A tunnel cross-section deformation monitoring method, device and system

Abstract

The present application relates to the technical field of tunnel engineering safety monitoring, in particular to a tunnel cross-section deformation monitoring method, device and system; comprising data acquisition and grid construction steps: tunnel cross-section laser point cloud and reflection intensity are acquired, a triangular mesh with a manifold topological structure is constructed and virtual physical properties are given; dynamic evolution step: a virtual shrinkage potential field is constructed, the mesh is iteratively evolved to a thermodynamic equilibrium state, and a denoising reconstructed mesh is obtained; energy mapping step: the strain energy density value of the mesh relative to the initial state is calculated, and a deformation thermodynamic map is generated; feature extraction and early warning step: the energy gradient feature is extracted based on the thermodynamic map, the deformation area and trend are determined, the early warning instruction is generated, and the scanning parameters of the acquisition device are adjusted; the present application solves the problem of traditional geometric filtering smoothing small deformation features by introducing a virtual physical field evolution mechanism, effectively reduces the false positive rate caused by environmental interference without reducing the detection sensitivity.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering safety monitoring technology, specifically to a method, equipment, and system for monitoring tunnel cross-sectional deformation. Background Technology

[0002] Tunnel cross-section deformation monitoring is a key means to ensure the structural safety of underground engineering projects. At present, non-contact measurement methods based on three-dimensional laser scanning are widely used. By collecting point cloud data of the tunnel inner wall, a geometric model is constructed and compared with the design model or historical data to obtain the deformation amount. It has the advantages of long detection distance, fast sampling speed and high data density.

[0003] With the increasing demand for refined tunnel operation and maintenance, data processing algorithms have gradually evolved from simple distance comparison to complex surface fitting and geometric filtering. However, in the actual tunnel operation environment, high dust concentration, high humidity, and unevenness of the shotcrete surface often introduce a lot of noise into the point cloud data. Existing geometric filtering algorithms, while smoothing out noise, tend to smooth out the micro deformation features that are crucial for early warning, making it difficult to effectively distinguish between environmental interference and structural stress concentration. This results in insufficient sensitivity for identifying early micro deformations and a high false alarm rate. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method, device, and system for monitoring tunnel cross-section deformation. Specifically, the technical solution of this invention is as follows:

[0005] A method for monitoring tunnel cross-sectional deformation, comprising:

[0006] Laser point cloud data and reflection intensity data of the tunnel cross section are collected as the original measurement set, and a triangular mesh with a manifold topology is constructed based on the original measurement set, and virtual physical properties are assigned to the vertices of the triangular mesh.

[0007] A virtual contraction potential energy field is constructed, and the triangular mesh is subjected to dynamic evolution iteration in the virtual contraction potential energy field until the triangular mesh reaches a thermodynamic equilibrium state, thereby obtaining a denoised reconstructed mesh.

[0008] Calculate the strain energy density value of the denoised reconstructed mesh relative to the initial state, and generate a deformation thermogram based on the strain energy density value;

[0009] Energy gradient features are extracted based on the deformation heatmap, and deformation regions and trends are determined according to the energy gradient features. When a deformation region is determined to exist, an early warning command is generated and the scanning parameters of the acquisition device are adjusted.

[0010] Preferably, the step of constructing a triangular mesh with a manifold topology based on the original measurement set and assigning virtual physical properties to the vertices of the triangular mesh includes:

[0011] The original measurement set is mapped to three-dimensional Euclidean space, and an initial topological connection is established through a triangulation algorithm to generate a triangular mesh model.

[0012] Each vertex of the triangular mesh model is defined as a mass point with a preset virtual mass, and the connecting edges between vertices are defined as spring connections with a preset virtual elastic coefficient.

[0013] A mapping relationship between reflection intensity data and elastic coefficient is established. The reflection intensity data is converted into weighting coefficients. The virtual elastic coefficients are then weighted and corrected using the weighting coefficients to obtain the triangular mesh with a manifold topology.

[0014] Preferably, the step of constructing a virtual contraction potential energy field and performing dynamic evolution iteration on the triangular mesh within the virtual contraction potential energy field includes:

[0015] A gravitational source is defined at the center of the tunnel cross-section, a global contraction force is generated pointing towards the gravitational source, and the Euclidean distance between the vertices of the triangular mesh and the corresponding nearest neighbor points in the original measurement set is calculated. A local repulsive force is generated based on the Euclidean distance.

[0016] The virtual contractile potential energy field is constructed by superimposing the global contractile force and the local repulsive force.

[0017] In the virtual contraction potential energy field, the resultant force vector of each vertex in the triangular mesh is calculated, and the spatial coordinates of each vertex are updated according to the resultant force vector to complete a single iteration;

[0018] Repeat the single iteration to calculate the total potential energy change rate of the system. When the total potential energy change rate is less than the preset equilibrium threshold, it is determined that the thermodynamic equilibrium state has been reached.

[0019] Preferably, the step of calculating the strain energy density value of the denoised reconstructed mesh relative to the initial state and generating a deformation thermogram based on the strain energy density value includes:

[0020] Obtain the current area and initial area of ​​each triangular cell in the denoised reconstructed mesh;

[0021] Based on the difference between the current area and the initial area, and in conjunction with the virtual elastic coefficient, the elastic potential energy of each triangular unit is calculated.

[0022] The elastic potential energy is normalized to obtain the strain energy density value, and the strain energy density value is mapped to the corresponding mesh position to generate the deformation heat map.

[0023] Preferably, the steps of extracting energy gradient features based on the deformation heatmap and determining the deformation region and deformation trend based on the energy gradient features include:

[0024] Frequency domain analysis was performed on the deformation thermogram to separate the high-frequency low-energy components from the low-frequency high-energy components.

[0025] The low-frequency high-energy component is used as an effective signal. Its energy gradient characteristics in spatial distribution are calculated, and the connected region area of ​​the low-frequency high-energy component is identified.

[0026] Set energy gradient threshold and area threshold;

[0027] If the energy gradient feature is greater than the energy gradient threshold, the corresponding location is determined to be a structural deformation region;

[0028] If the energy gradient feature is less than or equal to the energy gradient threshold, and the area of ​​the connected region is greater than the area threshold, then the corresponding location is determined to be a potential stress concentration region.

[0029] If the energy gradient feature is less than or equal to the energy gradient threshold, and the area of ​​the connected region is less than or equal to the area threshold, then the corresponding location is determined to be a surface roughness noise region.

[0030] Preferably, when a deformed area is determined to exist, the steps of generating an early warning command and adjusting the scanning parameters of the acquisition device include:

[0031] Based on the spatial location of the structural deformation region, calculate the inertial navigation correction amount of the acquisition device;

[0032] The attitude of the acquisition device is adjusted based on the inertial navigation correction, and the sampling frequency of laser scanning is increased for the structural deformation area;

[0033] A warning command containing the deformation location, deformation magnitude, and stress concentration direction is generated and sent to the monitoring terminal.

[0034] Preferably, the method further includes:

[0035] Using the spatial distribution characteristics of the strain energy density value, the geometric center of the deformation area is calculated with the strain energy density value as the weight, and this geometric center is determined as the center of the surrounding rock pressure applied outside the tunnel section.

[0036] Construct a surrounding rock pressure vector field pointing to the center of the surrounding rock pressure, and overlay the surrounding rock pressure vector field with the deformation thermogram for display.

[0037] A tunnel cross-section deformation monitoring system, comprising:

[0038] The data acquisition module is used to control the laser scanning equipment to acquire the original measurement set of the tunnel cross-section;

[0039] The manifold evolution calculation module is used to construct triangular meshes with manifold topology and virtual shrinking potential energy fields, and perform dynamic evolution to obtain denoised reconstructed meshes;

[0040] The deformation analysis module is used to calculate the strain energy density, generate a deformation thermogram, and determine the deformation region.

[0041] The feedback control module is used to generate early warning commands based on the deformation judgment results and adjust the scanning parameters of the data acquisition module.

[0042] An electronic device, comprising:

[0043] Memory and processor;

[0044] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of any one of the embodiments of a tunnel cross-section deformation monitoring method.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. This invention establishes a data quality-driven adaptive stiffness adjustment mechanism by constructing a virtual contraction potential energy field and introducing reflection intensity data to weight and correct the virtual elastic coefficients. In smooth areas with high reflection intensity, the virtual mesh stiffness is enhanced to strictly maintain the geometry; in dusty or damp areas with low reflection intensity, the virtual mesh stiffness is reduced to allow for greater smoothing and noise reduction. This mechanism overcomes the technical defect of traditional geometric filtering algorithms that tend to smooth out small deformation features when smoothing high dust or sprayed concrete surface noise, effectively filtering out unstructured high-frequency jitter noise while retaining physically meaningful structural micro-deformations.

[0047] 2. This invention abandons the traditional direct geometric distance comparison method and instead calculates the virtual elastic energy accumulated in the denoised reconstructed mesh from the initial state to the equilibrium state. Since energy and deformation have a nonlinear square relationship, this method can transform small geometric displacements into significant energy density values, thus amplifying early small deformations. By generating deformation heatmaps, it can intuitively reveal local tension concentration areas that are difficult to distinguish from simple displacement cloud maps, realizing the transformation from geometric observation to energy distribution observation, and greatly improving the ability to identify early tunnel defects.

[0048] 3. This invention combines frequency domain analysis, energy gradient characteristics, and connected domain area thresholds to establish a multi-level judgment logic. The system can distinguish between abrupt structural deformation and gradual potential stress concentration based on the spatial distribution of energy gradients, and accurately identify small-scale surface textures or construction unevenness as background noise by utilizing the physical relationship between connected domain area and lining thickness. This intelligent classification method based on physical scale effectively solves the problem of distinguishing between environmental interference and actual structural deformation on rough shotcrete surfaces, ensuring the reliability of monitoring results.

[0049] 4. This invention is not merely passive data processing; it can also generate early warning commands based on deformation judgment results and control the acquisition hardware in reverse. By calculating the inertial navigation correction, the system can automatically adjust the attitude of the acquisition equipment and increase the laser scanning sampling frequency for risk areas. This dynamic adjustment mechanism ensures that higher precision and more complete data details can be obtained immediately when an anomaly is detected, avoiding the risk of missing key information that may occur with traditional timed and fixed-point scanning, and significantly improving the emergency response speed and scientific decision-making of tunnel operation and maintenance. Attached Figure Description

[0050] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0051] Figure 1 This is a flowchart of the method of the present invention;

[0052] Figure 2 This is a structural diagram of the system of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0054] Example 1:

[0055] Please see Figure 1 A method for monitoring tunnel cross-sectional deformation, comprising:

[0056] Laser point cloud data and reflection intensity data of the tunnel cross section are collected as the original measurement set, and a triangular mesh with a manifold topology is constructed based on the original measurement set, and virtual physical properties are assigned to the vertices of the triangular mesh.

[0057] A virtual contraction potential energy field is constructed, and the triangular mesh is dynamically evolved and iterated in the virtual contraction potential energy field until the triangular mesh reaches a thermodynamic equilibrium state, thus obtaining a denoised reconstructed mesh.

[0058] Calculate the strain energy density of the denoised reconstructed mesh relative to the initial state, and generate a deformation thermogram based on the strain energy density.

[0059] Energy gradient features are extracted from deformation heatmaps. Deformation areas and trends are determined based on these energy gradient features. When a deformation area is identified, an early warning command is generated and the scanning parameters of the acquisition device are adjusted.

[0060] This embodiment provides a method for monitoring tunnel cross-section deformation. The method processes geometric measurement data by constructing a virtual physical field. The system executes a data acquisition step, and acquires laser point cloud data and reflection intensity data of the tunnel cross-section through a three-dimensional laser scanner LiDAR, i.e., the original measurement set. The original measurement set includes spatial coordinates and laser reflection intensity, and the manifold topology is represented by closed or semi-closed triangular meshes with local Euclidean spatial properties.

[0061] The system assigns virtual physical properties to the vertices of the triangular mesh, including virtual mass and virtual elastic coefficient, aiming to provide a physical parameter basis for subsequent dynamic simulations. A virtual contraction potential energy field is constructed, which simulates the combined effect of global gravity pointing towards the center of the tunnel and local repulsive forces from the original measurement points on the triangular mesh. Based on this, the system performs dynamic evolution iterations on the mesh in the potential energy field until the total potential energy of the system no longer changes significantly, that is, reaches a thermodynamic equilibrium state, thereby obtaining a denoised reconstructed mesh.

[0062] Instead of directly comparing geometric distances, the system calculates the virtual elastic energy accumulated by the mesh as it deforms from a stress-free initial state to the current equilibrium state, i.e., the strain energy density value, which reflects the degree of tension concentration in the local area; based on the deformation heat map, energy gradient features are extracted, and in response to the determination of the existence of deformation areas, early warning commands are generated and the scanning parameters of the acquisition equipment are adjusted;

[0063] This embodiment solves the technical problem that traditional geometric filtering algorithms often smooth out small deformation features when smoothing noise in high-dust tunnel environments by introducing a virtual physical field evolution mechanism. By utilizing the physical tension of the virtual mesh, this method can automatically adapt to the roughness of the tunnel surface and filter it out as high-frequency, low-energy noise. At the same time, it can treat structural micro-deformations as significant energy accumulation areas, thereby effectively reducing the false alarm rate caused by environmental interference without reducing the detection sensitivity.

[0064] The steps of constructing a triangular mesh with a manifold topology based on the original measurement set and assigning virtual physical properties to the vertices of the triangular mesh include:

[0065] The original measurement set is mapped to three-dimensional Euclidean space, and the initial topological connections are established through the triangulation algorithm to generate a triangular mesh model.

[0066] Each vertex of the triangular mesh model is defined as a mass point with a preset virtual mass, and the connecting edges between vertices are defined as spring connections with preset virtual elastic coefficients.

[0067] A mapping relationship between reflection intensity data and elastic coefficients is established. The reflection intensity data is converted into weighting coefficients. The virtual elastic coefficients are then weighted and corrected using the weighting coefficients to obtain a triangular mesh with a manifold topology.

[0068] This embodiment details the process of constructing the mesh and assigning attributes. The system uses the Delaunay triangulation algorithm or the rolling sphere method to connect the discrete original measurement set point cloud into continuous triangular patches, establishing an initial topological connection. To ensure that the subsequent virtual shrinkage field can correctly play the role of shrinkage envelope and avoid the force direction disorder caused by the mesh being interspersed between data points in the initial state, this embodiment performs a preprocessing expansion step after generating the initial triangular mesh model: calculating the normal vector of each vertex. The specific algorithm is as follows: traversing all triangular patches to which the vertex belongs, calculating the normal vector of each patch, and performing a weighted summation with the patch area as the weight. Finally, the result vector is normalized to ensure the geometric continuity of the expansion direction, and the vertex is translated outward along the normal by a preset buffer distance, such as 0.1 meters to 0.3 meters, thereby forming an initial envelope surface that completely surrounds the original measurement set.

[0069] This step ensures that the mesh is located outside the data points at the beginning of the dynamic evolution, allowing the global contraction force and local repulsion force to form an effective mechanical balance. Each vertex of the mesh model is defined as a point mass with a preset virtual mass, and the connecting edges are defined as springs with preset virtual elastic coefficients. To fully utilize the surface material characteristics contained in the laser reflection intensity information, the system establishes a mapping relationship between reflection intensity data and elastic coefficients. Specifically, the corrected virtual elastic coefficients of the connecting edges are calculated using the following formula:

[0070]

[0071] in, The corrected virtual elastic coefficient is derived from calculation results and its physical meaning is the actual stiffness of the connecting edge, with units of Newtons per meter. : Basic elastic coefficient, sourced from preset values, physical meaning is the default stiffness before correction, unit is Newtons per meter; Intensity influence factor, which is a preset constant, is recommended to be 0.2 to 0.8, preferably 0.5. It can be set through experimental calibration or experience. Physically, it is a coefficient that adjusts the weight of the influence of reflection intensity on stiffness, and is dimensionless. : Weighting coefficient, derived from reflection intensity calculation, physically represents the stiffness gain ratio based on material reflectivity, dimensionless;

[0072] The formula for calculating the weighting coefficient is as follows:

[0073]

[0074] in, Vertex reflection intensity value, sourced from laser scanner acquisition, physically represents the original laser echo intensity readings at both ends of the connecting edge, and is in the intensity unit defined by the device, such as a grayscale value of 0-255 or a 16-bit integer value of 0-65535; : Maximum reflection intensity scale value, which is derived from equipment parameters. Its physical meaning is the upper limit of the sensor's intensity range, such as 255 for 8-bit data or 65535 for 16-bit data. It is used as a normalized reference constant, and the unit is a dimensionless digital quantity.

[0075] To enable those skilled in the art to better implement this solution, preferred examples of parameter selection are provided below: basic elastic coefficient The recommended range of values ​​is: to In practical implementation, 200 Newtons per meter can be used; Intensity Influence Factor It is recommended to set it between 0.2 and 0.8, and 0.5 can be used in specific implementations; these values ​​are determined based on experiments on the laser reflection characteristics of typical concrete tunnel surfaces, which can ensure that the mesh achieves the best balance between maintaining geometric details and smoothing and denoising.

[0076] This embodiment achieves data quality-driven adaptive stiffness adjustment by introducing reflection intensity to weight the virtual elastic coefficient. In areas with high reflection intensity, such as smooth concrete surfaces, the virtual mesh stiffness is enhanced to strictly maintain the geometry and prevent excessive smoothing. In areas with low reflection intensity, such as damp or dusty areas, the virtual mesh stiffness is reduced, allowing for greater smoothing and noise reduction. This mechanism significantly improves the reconstruction accuracy and robustness for different surface material areas in complex tunnel environments.

[0077] The steps for constructing a virtual contraction potential energy field and performing dynamic evolution iterations on the triangular mesh within that field include:

[0078] A gravitational source is defined at the center of the tunnel cross-section, a global contraction force is generated pointing towards the gravitational source, and the Euclidean distance between the vertices of the triangular mesh and the corresponding nearest neighbor points in the original measurement set is calculated. A local repulsive force is generated based on the Euclidean distance.

[0079] By superimposing global contraction force and local repulsion force, a virtual contraction potential energy field is constructed;

[0080] In the virtual contraction potential energy field, the resultant force vector of each vertex in the triangular mesh is calculated, and the spatial coordinates of each vertex are updated according to the resultant force vector to complete a single iteration;

[0081] Repeat the single iteration to calculate the rate of change of the total potential energy of the system. When the rate of change of the total potential energy is less than the preset equilibrium threshold, it is determined that the thermodynamic equilibrium state has been reached.

[0082] This embodiment details the process of constructing the potential energy field and the dynamic evolution. The system defines a gravitational source at the center of the tunnel cross-section, generating a global contraction force pointing towards this center to simulate the adsorption effect of the contraction membrane. Simultaneously, it calculates the Euclidean distance between the grid vertex and the corresponding nearest neighbor point in the original measurement set, and generates a local repulsive force based on this distance, simulating the blocking effect of the data boundary. The global contraction force, the local repulsive force, and the internal spring force from adjacent vertices are superimposed to construct a virtual contraction potential energy field. To ensure the physical realism of the dynamic evolution, the system uses the following formula to calculate the force acting on any vertex of the grid. The resultant force vector on :

[0083]

[0084] The specific calculation models for each force are as follows:

[0085]

[0086]

[0087]

[0088] in, Represents the relationship between vertices in the mesh. The set of directly connected first-order adjacent vertices; Global contraction coefficient, derived from preset parameters, physically represents constant adsorption tension, measured in Newtons, with a recommended value range of [value missing]. to In specific implementation, it is advisable to adopt ; : Coordinates of the center of the tunnel cross section; The local repulsion strength coefficient, derived from preset parameters, is the physical constant of the force field that generates repulsion, measured in Newton-square meters. A recommended value range is [insert range here]. to ; Regularization parameter, derived from a preset small quantity, physically means to prevent numerically stable terms with a denominator of zero. The unit is meters, and a recommended value is [value missing]. ; : Euclidean distance from the vertex to the nearest measurement point, in meters, used as a scalar input when calculating the magnitude of the repulsive force; The unit direction vector along the measurement point towards the grid vertex is calculated using the following formula: ;

[0089] To prevent numerical singularity, when the distance When the vertex is in meters, the current normal vector of that vertex is forcibly taken as the direction; The effective cutoff distance for the repulsive force is calculated from a preset value and is in meters. It is recommended to set it to 3 to 5 times the average spacing between points in the original measurement set, for example, 0.05 meters, in order to construct an effective repulsive buffer zone.

[0090] To ensure the adaptability of this parameter, the average point spacing is obtained by building a KD-Tree spatial index for the original measurement set, randomly sampling at least 1000 points, and calculating the arithmetic mean of their respective nearest neighbor distances. Connecting edges The virtual elasticity coefficient; Connecting edges The initial natural length is the side length at the initial moment, which remains constant throughout the evolution process. To prevent mesh degradation or overlap during shrinkage, this embodiment sets this value to the length of the corresponding connecting edge in the initial triangular mesh model, i.e. This causes the spring force to tend to maintain the initial topological shape of the mesh, relying on global contraction force. The entire mesh moves inward, rather than relying on spring contraction; : Mesh vertex index subscript; : Indicates connection to vertices and The edge; : Minimum value index;

[0091] In this resultant force field, the system uses a semi-implicit Euler method to update its position coordinates. To address the potential perpetual oscillations of a conservative force field system composed of spring potential energy and gravitational potential energy under dissipation-free conditions, this embodiment explicitly introduces a non-conservative numerical damping mechanism. Kinetic energy dissipation is achieved by multiplying the sum of the previous velocity and the current acceleration by a velocity decay coefficient, ensuring the system converges to the static equilibrium point. The specific update formula is as follows:

[0092]

[0093]

[0094] in, Vertex velocity vector, derived from iterative calculations, physically represents the velocity of a particle, measured in meters per second; Force per unit mass, i.e., acceleration vector, where The total driving force acting on the particle. For virtual quality; Vertex position coordinate vector, sourced from iterative updates, physically represents the spatial position of a particle, and is measured in meters; : Iteration step index, representing the current time step and the next time step respectively; : The resultant force vector, derived from the above force field superposition calculation, physically represents the total driving force on a particle, and its unit is Newton; Time step, sourced from a preset value, physically represents the time interval between a single iteration, and is measured in seconds; Velocity damping coefficient, derived from a preset constant, physically represents the viscous resistance of the simulating medium to eliminate oscillations; it is dimensionless and its value range is [value missing]. Preferred ;

[0095] The system repeats the above iterations and calculates the rate of change of the system's total potential energy. To accurately quantify the system's energy state and determine convergence, this embodiment explicitly defines the system's total potential energy. The calculation formula is as follows, which strictly corresponds to the potential energy integral of each of the above conservative force fields, where the local repulsive potential energy is the original function of the local repulsive force, and a cutoff distance criterion is introduced to ensure consistency with the definition of the force field:

[0096]

[0097] in, The term represents the local repulsive potential energy and needs to be combined with the formula for the repulsive force field. The integral form remains consistent, that is The definition is as follows: when hour,

[0098]

[0099] when hour,

[0100]

[0101] That is, constant potential energy, corresponding to the zero-force region, ensuring the continuity of the energy function and the consistency of the force field;

[0102] Based on this total potential energy, calculate the rate of change of total potential energy. :

[0103]

[0104] Among them, superscript and These represent the current iteration step and the previous iteration step, respectively; in response to the rate of change of total potential energy. Less than the preset balance threshold, for example The system determines that the grid has reached thermodynamic equilibrium and stops evolving;

[0105] This embodiment simulates the adsorption behavior of the elastic film on the tunnel wall using the dynamic evolution process; the global contraction force ensures the large deformation fitting ability of the mesh, while the local repulsion force ensures the fidelity of the measurement data; through this physical field balance mechanism, the denoised mesh naturally has continuity, effectively eliminating high-frequency jitter noise in the discrete point cloud, and providing a high-quality, smooth reference surface for subsequent micro-deformation analysis.

[0106] The steps of calculating the strain energy density of the denoised reconstructed mesh relative to the initial state and generating a deformation thermogram based on the strain energy density include:

[0107] Obtain the current area and initial area of ​​each triangle cell in the denoised reconstructed mesh;

[0108] Based on the difference between the current area and the initial area, and combined with the virtual elastic coefficient, the elastic potential energy of each triangular unit is calculated.

[0109] The elastic potential energy is normalized to obtain the strain energy density value, and the strain energy density value is mapped to the corresponding mesh position to generate a deformation heat map.

[0110] This embodiment details how to convert geometric shapes into physical energy indices; the system obtains the current area of ​​each triangular element in the denoised reconstructed mesh and retrieves the initial reference area under the assumption of no deformation; based on the difference between the current area and the initial area, combined with thin film strain theory, the elastic potential energy of each triangular element is calculated, as shown in the following formula:

[0111]

[0112] in, Elastic potential energy, derived from calculations, is the energy stored in a unit due to deformation, measured in joules. The average virtual elastic coefficient is derived from the average of the coefficients of the three sides of the element. Its physical meaning is the average stiffness of the element, and its unit is Newtons per meter. : Current area, sourced from equilibrium mesh calculation, physically refers to the area of ​​the deformed element, in square meters; Initial reference area, which is the area of ​​ideal grid cells generated based on standard tunnel design parameters, or the area of ​​the denoised and reconstructed grid in a naturally relaxed state after removing external force field constraints. Physically, it is the geometric area of ​​the grid in a state without deformation stress, and the unit is square meters.

[0113] To eliminate the inconsistency in the effect of mesh cell area size on deformation measurement, the system further calculates the energy distribution per unit area, i.e., strain energy density:

[0114]

[0115] in, Strain energy density in a physical sense, with units of The system will use physical strain energy density Normalization is performed; to avoid false-color rendering anomalies caused by numerical calculation noise in flat, undeformed tunnel areas, i.e., when At this time, minute numerical fluctuations are stretched to the full color gamut. This embodiment introduces a minimum dynamic range threshold mechanism: calculating the global dynamic range. ,like ,in, For example, a preset noise threshold. This forces the normalization value of all grid cells. Otherwise, execute the following specific normalization formula:

[0116]

[0117] in, and These are the global minimum and global maximum values ​​of strain energy density calculated by traversing all triangular elements in the current denoised and reconstructed mesh. This calculation range covers the entire mesh model of the current section, ensuring the global consistency of the normalized benchmark. The numerical stability constant is defined, and its unit is the same as that of strain energy density. Maintain consistency, that is The recommended value is The dimensionless normalized strain energy density is obtained through this calculation. ; symbol here This is to unify the signs of input variables in subsequent frequency domain analysis and gradient calculation, and to convert physical quantities into relative intensity values ​​required for signal processing, thereby preserving the deformation distribution characteristics while adapting to the requirements of image processing algorithms;

[0118] Normalized strain energy density value Mapped to the corresponding grid positions, the final rendered heatmap is generated. The rendering process is not simply a numerical display; instead, it employs a piecewise linear interpolation algorithm to normalize the values. Convert to RGB color values: Settings Blue indicates the safe zone. The green area represents the transition zone. The color is red, indicating the danger zone. The intermediate value is calculated using linear interpolation to generate a pseudo-color cloud map that intuitively reflects the strain energy gradient.

[0119] This embodiment transforms traditional geometric displacement observation into energy distribution observation by calculating strain energy density. Since energy is proportional to the square of deformation, this method has a nonlinear amplification effect on small deformations. In tunnel monitoring scenarios, this energy perspective can significantly enhance the identification of early small deformations, such as local stress release caused by microcracks, and solve the problem that simple displacement cloud maps cannot distinguish between overall displacement and local bulges.

[0120] The steps for extracting energy gradient features from deformation heatmaps and determining deformation regions and trends based on these energy gradient features include:

[0121] Frequency domain analysis was performed on the deformation thermogram to separate the high-frequency low-energy components from the low-frequency high-energy components.

[0122] Using low-frequency high-energy components as effective signals, calculate their energy gradient characteristics in spatial distribution and identify the connected region area of ​​low-frequency high-energy components.

[0123] Set energy gradient threshold and area threshold;

[0124] If the energy gradient feature is greater than the energy gradient threshold, the corresponding location is determined to be a structural deformation region;

[0125] If the energy gradient feature is less than or equal to the energy gradient threshold, and the area of ​​the connected region is greater than the area threshold, then the corresponding location is determined to be a potential stress concentration region.

[0126] If the energy gradient feature is less than or equal to the energy gradient threshold, and the area of ​​the connected region is less than or equal to the area threshold, then the corresponding location is determined to be a surface roughness noise region.

[0127] This embodiment details how to perform intelligent discrimination based on energy characteristics; it utilizes graph Fourier transform to perform frequency domain analysis on deformed thermograms; the specific execution steps are as follows: constructing a combined Laplace matrix of triangular meshes. ;in, Let be a degree matrix, and its diagonal elements are defined as follows: Off-diagonal elements are 0; adjacency matrix A Gaussian kernel function is used to construct a structure that reflects the local geometric weights of the mesh, i.e., if the vertex... If connected, then To ensure that frequency domain analysis can accurately capture the characteristics of the deformed manifold, the formula... It is explicitly defined as a denoised and reconstructed mesh, that is, the mesh after reaching thermodynamic equilibrium, corresponding to the Euclidean distance between vertices, rather than the initial mesh distance; Set to the arithmetic mean of the lengths of all connected edges in the denoised mesh;

[0128] calculate eigenvector matrix diagonal matrix with eigenvalues Normalized strain energy density signal at the grid vertices Projecting onto the frequency domain yields the spectral coefficients. According to the set cutoff frequency parameters Constructing an ideal low-pass filter mask To resolve the mapping relationship between parameter definitions and calculation formulas, this embodiment explicitly defines... The spectral cutoff ratio is the proportion of low-frequency components retained in the total spectrum; a recommended value is [value to be filled in]. ;

[0129] Specifically, first sort the eigenvalues ​​in ascending order. According to parameters Calculate the cutoff threshold and select the first... There are eigenvalues, among which... The total number of eigenvalues ​​corresponds to the low-frequency portion and serves as the passband range, i.e., when the eigenvalue index... At that time, diagonal matrix elements The rest are Using this mathematical definition, the coefficients corresponding to low frequencies are retained, while the high-frequency coefficients are set to zero. An inverse graphical Fourier transform is then performed to reconstruct the spatial signal, yielding the low-frequency, high-energy components. This component is the effective deformed signal after filtering out random noise;

[0130] In this frequency domain analysis context, separating high-frequency low-energy components from low-frequency high-energy components is specifically achieved through low-pass filtering. The low-frequency high-energy components refer to the retained low-frequency signals that constitute the majority of the energy in the spectrum, corresponding to continuous structural deformations in space. The high-frequency low-energy components, on the other hand, refer to the filtered-out noise signals, typically with small amplitudes and drastic variations, corresponding to surface roughness. This step utilizes normalized data... The relative difference solves the problem of the lack of absolute energy scale, maintains the consistency of the physical basis in logic, and thus successfully separates the effective signal representing the real deformation;

[0131] To simultaneously identify both minute cracks and overall deformation, the system performs dual-channel feature extraction: on the one hand, it calculates the normalized strain energy density value. The energy gradient characteristics in the spatial distribution are used to capture local energy abrupt changes; on the other hand, the low-frequency high-energy components reconstructed above are utilized. Identify the area of ​​connected components to determine the spatial extent of the deformation; adjust the formula for calculating the energy gradient characteristics to be based on the original signal. ;

[0132] The calculation formula is:

[0133]

[0134] in, For unit Energy gradient characteristics; Representation of triangular units A set of adjacent cells that share a common edge; This is the normalized strain energy density value, i.e., the normalized dimensionless value; The geodesic distance, in meters, between the geometric centers of adjacent units is calculated on the triangular mesh using Dijkstra's algorithm. Simultaneously, to quantify the spatial extent of deformation to support subsequent classification, the system employs an adaptive threshold segmentation method, for example, setting... ,in, for The mean and standard deviation of the signal are used to binarize the signal, and the flooding algorithm or disjoint-set data structure algorithm is used to identify the connected components in the region above the threshold, and the geometric area of ​​the connected components is calculated.

[0135] Based on this, the system sets energy gradient threshold and area threshold; given the denominator in the aforementioned formula... The calculated energy gradient characteristics have the dimension of length. It must have the reciprocal dimension of length, that is... To address the issue of inconsistent dimensions and to assign a clear physical meaning to the threshold, this embodiment provides a set of standards applicable to normalized signals. Threshold setting scheme: Set the energy gradient threshold to That is, the change in normalized strain energy density per meter of geodesic distance reaches 5%; regarding the area threshold, in order to ensure the uniformity of the technical solution and avoid implementation confusion that may be caused by fixed values, this embodiment specifies an area threshold. The calculation should be based on the physical properties of the tunnel lining, and the specific calculation formula is shown below, rather than using a single fixed empirical value and executing multi-level judgment logic.

[0136] When the energy gradient feature is greater than the energy gradient threshold, the system determines that the corresponding location is a structural deformation region, such as a crack or a fault. Such deformation has the characteristic of abrupt energy change.

[0137] When the energy gradient feature is less than or equal to a threshold, but the area of ​​the connected domain is greater than the area threshold, the system determines the corresponding location as a potential stress concentration region. The physical basis for this determination is that a low gradient feature indicates that the deformation is spatially gradual and continuous, rather than an abrupt fracture. When the area of ​​the connected domain for such gradual deformation is sufficiently large, exceeding the structural feature scale, according to the principles of continuum mechanics, this corresponds to the overall elastic deformation or generalized rheology of the lining structure, such as cross-sectional convergence caused by surrounding rock compression, i.e., macroscopic potential stress concentration. To ensure the objectivity of this determination, an area threshold is used. The calculation formula is:

[0138]

[0139] in, : Structural influence coefficient, dimensionless, used to adjust the threshold for determining the area of ​​the deformed region. It is recommended to take a value of 1.0~3.0, with 1.5 being preferred. The design thickness of the tunnel lining is, for example, 0.4 meters. The physical derivation of this calculation formula is based on Saint-Venant's principle of elasticity and the theory of thick-walled cylinders: in tunnel lining structures, local stress disturbances caused by surface defects or roughness typically only occur at a thickness equivalent to the structural thickness. The characteristic scale is significant within a certain range, and decays rapidly with increasing distance; conversely, if the area of ​​the connected domain of the deformed region is... satisfy That is, the corresponding feature scale This indicates that the deformation has crossed the boundary of local disturbance and belongs to generalized rheological or bending deformation caused by surrounding rock pressure or structural instability.

[0140] coefficient The value of was calibrated through numerous tunnel model tests: when At times, the false alarm rate is relatively high, that is, construction traces are misjudged as deformation; when When the sensitivity decreases, take Able to distinguish surface textures, usually , and structural bulges, usually The optimal balance is achieved between these factors. According to the formula, for a typical 0.4-meter-thick lining, the area threshold is approximately 0.24 square meters. This means that only gentle deformations with a spatial distribution range exceeding the square of the lining thickness are considered to be stress accumulation with structural mechanical significance, thus effectively distinguishing the actual surrounding rock pressure effect.

[0141] When both the energy gradient feature and the area of ​​the connected region are less than or equal to their respective thresholds, the system determines the corresponding location as a surface roughness noise region; this is because within a range smaller than the structural feature scale, i.e., the area is smaller than... Low-gradient energy fluctuations typically correspond to construction textures on concrete surfaces, local unevenness in shotcrete, or statistical fluctuations in laser spots. They lack the geometric continuity conditions to form an effective structural stress field, and are therefore treated as background noise. By introducing the spatial dimension index of connected domain area and combining it with the principle of lining thickness, this method effectively solves the logical defect that energy values ​​alone cannot distinguish between large-scale compression and local unevenness, and achieves intelligent and accurate identification of deformation types.

[0142] When a deformed area is determined to exist, the steps for generating an early warning command and adjusting the scanning parameters of the acquisition device include:

[0143] Calculate the inertial navigation correction amount of the acquisition device based on the spatial location of the structural deformation area;

[0144] The attitude of the acquisition device is adjusted based on the inertial navigation correction, and the sampling frequency of laser scanning is increased for structurally deformable areas;

[0145] Generate early warning instructions that include the deformation location, deformation magnitude, and stress concentration direction, and send the early warning instructions to the monitoring terminal.

[0146] This embodiment involves a closed-loop feedback control process; the system calculates the inertial navigation correction amount of the acquisition device based on the spatial location of the identified structural deformation region; this calculation process is not a direct numerical mapping, but is based on the principle of geometric line-of-sight tracking, and the specific steps include: extracting the geometric centroid coordinates of the mesh vertices of the structural deformation region. ; Obtain the current coordinates of the optical axis origin of the acquisition device. and the current attitude angle, i.e., yaw angle Pitch angle Construct the target line-of-sight vector from the device to the deformation center. ; target line-of-sight vector Convert to the ideal target angle in spherical coordinates:

[0147]

[0148]

[0149] Calculate the inertial navigation correction: , The inertial navigation correction here specifically refers to the angle control increment required to eliminate the deviation between the current inertial attitude and the target observation attitude, rather than the calibration of the drift error of the inertial navigation system itself.

[0150] Based on this correction amount The system drives the hardware to adjust the posture of the acquisition device, such as controlling the rotation of the laser scanner's pan-tilt head to precisely align its optical axis with the center of the deformation area. At the same time, for this structurally deformed area, the system automatically increases the sampling frequency of the laser scan, such as improving the angular resolution. It generates an early warning command that includes the deformation location, deformation magnitude, and stress concentration direction, and sends the command to the monitoring terminal in real time.

[0151] This embodiment realizes real-time closed-loop control from monitoring to analysis to optimization; unlike the traditional blind scan mode, this method can dynamically adjust hardware parameters based on preliminary analysis results and conduct in-depth investigations of suspicious areas; this mechanism greatly improves the data accuracy and early warning reliability of key risk areas without significantly increasing the overall data volume, and ensures rapid response capability to sudden deformations;

[0152] The method also includes:

[0153] Using the spatial distribution characteristics of strain energy density values, the geometric center of the deformation area is calculated with strain energy density values ​​as weights, and this geometric center is determined as the center of the surrounding rock pressure applied outside the tunnel section.

[0154] Construct a surrounding rock pressure vector field pointing towards the center of the surrounding rock pressure, and overlay the surrounding rock pressure vector field with the deformation heat map for display.

[0155] This embodiment provides an inversion analysis function based on monitoring data; utilizing the spatial distribution characteristics of strain energy density values, the geometric center of the deformation area is calculated using this value as a weight, and it is determined as the center of the surrounding rock pressure applied outside the tunnel cross-section. The calculation formula is as follows:

[0156]

[0157] in, : Coordinates of the center of pressure on the surrounding rock, derived from weighted calculation, physically representing the equivalent point of application of external pressure, in meters; The geometric center coordinates of a triangular element are derived from grid computing and physically represent the spatial location of the element, measured in meters. The strain energy density value refers to the physical strain energy density calculated above, or a normalized value. Alternatively, this embodiment preferably uses physical values. To maintain the dimensional meaning, the source is the preceding steps, the physical meaning is the weighting factor, and the unit is joules per square meter;

[0158] The system constructs a surrounding rock pressure vector field pointing towards the center of the surrounding rock pressure; specifically, for each cell in the grid... Generate a pressure indication vector Its definition The direction of this vector is from the center of the unit to the center of the surrounding rock pressure. The modulus is proportional to the strain energy density of the unit, thus intuitively representing the contribution of the stress in each local area to the overall pressure center. This vector field is displayed as an arrow layer overlaid with the deformation thermogram to intuitively present the stress state.

[0159] This embodiment realizes the inversion analysis of external surrounding rock pressure derived from internal wall deformation; by visualizing the center and direction of the surrounding rock pressure, engineers can not only locate the deformation position inside the tunnel, but also intuitively understand the source direction of the external pressure; this visualized mechanical basis has important guiding significance for formulating targeted grouting reinforcement or support schemes, and effectively improves the scientific nature of tunnel maintenance decisions.

[0160] Example 2:

[0161] Please see Figure 2 A tunnel cross-section deformation monitoring system, comprising:

[0162] The data acquisition module is used to control the laser scanning equipment to acquire the original measurement set of the tunnel cross-section;

[0163] The manifold evolution calculation module is used to construct triangular meshes with manifold topology and virtual shrinking potential energy fields, and perform dynamic evolution to obtain denoised reconstructed meshes;

[0164] The deformation analysis module is used to calculate the strain energy density, generate a deformation thermogram, and determine the deformation region.

[0165] The feedback control module is used to generate early warning commands based on the deformation judgment results and adjust the scanning parameters of the data acquisition module.

[0166] This embodiment provides a tunnel cross-section deformation monitoring system. The system is based on a modular architecture design, which aims to transform the aforementioned theoretical algorithms into an engineering-deployable physical system. The system mainly consists of a data acquisition module, a manifold evolution calculation module, a deformation analysis module, and a feedback control module. Each module interacts with data through a high-bandwidth internal bus or network protocol.

[0167] Data Acquisition Module: This module serves as the system's sensing front end, directly interfacing with the laser scanning hardware interface. Specifically, it integrates SDK drivers for multiple brands of LiDAR, enabling real-time acquisition of laser point cloud data and reflection intensity data from the tunnel cross-section. The module incorporates a time synchronization subsystem, employing PTP to ensure microsecond-level timestamp accuracy for the point cloud data. Furthermore, this module is responsible for establishing a spatial index for the original measurement set, providing a rapid nearest neighbor search service for subsequent calculations—the foundation for generating local repulsive forces.

[0168] Manifold Evolution Computation Module: This module is the core computing unit of the system, responsible for performing high-density physical simulation calculations. Considering that dynamic evolution iteration involves a large number of matrix operations, this module adopts a GPU parallel computing architecture in this embodiment. It maps the original measurement set to video memory, quickly constructs a triangular mesh with a manifold topology, and calculates the resultant force vector of each vertex in parallel. This module has a built-in physics engine solver, uses the semi-implicit Euler method for position updates, and monitors the total potential energy change rate of the system in real time. When the change rate is lower than a preset threshold, the module outputs a denoised reconstructed mesh data stream that has reached thermodynamic equilibrium.

[0169] Deformation Analysis Module: This module receives denoised mesh data and is primarily responsible for converting geometric information into energy information. It implements the calculation logic for strain energy density by comparing the current area of ​​the mesh cells with the initial area to calculate the elastic potential energy. This module integrates a frequency domain analysis engine, which uses graphical Fourier transform to separate high-frequency noise from low-frequency deformation signals and intelligently identifies the deformation type based on energy gradient characteristics. In addition, this module also includes a rendering engine, which is responsible for mapping the normalized strain energy density into a pseudo-color deformation heatmap and overlaying the surrounding rock pressure vector field to present it to the user in an intuitive visualization form.

[0170] Feedback Control Module: This module implements the closed-loop control function of the system. When the deformation analysis module determines the existence of a structural deformation region, this module responds immediately and calculates the inertial navigation correction of the acquisition equipment. Specifically, it calculates the target yaw angle based on the centroid coordinates of the deformation region. and pitch angle The module generates motor drive commands through a PID control algorithm to adjust the scanner's posture to align with abnormal areas. Simultaneously, it automatically issues commands to modify scanning parameters, such as reducing the scanning speed to increase point cloud density, and generates early warning data packets containing deformation magnitude and direction, which are sent to the remote monitoring terminal via TCP / IP protocol.

[0171] Example 3:

[0172] An electronic device, comprising:

[0173] Memory and processor;

[0174] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of any one of the embodiments of a tunnel cross-section deformation monitoring method.

[0175] This embodiment provides an electronic device as the hardware carrier of the above-mentioned tunnel cross-section deformation monitoring method;

[0176] Hardware architecture: The electronic device includes at least one processor and memory;

[0177] Processor: Considering the high computational requirements of dynamic evolution iteration and frequency domain analysis, the processor preferably adopts a heterogeneous computing architecture, including a high-performance central processing unit for logic control and data scheduling, and a graphics processing unit or tensor processing unit for accelerating large-scale matrix operations and parallel physical simulation.

[0178] Memory: Memory is divided into volatile memory (RAM) and non-volatile memory; RAM capacity needs to meet the storage requirements of large-scale triangular mesh models and temporary physical attribute data generated during the evolution process; non-volatile memory stores computer-executable instructions;

[0179] Communication interface: The device is equipped with a high-speed I / O interface for connecting to a laser scanner to receive raw measurement sets, and has a network communication module for sending early warning commands;

[0180] Software instruction execution logic:

[0181] When the processor executes instructions from memory, it reads the laser point cloud from the I / O interface and constructs a virtual contraction potential energy field in memory;

[0182] The processor calls the physical simulation instruction set, adjusts the virtual elastic coefficients based on the reflection intensity data, and performs iterative calculations until the system potential energy converges;

[0183] The processor executes analysis instructions, calculates strain energy density, and identifies deformation properties through threshold judgment logic based on energy gradient characteristics;

[0184] If a risk is identified, the processor calculates the inertial navigation correction amount and sends hardware adjustment commands to the external scanning device through the control interface to achieve automated monitoring closed loop;

[0185] The electronic device can be an industrial control computer deployed on a tunnel inspection vehicle, an embedded edge computing node, or a back-end analysis server. As long as it has the ability to execute the above-mentioned tunnel cross-section deformation monitoring method steps, it is included within the protection scope of this embodiment.

[0186] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for monitoring tunnel cross-sectional deformation, characterized in that, include: Laser point cloud data and reflection intensity data of the tunnel cross section are collected as the original measurement set, and a triangular mesh with a manifold topology is constructed based on the original measurement set, and virtual physical properties are assigned to the vertices of the triangular mesh. A virtual contraction potential energy field is constructed, and the triangular mesh is subjected to dynamic evolution iteration in the virtual contraction potential energy field until the triangular mesh reaches a thermodynamic equilibrium state, thereby obtaining a denoised reconstructed mesh. Calculate the strain energy density value of the denoised reconstructed mesh relative to the initial state, and generate a deformation thermogram based on the strain energy density value; Energy gradient features are extracted based on the deformation heatmap, and deformation regions and trends are determined according to the energy gradient features. When a deformation region is determined to exist, an early warning command is generated and the scanning parameters of the acquisition device are adjusted.

2. The method for monitoring tunnel cross-section deformation as described in claim 1, characterized in that, The steps of constructing a triangular mesh with a manifold topology based on the original measurement set and assigning virtual physical properties to the vertices of the triangular mesh include: The original measurement set is mapped to three-dimensional Euclidean space, and an initial topological connection is established through a triangulation algorithm to generate a triangular mesh model. Each vertex of the triangular mesh model is defined as a mass point with a preset virtual mass, and the connecting edges between vertices are defined as spring connections with a preset virtual elastic coefficient. A mapping relationship between reflection intensity data and elastic coefficient is established. The reflection intensity data is converted into weighting coefficients. The virtual elastic coefficients are then weighted and corrected using the weighting coefficients to obtain the triangular mesh with a manifold topology.

3. The method for monitoring tunnel cross-sectional deformation as described in claim 2, characterized in that, The steps of constructing a virtual contraction potential energy field and performing dynamic evolution iterations on the triangular mesh within the virtual contraction potential energy field include: A gravitational source is defined at the center of the tunnel cross-section, a global contraction force is generated pointing towards the gravitational source, and the Euclidean distance between the vertices of the triangular mesh and the corresponding nearest neighbor points in the original measurement set is calculated. A local repulsive force is generated based on the Euclidean distance. The virtual contractile potential energy field is constructed by superimposing the global contractile force and the local repulsive force. In the virtual contraction potential energy field, the resultant force vector of each vertex in the triangular mesh is calculated, and the spatial coordinates of each vertex are updated according to the resultant force vector to complete a single iteration; Repeat the single iteration to calculate the total potential energy change rate of the system. When the total potential energy change rate is less than the preset equilibrium threshold, it is determined that the thermodynamic equilibrium state has been reached.

4. The method for monitoring tunnel cross-section deformation as described in claim 3, characterized in that, The steps of calculating the strain energy density of the denoised reconstructed mesh relative to the initial state and generating a deformation thermogram based on the strain energy density include: Obtain the current area and initial area of ​​each triangular cell in the denoised reconstructed mesh; Based on the difference between the current area and the initial area, and in conjunction with the virtual elastic coefficient, the elastic potential energy of each triangular unit is calculated. The elastic potential energy is normalized to obtain the strain energy density value, and the strain energy density value is mapped to the corresponding mesh position to generate the deformation heat map.

5. The method for monitoring tunnel cross-sectional deformation as described in claim 4, characterized in that, The steps of extracting energy gradient features based on the deformation heatmap and determining the deformation region and deformation trend based on the energy gradient features include: Frequency domain analysis was performed on the deformation thermogram to separate the high-frequency low-energy components from the low-frequency high-energy components. The low-frequency high-energy component is used as an effective signal. Its energy gradient characteristics in spatial distribution are calculated, and the connected region area of ​​the low-frequency high-energy component is identified. Set energy gradient threshold and area threshold; If the energy gradient feature is greater than the energy gradient threshold, the corresponding location is determined to be a structural deformation region; If the energy gradient feature is less than or equal to the energy gradient threshold, and the area of ​​the connected region is greater than the area threshold, then the corresponding location is determined to be a potential stress concentration region. If the energy gradient feature is less than or equal to the energy gradient threshold, and the area of ​​the connected region is less than or equal to the area threshold, then the corresponding location is determined to be a surface roughness noise region.

6. The method for monitoring tunnel cross-section deformation as described in claim 5, characterized in that, When a deformed area is determined to exist, the steps for generating an early warning command and adjusting the scanning parameters of the acquisition device include: Based on the spatial location of the structural deformation region, calculate the inertial navigation correction amount of the acquisition device; The attitude of the acquisition device is adjusted based on the inertial navigation correction, and the sampling frequency of laser scanning is increased for the structural deformation area; A warning command containing the deformation location, deformation magnitude, and stress concentration direction is generated and sent to the monitoring terminal.

7. The method for monitoring tunnel cross-section deformation as described in claim 6, characterized in that, The method further includes: Using the spatial distribution characteristics of the strain energy density value, the geometric center of the deformation area is calculated with the strain energy density value as the weight, and this geometric center is determined as the center of the surrounding rock pressure applied outside the tunnel section. Construct a surrounding rock pressure vector field pointing to the center of the surrounding rock pressure, and overlay the surrounding rock pressure vector field with the deformation thermogram for display.

8. A tunnel cross-section deformation monitoring system, using the method described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to control the laser scanning equipment to acquire the original measurement set of the tunnel cross-section; The manifold evolution calculation module is used to construct triangular meshes with manifold topology and virtual shrinking potential energy fields, and perform dynamic evolution to obtain denoised reconstructed meshes; The deformation analysis module is used to calculate the strain energy density, generate a deformation thermogram, and determine the deformation region. The feedback control module is used to generate early warning commands based on the deformation judgment results and adjust the scanning parameters of the data acquisition module.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the tunnel cross-section deformation monitoring method according to any one of claims 1 to 7.

Images