is used to extract the large-scale boundary for the first time, and 0.7 is used to refine the edge for the second time. Finally, the multiple results are superimposed and fused to retain the recurring high-frequency edge features (such as the arc-shaped boundary that appears in three consecutive extractions). The medium-density area directly uses a fixed Alpha value of 0.5 to perform a single Alpha shape calculation to generate a smooth and coherent closed contour line. For example, if the point cloud in a certain area is evenly distributed, an Alpha value of 0.5 just connects adjacent points to form a complete polygon. Low-density areas start with an alpha value of 0.2 and increase in steps of 0.15 to 0.5, gradually filtering out loose outliers. For example, 0.2 is used to remove isolated points far from the main group for the first time, and 0.35 is used a second time to further remove minor scattered points. Finally, the core point set is retained and a simplified boundary is generated using 0.5. During this process, the condition that the upper bound of the high-density decreasing sequence, 0.8, is less than the fixed value of 0.5 for the medium density does not hold. In fact, it is corrected to the upper bound of the decreasing sequence, 0.3, which is less than the fixed value of 0.5, and the lower bound of the increasing sequence, 0.2, which is greater than the fixed value of 0.5. Through parameter constraints, high-density areas focus on capturing details and low-density areas strengthen noise filtering.

[0101] The beneficial effect of this technical solution is that the upper bound of the decreasing alpha value sequence is less than the fixed alpha value and greater than the lower bound of the increasing alpha value sequence. This allows for a natural transition between density regions, reduces sudden changes, enhances the smoothness and consistency of overall boundary extraction, and improves the quality of final boundary recognition. This method not only optimizes boundary extraction accuracy but also reduces redundant data while maintaining computational efficiency, improving stability and adaptability when handling complex scenarios.

[0102] In another embodiment, the adaptive adjustment submodule includes:

[0103] The topology subgraph optimization unit is used to configure the strategy based on the Alpha parameter of each density area, perform multi-layer boundary fusion calculation on the high-density area to retain complex structural details, perform contour smoothing optimization processing on the medium-density area, and implement topology connection compensation on the low-density area to fill the breakpoints in the sparse area, thereby generating optimized topology subgraphs for each partition;

[0104] The topology reconstruction unit is used to input the optimized topology subgraphs of each partition into the spatial superposition submodule, eliminate the topological faults of adjacent areas through the density weighted interpolation algorithm, and perform geometric consistency correction on the cross nodes in combination with the curvature continuity constraint to generate a preliminary reconstructed topology map of the entire domain;

[0105] The final reconstructed topology map unit is used to verify the topological connectivity based on the preliminary reconstructed topology map of the entire domain. If an unclosed boundary or isolated node is detected, the secondary adjustment mechanism of the Alpha parameter of the corresponding density area is triggered. The boundary extraction and topology optimization are iteratively performed until the preset integrity threshold is met to generate the final reconstructed topology map.

[0106] The technical solution works by performing multi-layer boundary fusion calculations on high-density areas. The Alpha shape algorithm (a geometric construction method that uses parameters to control the tightness of boundaries) is run multiple times using a decreasing sequence of alpha values ​​(e.g., 0.8, 0.7, 0.6). Each time, contour lines of varying degrees of detail are extracted. The results are then superimposed and a voting mechanism is used to retain boundary features with a repetition rate exceeding 70% (e.g., if a curved edge appears twice in three extractions, it is retained). Morphological closing operations are then used to fill in small gaps. For medium-density areas, an initial contour is generated using a fixed alpha value of 0.5. A B-spline curve fitting algorithm is then used to smooth jagged edges (e.g., replacing broken line corners with arcs of continuous curvature). Isolated short edges less than 5 cm in length are removed. When breakpoints still exist in low-density areas after the Alpha value increases to 0.5, the system automatically identifies breaks where the distance between adjacent point clouds is less than 20 cm, inserts virtual connecting line segments, and extends them along the normal vector direction to the actual point cloud (for example, a straight line connection is generated between two breakpoints 15 cm apart, and extended 10 cm on both sides to connect with the actual point cloud), forming a closed topological structure.

[0107] Each partitioned subgraph is fed into the spatial overlay submodule, which interpolates the boundaries of adjacent regions using point cloud density as a weight. For example, at the intersection of a high-density region with a weight of 0.8 and a medium-density region with a weight of 0.5, an intermediate transition curve is generated based on the weight ratio. Curvature differences at intersection nodes are also detected (e.g., the difference between a high-density region with a curvature of 0.3 and a medium-density region with a curvature of 0.5). Curve control points are adjusted to maintain the curvature change rate within 0.2 per second (similar to the smooth turning of a car's steering wheel when turning). After generating a preliminary global topology reconstruction, the system automatically traverses all boundary segments, marking unclosed areas (e.g., where the distance between the start and end points of a boundary segment is greater than 1 mm) and isolated node clusters (e.g., with fewer than 10 points and a distance greater than 50 cm from the main region). This triggers a secondary adjustment of the alpha parameters for these corresponding regions: the lower limit of the alpha value is lowered to 0.2 in high-density regions to capture finer features, while the upper limit is raised to 0.6 in low-density regions to strengthen connectivity. The fusion and compensation process is then repeated until the proportion of unclosed boundaries drops below 0.1% and the number of isolated nodes is reduced to zero. Finally, a complete reconstructed topology is output.

[0108] The beneficial effect of the above technical solution is that the generated reconstructed topology map not only meets the accuracy requirements, but also effectively avoids topological faults or isolated areas, thereby improving the practical application value of the topology map.

[0109] In another embodiment, the feature quantization submodule includes:

[0110] The displacement vector unit is used to establish a three-dimensional coordinate system in the spatial distribution model and calculate the displacement vector of each node in the mud-filled fracture zone relative to the benchmark model;

[0111] The high-frequency laser scanning unit is used to perform spatiotemporal interpolation of the displacement vector to generate the deformation rate field and deformation gradient field. When the deformation rate is detected to exceed the preset safety threshold, the local monitoring instruction is triggered and the high-frequency laser scanning is started.

[0112] The working principle of this technical solution is as follows: the system first selects an undamaged, stable area in the reconstructed 3D topology as the origin of the reference coordinate system (for example, the center of the tunnel vault). The coordinates of the nodes in the currently scanned fracture zone are precisely aligned with the historical reference model using the ICP point cloud registration algorithm. (For example, the original coordinates of a node (10.2, 3.5, 0.0) become (10.3, 3.4, -0.1), and the displacement vector is calculated as (0.1, -0.1, -0.1) meters.) The 3D displacement of each node is interpolated in the time dimension. If the interval between two scans is 30 days, the deformation rate is calculated as displacement / 30 (for example, a node displacement of 0.3 meters will have a rate of 0.01 meters per day). The system also divides the spatial area into 0.5-meter grids along the X, Y, and Z axes. The displacement change rate of adjacent nodes within each grid is calculated as the deformation gradient (for example, a displacement difference of 0.02 meters / 0.5 meters between adjacent nodes in the X direction will have a gradient of 0.04 meters / meter). When the deformation rate of a certain area exceeds the preset threshold (such as 0.005 meters / day stipulated in tunnel safety standards) for three consecutive times, the system automatically adjusts the lidar scanning frequency of the area from once per hour to once per minute, and enables high-precision mode (resolution increased from 5 mm to 1 mm), tracking the changing trend of the displacement vector in real time (for example, the rate at a certain point accelerates from 0.006 meters / day to 0.008 meters / day) until the rate drops below the threshold and normal monitoring is resumed.

[0113] The beneficial effects of this technical solution include not only improving the accuracy and response speed of deformation monitoring but also enhancing the safety of the overall structure. The deformation rate and gradient fields generated through spatiotemporal interpolation can effectively guide subsequent engineering optimization, risk management, and emergency response, ensuring that necessary protective measures are implemented in high-risk situations.

[0114] In another embodiment, the feature fusion submodule includes:

[0115] Based on the preset fracture zone classification conditions, the mud-fill fracture zone is divided into multiple graded areas;

[0116] The conditions for classifying the broken zones include: areas with porosity higher than the first preset threshold, permeability higher than the second preset threshold, and deformation rate higher than the third preset threshold are classified as high-grade broken zones; areas with decreasing porosity, permeability, and deformation rate are respectively classified as medium-grade broken zones and low-grade broken zones.

[0117] The working principle of the above technical solution is to traverse all the fracture zones and calculate the porosity (the percentage of void volume to the total volume of the area, for example, 35% of the volume of a certain area is voids), permeability (the ability of fluid to pass through the area, unit square meter, for example, 1×10⁻¹²m² means that one trillionth of a cubic meter of fluid can pass through each square meter of cross-section per second) and deformation rate (the number of millimeters of displacement change per day, for example, a certain area moves 6 mm per day). If a region satisfies the following conditions simultaneously: porosity > 30%, permeability > 1×10⁻¹²m², and deformation rate > 5mm / day (e.g., porosity 35%, permeability 1.5×10⁻¹²m², and deformation rate 6mm / day), it is marked as a primary fracture zone and highlighted in red in the 3D model. If the porosity is between 15% and 30%, the permeability is between 1×10⁻¹³ and 1×10⁻¹²m², and the deformation rate is 2-5mm / day (e.g., porosity 25%, permeability 5×10⁻¹³m², and deformation rate 3mm / day), it is classified as a secondary fracture zone and marked in yellow. If the porosity is <15%, permeability <1×10⁻¹³m², and deformation rate <2mm / day (e.g., porosity 10%, permeability 8×10⁻¹³m², and deformation rate 10mm / day), it is classified as a secondary fracture zone and highlighted in yellow. 4 m², deformation rate 1 mm / day), is classified as a Level 3 fracture zone and highlighted in green. For situations where parameters cross levels (e.g., porosity 28%, permeability 1.2×10⁻¹²m², deformation rate 4 mm / day), the system prioritizes the deformation rate (4 mm / day falls within the Level 2 range), then checks the permeability (1.2×10⁻¹²m² exceeds the Level 1 threshold). Ultimately, because the porosity does not reach 30%, the zone is downgraded to Level 2. All assessment results are mapped in real time to a 3D thermal map, allowing engineers to visually identify high-risk areas by color (e.g., red areas require priority reinforcement).

[0118] The beneficial effects of the above technical solution are: the classification of broken zones can not only improve the ability to identify deformation and risks in different areas, but also optimize resource allocation, improve response speed and enhance safety, providing reliable data support and decision-making basis for various projects and natural disaster prevention.

[0119] In another embodiment, the risk marking unit includes:

[0120] The dynamic threshold is adjusted dynamically based on the grade of the broken zone. The dynamic threshold corresponding to the high-grade broken zone is lower than that of the medium and low-grade broken zones.

[0121] The crack distribution density is obtained by calculating the ratio of the total length of cracks per unit area to the area of ​​the region;

[0122] The local deformation gradient is calculated by differentiating the deformation rates of continuous time series, and the size of the gradient calculation window is negatively correlated with the grade of the fracture zone.

[0123] The working principle of this technical solution is as follows: the system first dynamically sets the breakage determination threshold based on the level of the broken zone. For example, a first-level high-level broken zone uses a curvature dynamic threshold of 0.3 (the historical average curvature is 0.2±0.05), a second-level medium-level threshold of 0.4, and a third-level low-level threshold of 0.5, ensuring that high-risk areas trigger warnings more sensitively. For crack distribution density, the system automatically identifies linear cracks with missing continuous point clouds or sudden color changes within each square meter. It measures the pixel length of each crack and accumulates it (for example, three cracks are detected in a 1㎡ area, with lengths of 0.5m, 0.8m, and 1.2m, respectively, for a total length of 2.5m → density 2.5m / ㎡), then divides it by the area to obtain the density value. When calculating local deformation gradients, a three-day time window is used for level one fracture zones (the interval for differential deformation rates is short, for example, a displacement of 2mm on day one, 3mm on day two, and 5mm on day three → a gradient of (5-3) / 1 = 2mm / day²), a five-day window is used for level two, and a seven-day window is used for level three. Smaller windows are more sensitive to transient changes (for example, a sudden increase in gradient of 2mm / day² in a level one fracture zone triggers an alarm, while a level three zone requires a gradient >1mm / day² for seven consecutive days). When an area simultaneously meets the conditions of a curvature exceeding a dynamic threshold, a crack density exceeding 1.2 times the level average, and a sustained positive gradient increase, the system marks it as a fallout risk area. For example, a level one zone with a curvature of 0.35, a crack density of 3.0m / m² (average 2.5m / m²), and a gradient increasing from 0.5 to 0.8mm / day² is considered high risk.

[0124] The beneficial effect of this technical solution is that by setting multiple criteria, including dynamic thresholds, crack distribution density, and local deformation gradients, it can significantly improve the accuracy and real-time performance of fallout risk assessment. This method not only provides a scientific basis for accurately identifying high-risk areas, but also improves the monitoring system's response speed and processing efficiency, ensuring the safety of the fracture zone.

[0125] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A lining detachment detection system for thick tunnels in earthquake-affected mud-filled fractured zones, characterized by: include: An acquisition module is used to scan the tunnel surface based on multiple sets of lidar arrays to obtain point cloud data; The reconstruction module is used to reconstruct the point cloud data based on the point cloud density adaptive algorithm to generate a reconstructed topological map of the tunnel's wide and thick earthquake zone; The detection module is used to analyze the mud-filled fractured zones in the reconstructed topological map, identify lining failure areas based on changes in the lining surface morphology in the mud-filled fractured zones, and generate a graded early warning report. The detection module includes: The data modeling submodule is used to extract the three-dimensional spatial data of the mud-filled fracture zone from the reconstructed topological map, identify the fracture zone boundary based on the region growing algorithm, and generate a spatial distribution model containing length, width, and volume parameters; The feature quantification submodule is used to analyze the geometric characteristics of the spatial distribution model. It determines the major axis length and minor axis width through principal component analysis, calculates the cross-sectional area fluctuation coefficient and surface roughness parameters to generate geometric feature vectors, and simultaneously extracts the maximum deformation, deformation rate, and deformation gradient parameters of the mud-filled fracture zone to construct a deformation feature matrix. The feature fusion submodule is used to perform multi-dimensional fusion of geometric feature vectors and deformation feature matrices to generate comprehensive assessment indicators for mud-filled fracture zones to classify risk levels and output a three-dimensional heat map containing level labels; The early warning generation submodule scans the lining surface morphology in high-grade fracture zones, extracts curvature anomalies, crack density, and deformation gradient parameters, and combines coverage area, deformation acceleration, and crack expansion trend to generate a graded early warning report containing warning level, risk location, and measures for areas that meet the fall-off judgment conditions.

2. The lining fall-off detection system for thick tunnel mud-filled broken zone in earthquake-affected areas according to claim 1 is characterized in that: The acquisition module includes: The array layout submodule is used to layout multiple groups of lidar arrays along the tunnel axis, obtain the scanning coverage of each array, and generate a full-segment scanning path based on the array spacing; The scanning submodule is used to control multiple groups of lidar arrays to perform synchronous scanning based on the full scanning path, and obtain point cloud data of the corresponding scanning area of ​​each array.

3. The lining detachment detection system for thick tunnel mud-filled broken zone in earthquake zone according to claim 1 is characterized in that: The reconstruction module includes: The density distribution submodule is used to calculate the local density distribution of point cloud data based on the VAlpha-Shapes algorithm, and divide the tunnel into high-density areas, medium-density areas, and low-density areas according to the density distribution; The adaptive adjustment submodule is used to dynamically adjust the Alpha parameter threshold of the VAlpha-Shapes algorithm according to the density distribution characteristics of each area to generate a reconstructed topology map.

4. The lining fall-off detection system for thick tunnel mud-filled broken zone in earthquake-affected areas according to claim 1 is characterized in that: The warning generation submodule includes: The broken zone scanning unit is used to preferentially perform distributed scanning of the lining surface morphology in high-grade broken zones, generate corresponding morphological change data sets, and extract curvature anomalies, crack distribution density, and local deformation gradients of the lining surface based on the morphological change data sets; The risk marking unit is used to mark the area with the risk of lining falling off as an area with the risk of lining falling off, based on the preset falling off judgment conditions, if the curvature anomaly value exceeds the dynamic threshold, the crack distribution density is higher than the regional average, and the local deformation gradient continues to increase; The early warning information generation unit is used to count the coverage area, deformation acceleration and crack growth trend parameters of each lining loss risk area, and generate a graded early warning signal for the lining loss risk area based on the ratio of the coverage area to the preset area threshold, the deviation of the deformation acceleration from the preset safety value, and the correlation model between the crack growth trend and the structural stability; The graded warning report output unit is used to integrate all graded warning signals and the corresponding spatial distribution data of risk areas, and output a graded warning report containing warning levels, risk locations and recommended measures.

5. The lining detachment detection system for thick tunnel mud-filled broken zone in earthquake zone according to claim 3 is characterized in that: The density distribution submodule includes: A decreasing Alpha value sequence is used for multi-scale boundary extraction in high-density areas, a fixed Alpha value is used for contour fitting in medium-density areas, and an increasing Alpha value is used for redundant filtering in low-density areas. The upper limit of the decreasing sequence is less than the fixed Alpha value and greater than the lower limit of the increasing sequence.

6. The lining fall-off detection system for thick tunnel mud-filled broken zone in earthquake-affected areas according to claim 3 is characterized in that: The adaptive adjustment submodule includes: The topology subgraph optimization unit is used to configure the strategy based on the Alpha parameter of each density area, perform multi-layer boundary fusion calculation on the high-density area to retain complex structural details, perform contour smoothing optimization processing on the medium-density area, and implement topology connection compensation on the low-density area to fill the breakpoints in the sparse area, thereby generating optimized topology subgraphs for each partition; The topology reconstruction unit is used to input the optimized topology subgraphs of each partition into the spatial superposition submodule, eliminate the topological faults of adjacent areas through the density weighted interpolation algorithm, and perform geometric consistency correction on the cross nodes in combination with the curvature continuity constraint to generate a preliminary reconstructed topology map of the entire domain; The final reconstructed topology map unit is used to verify the topological connectivity based on the preliminary reconstructed topology map of the entire domain. If an unclosed boundary or isolated node is detected, the secondary adjustment mechanism of the Alpha parameter of the corresponding density area is triggered. The boundary extraction and topology optimization are iteratively performed until the preset integrity threshold is met to generate the final reconstructed topology map.

7. The lining detachment detection system for thick tunnel mud-filled broken zone in earthquake zone according to claim 1 is characterized in that: The feature quantization submodule includes: The displacement vector unit is used to establish a three-dimensional coordinate system in the spatial distribution model and calculate the displacement vector of each node in the mud-filled fracture zone relative to the benchmark model; The high-frequency laser scanning unit is used to perform spatiotemporal interpolation of the displacement vector to generate the deformation rate field and deformation gradient field. When the deformation rate is detected to exceed the preset safety threshold, the local monitoring instruction is triggered and the high-frequency laser scanning is started.

8. The lining fall-off detection system for thick tunnel mud-filled broken zone in earthquake-affected areas according to claim 1 is characterized in that: The feature fusion submodule includes: Based on the preset fracture zone classification conditions, the mud-fill fracture zone is divided into multiple graded areas; The conditions for classifying the broken zones include: areas with porosity higher than the first preset threshold, permeability higher than the second preset threshold, and deformation rate higher than the third preset threshold are classified as high-grade broken zones; areas with decreasing porosity, permeability, and deformation rate are respectively classified as medium-grade broken zones and low-grade broken zones.

9. The lining fall-off detection system for thick tunnel mud-filled broken zone in earthquake-affected areas according to claim 4 is characterized in that: Risk marking units include: The dynamic threshold is adjusted dynamically based on the grade of the broken zone. The dynamic threshold corresponding to the high-grade broken zone is lower than that of the medium and low-grade broken zones. The crack distribution density is obtained by calculating the ratio of the total length of cracks per unit area to the area of ​​the region; The local deformation gradient is calculated by differentiating the deformation rates of continuous time series, and the size of the gradient calculation window is negatively correlated with the grade of the fracture zone.

Images