A Drilling Positioning Method and System Based on Spatial Scanning and Surrounding Rock Characteristic Survey

Abstract

This invention discloses a drilling positioning method and system based on spatial scanning and surrounding rock characteristic surveying, belonging to the field of digital twin technology. The method includes: performing multimodal 3D scanning of the rock wall; fusing and registering point cloud data to obtain a 3D model; dividing the 3D model into voxels; inputting real-time acquired geological signals into the 3D model; extracting features and calculating mechanical parameters to obtain a 3D digital twin model; constructing a comprehensive objective function to optimize the drilling scheme; using a multi-agent algorithm to update the 3D digital twin model; and iteratively optimizing to obtain the final drilling scheme. The method described in this invention effectively guides the differentiated drilling density and sequence in different regions based on the 3D digital twin model, improving the safety, precision, and resource utilization efficiency of drilling operations.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, specifically to a drilling positioning method and system based on spatial scanning and surrounding rock characteristic surveying. Background Technology

[0002] Digital twin technology provides precise real-time data support for geological surveying and drilling positioning by synchronizing physical entities and virtual models in real time. Point cloud data based on 3D scanning technology is increasingly being applied to surrounding rock surveying. Through data fusion and precise registration, highly realistic 3D digital models are constructed, and combined with digital twin technology, real-time updates and dynamic monitoring are achieved.

[0003] Most existing digital twin rock mass models lack sufficient methods for calculating mechanical parameters and assessing stability, failing to reflect the dynamic changes of the surrounding rock in real time during construction. This results in an inability to fully consider the dynamic changes and local characteristics of the surrounding rock when optimizing drilling schemes. Moreover, for highly complex scenarios such as large-section shield tunneling, existing methods suffer from disconnects in the processes of data acquisition and parameter processing, physical property parameter inversion, drilling spatial zoning, and drilling scheme optimization in digital twin models, limiting the level of intelligence in simulation. Consequently, it becomes difficult to balance the safety, cost, and operational efficiency of the drilling process. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is that existing drilling positioning technologies based on spatial scanning and surrounding rock characteristic surveys generally suffer from problems such as insufficient utilization of spatial information, insufficient intelligence of parameter zoning, and limited optimization constraint dimensions.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a drilling positioning method based on spatial scanning and surrounding rock characteristic survey, comprising: performing multimodal three-dimensional scanning of the rock wall, fusing and registering point cloud data to obtain a three-dimensional model;

[0007] The three-dimensional model is divided into voxels, and the real-time acquired geological signals are input into the three-dimensional model to extract features and calculate mechanical parameters, thus obtaining a three-dimensional digital twin model.

[0008] Calculate voxel physical property parameters, classify voxels, evaluate voxel stability using stability scores, and divide different drilling areas according to voxel category and stability.

[0009] A comprehensive objective function is constructed to optimize the drilling scheme. A multi-agent algorithm is used to update the 3D digital twin model, and the final drilling scheme is obtained through iterative optimization.

[0010] As a preferred embodiment of the drilling positioning method based on spatial scanning and surrounding rock characteristic surveying described in this invention, the multimodal three-dimensional scanning includes: using a LiDAR sensor to scan the rock wall along the rear edge of the drilling rig to generate a point cloud set; a structured light projection module to simultaneously supplement the point cloud set of the unscanned shadow area; and merging them into a global point cloud set; while using ground-penetrating radar (GPR) to record reflected geological signals.

[0011] Based on the fused global point cloud set and the initial attitude estimation provided by the inertial measurement unit (IMU), a rigid body transformation is performed on the newly acquired point cloud set to complete coarse registration; then, the ICP iterative nearest neighbor algorithm is used to find the nearest neighbor corresponding point of each point in the new point cloud set in the global point cloud, and the rigid body transformation matrix is ​​iteratively updated to perform fine registration.

[0012] The registered point cloud set and the shadow region point cloud are merged into the historical point cloud to obtain a temporary fusion set; the merged point cloud set is voxelized and resampled according to a fixed voxel size, retaining the point closest to the center in each voxel, and generating an updated global point cloud as a 3D model.

[0013] The tunnel boring machine advances a fixed distance along a predetermined trajectory, then enters the next scanning position, triggering a multimodal data acquisition and registration process to update the 3D model.

[0014] As a preferred embodiment of the drilling positioning method based on spatial scanning and surrounding rock characteristic surveying described in this invention, the three-dimensional digital twin model includes: inputting the real-time geological signal data acquisition results of GPR into the three-dimensional model; for each voxel in the three-dimensional model, collecting the amplitude and arrival time information of all intersecting rays, and calculating the four signal characteristics of the voxel: average amplitude, amplitude variance, arrival time skewness, and spectral centroid.

[0015] Four signal features are input into a multilayer perceptron regression model pre-trained on historical data to obtain the dielectric constant estimate for each voxel. Using the linear and power mapping coefficients calibrated by geological engineering, the dielectric constant estimate is successively converted into uniaxial compressive strength and elastic modulus. The uniaxial compressive strength and elastic modulus results of the voxels are smoothed based on neighborhood averaging to eliminate isolated outliers and output a three-dimensional digital twin model.

[0016] As a preferred embodiment of the drilling positioning method based on spatial scanning and surrounding rock characteristic surveying described in this invention, the voxel classification includes: calculating the local curvature of the surface geometry points of each voxel in the three-dimensional digital twin model; simultaneously mapping the uniaxial compressive strength and elastic modulus from the three-dimensional digital twin model; the three indices of uniaxial compressive strength, elastic modulus, and local curvature correspond to three dimensions and are normalized according to a pre-set value range; each voxel corresponds to a three-dimensional feature vector;

[0017] Construct a similarity network to obtain an adjacency matrix, where nodes represent voxels and edge weights represent the similarity between two voxels in the feature space, resulting in a similarity graph. Based on the adjacency matrix of the similarity graph, construct a degree matrix. The degree value of each node in the degree matrix is ​​the sum of its similarity to all other nodes. Combine the degree matrix and the adjacency matrix to generate a Laplacian matrix.

[0018] Calculate the difference between the feature vectors corresponding to any two voxels, convert the distance into similarity, and use the k-nearest neighbor method to filter edges; for each voxel, only keep the similarity between the k closest voxels as edges, and set the other edges to zero;

[0019] For the Laplace matrix, before calculation The eigenvectors corresponding to the smallest eigenvalues, each eigenvector representing a scalar set with a length equal to the total number of voxels; take the first... Individual elements are spliced ​​together to form voxels in spectral space. 3D embedded coordinates;

[0020] get After embedding coordinates, the embedding vector corresponding to each voxel is normalized, retaining only the orientation information; after spectral embedding... In the 3D space, each voxel is treated as a point, and K initial cluster centers are randomly selected; the embedding vector of each voxel is assigned to the cluster corresponding to the nearest center; based on the assignment result, the center position of each cluster is updated to the average value of all embedding vectors of the cluster; the assignment and update steps are repeated until the center positions converge; the voxels are divided into K classes.

[0021] As a preferred embodiment of the drilling positioning method based on spatial scanning and surrounding rock characteristic surveying described in this invention, the step of dividing different drilling areas according to voxel category and stability includes: performing connectivity retrieval between voxels according to the improved orientation rules of the three-dimensional mesh in the three-dimensional digital twin model to determine whether each voxel is connected to any adjacent voxel; calculating a stability score for each voxel using the finite element analysis method and setting a local stability threshold; if two voxels belong to the same cluster, the difference in stability between the two voxels is less than the local stability threshold, and they are neighbors on the mesh, then they are classified as the same connected component.

[0022] Traverse all voxels and extract interconnected sub-regions from the same cluster; voxels within each sub-region are semantically and spatially consistent.

[0023] For each connected sub-region, the sum of the stability scores of all voxels within the sub-region is calculated to obtain the region stability score of the sub-region.

[0024] Based on the comparison results of the regional stability score and two step thresholds, all voxels are divided into high-stability, medium-stability, and low-stability regions, and the partition representation of the 3D digital twin model is output. Each region has a regional stability score, a list of candidate holes, regional center coordinates, and spatial topological relationships with adjacent regions. Among them, the high-stability region is equipped with low-density and conventional hole placement, the medium-stability region is equipped with conventional density hole placement, and the low-stability region needs to be reinforced first.

[0025] As a preferred embodiment of the drilling positioning method based on spatial scanning and surrounding rock characteristic surveying described in this invention, the construction of the comprehensive objective function includes: the comprehensive objective value of the region is weighted and synthesized by stability penalty, drilling movement cost and cross-regional constraint penalty, and the optimization objective is to minimize the comprehensive objective value; the cross-regional constraint penalty is calculated based on the degree of violation of cross-regional constraints by the regional drilling scheme;

[0026] The stability penalty consists of the overall stability of the region and the micro-stability of the location of the hole; the cross-region constraints include the minimum distance constraint between holes, the upper limit constraint of the total number of holes, and the region priority constraint.

[0027] As a preferred embodiment of the drilling location method based on spatial scanning and surrounding rock characteristic survey described in this invention, the method of using a multi-agent algorithm to update the three-dimensional digital twin model includes treating each connected region obtained by spectral clustering as an agent and independently maintaining a list of candidate borehole locations and a drilling location scheme within the region.

[0028] Each agent selects several candidate perforation schemes within the region and checks whether the candidate perforation schemes in this region meet the stability penalty requirements. If there is a conflict with the already determined perforations in neighboring regions, cross-regional constraints and key region priority constraints are added for recalculation to obtain local candidate schemes. The comprehensive objective function of this region is constructed to obtain the comprehensive objective value of the region.

[0029] The solution with the lowest comprehensive objective score is selected from the local candidate solutions as the current optimal solution, and the current optimal solution is sent to the neighboring Agents. Each Agent continuously receives the current optimal solution from its neighbors. When it finds that there is a conflict between the current optimal solution of the neighbor and the local current optimal solution in terms of cross-regional hole distance or cross-regional priority order, it automatically adjusts the local candidate solution according to the cross-regional constraints.

[0030] Every few rounds, the coordinator feeds back the current optimal solution to the 3D digital twin model, updates the digital twin model, and collects the updated current optimal solution information of all agents. It checks whether the solution complies with cross-regional constraints and issues local candidate solution adjustment suggestions to agents that violate cross-regional constraints.

[0031] When the preset number of iterations is reached, the iteration stops; each Agent submits the final optimal drilling plan; after the coordinator checks the plans output by all regions, it ensures that the cross-regional hole positions meet the cross-regional constraints, the comprehensive target value of the entire region is minimized, and the overall hole layout plan is generated.

[0032] As a preferred embodiment of the drilling positioning system based on spatial scanning and surrounding rock characteristic surveying described in this invention, it includes a multimodal data acquisition module, a digital twin module, a stability assessment module, and a drilling scheme generation module.

[0033] The multimodal data acquisition module is used for the automatic triggering, acquisition, registration, and fusion of multi-sensor data during the tunnel boring machine's advancement, and outputs a global point cloud.

[0034] The digital twin module is used to map the fused point cloud and GPR signal to voxels, and through mapping, generate a three-dimensional digital twin model with mechanical parameter annotations.

[0035] The stability evaluation module is used to calculate voxel-level stability features on a three-dimensional digital twin model and obtain connected punched partitions through spectral clustering.

[0036] The aperture layout generation module is used to construct a comprehensive objective function based on the stability and candidate aperture positions of each region, and output the optimal aperture layout scheme for each region through multi-agent collaborative search.

[0037] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of a drilling positioning method based on spatial scanning and surrounding rock characteristic survey.

[0038] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a drilling positioning method based on spatial scanning and surrounding rock characteristic surveying.

[0039] The beneficial effects of this invention are as follows: By combining the fusion and registration of point cloud data, a high-precision three-dimensional digital twin model of the surrounding rock is constructed. Utilizing digital twin technology, real-time acquired geological signals are input into the three-dimensional model, continuously updating and dynamically adjusting the mechanical properties of the surrounding rock. This allows the surrounding rock to reflect its true state at any time during drilling, enabling accurate assessment of its local stability. It effectively identifies potential risk areas within the surrounding rock, avoiding the inaccuracies in rock characteristic estimations found in traditional methods, thus improving safety. By comprehensively considering multiple factors such as surrounding rock stability, construction efficiency, and drilling safety, and dynamically interacting with the digital twin model, the optimal drilling scheme is obtained under multiple constraints, maximizing construction efficiency and minimizing the impact on the surrounding environment. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 The first embodiment of the present invention provides an overall flowchart of a drilling positioning method based on spatial scanning and surrounding rock characteristic survey. Detailed Implementation

[0042] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0043] Example 1, referring to Figure 1 As an embodiment of the present invention, a drilling positioning method based on spatial scanning and surrounding rock characteristic survey is provided, comprising:

[0044] S1: Perform multimodal 3D scanning on the rock wall, fuse and register the point cloud data to obtain a 3D model.

[0045] Furthermore, the multimodal 3D scanning includes generating a point cloud set by scanning the rock wall along the trailing edge of the drilling rig using a LiDAR sensor, and simultaneously supplementing the point cloud set of the unscanned shadow area using a structured light projection module, merging them into a global point cloud set; at the same time, the reflected geological signals are recorded using ground-penetrating radar (GPR).

[0046] Based on the fused global point cloud set and the initial attitude estimation provided by the inertial measurement unit (IMU), a rigid body transformation is performed on the newly acquired point cloud set to complete coarse registration. The ICP iterative nearest neighbor algorithm is used to find the nearest neighbor corresponding point of each point in the new point cloud set in the global point cloud, and the rigid body transformation matrix is ​​iteratively updated to perform fine registration.

[0047] The registered point cloud set and the point cloud in the shadow area are merged into the historical point cloud to obtain a temporary fusion set. The merged point cloud set is then voxelized and resampled according to a fixed voxel size, retaining the point closest to the center in each voxel, and generating an updated global point cloud as a 3D model.

[0048] The tunnel boring machine advances a fixed distance along a predetermined trajectory, then enters the next scanning position, triggering a multimodal data acquisition and registration process to update the 3D model.

[0049] A tunnel boring machine (TBM) consists of a cutterhead and a body. The cutterhead, located at the very front of the tunnel face, is used to break and cut the rock and soil to form the tunnel cross-section. The rear edge of the cutterhead is essentially the first line of defense when the TBM contacts the surrounding rock; various sensors installed at this location can be the first to capture information about the rock face. The body is a closed cylinder, housing the propulsion jacks, motors, cables, and control cabinets. During the TBM's advancement, the cutterhead continuously rotates, breaking the rock, and the cuttings are transported to the rear of the body and then carried away by the muck removal system. Simultaneously, the body itself moves forward to maintain mechanical balance with the excavated section.

[0050] Furthermore, high-precision LiDAR scanning units are compactly arranged on one side of the rear edge of the tunnel boring machine's cutterhead. The LiDAR maintains a fixed distance between the excavation face and the surrounding rock surface, performing measurements in a 360° or semi-circular manner, continuously emitting lasers and receiving echoes to generate a real-time set of three-dimensional point clouds on the surrounding rock surface. Simultaneously, a structured light projection module projects predefined gratings or spots onto the rock wall, and a fixed-view camera located inside the tunnel boring machine captures deformed light patterns, forming a supplementary three-dimensional point cloud. This effectively fills in shadowed areas or obscured corners that the LiDAR cannot reach, resulting in a more complete geometry of the surrounding rock surface.

[0051] The ground-penetrating radar transmitting and receiving array installed at the tunnel boring machine face continuously emits high-frequency electromagnetic waves and records the reflected waveforms from various layers inside the rock mass during brief moments when the radar antenna is relatively stationary relative to the surrounding rock surface. The radar system stores the echo signals in pipeline or sector scanning mode with timestamps, including amplitude information and travel time for each line, to prepare raw data for subsequent physical property inversion calculations of voxel mechanical parameters.

[0052] The real-time geological signal data acquisition results of GPR are input into the three-dimensional model; for each voxel in the three-dimensional model, the amplitude and arrival time information of all intersecting rays are collected, and the four signal characteristics of the voxel, namely the average amplitude, amplitude variance, arrival time skewness and spectral centroid, are calculated respectively.

[0053] Four signal features are input into a multilayer perceptron regression model pre-trained on historical data to obtain the dielectric constant estimate for each voxel. Using the linear and power mapping coefficients calibrated by geological engineering, the dielectric constant estimate is successively converted into uniaxial compressive strength and elastic modulus. The uniaxial compressive strength and elastic modulus results of the voxels are smoothed based on neighborhood averaging to eliminate isolated outliers and output a three-dimensional digital twin model.

[0054] Inside the tunnel boring machine (TBM) control cabinet, the inertial measurement unit (IMU) outputs the TBM's three-dimensional attitude information (including pitch, roll, and yaw angles) in real time. This attitude data is used to initially stitch together the point cloud acquired by LiDAR and structured light in the TBM's own coordinate system. By reading the most recent IMU data, the newly acquired point cloud set can be rigidly transformed according to attitude transformation rules, undergoing approximate rotation and translation to keep it approximately aligned with the previous geometric position.

[0055] Based on the coarse registration results, the Iterative Nearest Point (ICP) algorithm is used to perform fine registration between the newly acquired point cloud set and the currently accumulated global point cloud. Specifically, each point in the newly acquired point cloud is matched with its nearest neighbor in the global point cloud, and rigid body transformations (rotation matrix and translation vector) are calculated to minimize the Euclidean distance error between them. The transformation matrix is ​​iteratively updated repeatedly until the error converges or the preset number of iterations is reached. This eliminates the bias caused by the initial pose estimation and accurately aligns the old and new point clouds.

[0056] The transformed point cloud obtained by the ICP algorithm is merged with the point cloud of the shadow area supplemented by structured light, and then added to the historical accumulated point cloud of the tunnel boring machine to form a temporary fusion result. This temporary fusion result includes both high-precision LiDAR measurement points and structured light supplementation points, representing the complete three-dimensional geometry of the current tunnel face and its surroundings.

[0057] It should be noted that, in order to limit the size of the point cloud data and provide standardized spatial units for subsequent calculations, the fused point cloud needs to be processed into a voxel mesh. A fixed voxel side length is selected, and the space of the 3D model is divided into a series of uniform cubic voxels. For each voxel, only the point closest to the voxel center is retained, and all other points are discarded, thus ensuring that each voxel has only one representative point.

[0058] S2: Divide the 3D model into voxels, input the real-time acquired geological signals into the 3D model, and obtain a 3D digital twin model by feature extraction and mechanical parameter calculation.

[0059] Furthermore, the three-dimensional digital twin model includes applying the real-time reflection data acquisition results of GPR to the three-dimensional model; for each voxel in the three-dimensional model, collecting the amplitude and arrival time information of all intersecting rays, and calculating the four signal characteristics of the voxel: average amplitude, amplitude variance, arrival time skewness, and spectral centroid.

[0060] Four signal features are input into a multilayer perceptron regression model pre-trained on historical data to obtain the dielectric constant estimate for each voxel. Using the linear and power mapping coefficients calibrated by geological engineering, the dielectric constant estimate is successively converted into uniaxial compressive strength and elastic modulus. The uniaxial compressive strength and elastic modulus results of the voxels are smoothed based on neighborhood averaging to eliminate isolated outliers and output a three-dimensional digital twin model.

[0061] Furthermore, the voxelized global point cloud is correlated one-to-one with the ground-penetrating radar (GPR) data. Specifically, a 3D voxel mesh with the same side length as the point cloud is constructed in the same coordinate system of the 3D model, and a set of radar rays, also known as ray paths, is determined for each voxel. By projecting the GPR echo data into the spacetime, the time depth (arrival time) and amplitude information of all reflected signals belonging to that voxel can be obtained.

[0062] For all intersecting GPR rays within each voxel, the mean and variance of their echo amplitudes are statistically analyzed to characterize the concentration and dispersion of reflection intensity. Simultaneously, the skewness of these rays' arrival time information is calculated to reflect the asymmetric distribution within the medium. Finally, spectral analysis is used to extract the signal's spectral centroid, reflecting the medium's attenuation characteristics for signals of different frequencies. These four statistical features form a unified signal feature matrix across all voxels.

[0063] The four-dimensional signal features obtained from the above statistics are used as input to a multilayer perceptron (MLP) regression model pre-trained based on large-scale historical geological profile data. This neural network model has undergone supervised training using geological borehole samples and measured radar data, and can output an estimated dielectric constant value for each voxel. This dielectric constant represents the electrical properties of the rock or soil, providing a basis for subsequent calculations of rock mass mechanical parameters.

[0064] Based on empirical mapping rules commonly used in geological engineering, the dielectric constant output by the MLP is successively converted into uniaxial compressive strength and elastic modulus. Specifically, a set of linear mapping coefficients is first used to convert the dielectric constant into an approximate value of the compressive strength, and then mapping coefficients based on a power function (power curve) are used to map the compressive strength to the elastic modulus. This allows us to obtain the set of mechanical parameters for each voxel at the global voxel scale.

[0065] The four signal features are input into a multilayer perceptron regression model pre-trained on historical data to obtain the dielectric constant estimate of each voxel; using the linear and power mapping coefficients calibrated by geological engineering, the estimated dielectric constant is successively converted into uniaxial compressive strength and elastic modulus.

[0066] Linear and power-law mapping coefficients, geological engineering calibration studies typically indicate that the dielectric constant... Uniaxial compressive strength of rock mass There is a linear relationship, and the formula is in the form of:

[0067]

[0068] Where a and b are constants obtained through regression of a large number of measured samples.

[0069] Similarly, the dielectric constant and the elastic modulus E can be approximated by a power-law mapping:

[0070]

[0071] Where c and This is an empirical coefficient, obtained through indoor experiments or backtesting of core samples from the field.

[0072] The calculation of voxel mechanical parameters involves first using the estimated dielectric constant of each voxel in the mesh. Calculate the estimated value of the corresponding uniaxial compressive strength. :

[0073]

[0074] And estimate the elastic modulus based on the same dielectric constant. :

[0075]

[0076] The compressive strength and elastic modulus results for all elements are smoothed using a neighborhood averaging method to eliminate isolated outliers, outputting a three-dimensional digital twin model of the physical property parameters. The propagation speed of electromagnetic waves emitted by ground-penetrating radar in the underground medium is affected by the dielectric constant. The propagation speed is estimated by measuring the propagation time and a known distance. The estimated dielectric constant is calculated based on the propagation speed and the inherent properties of the electromagnetic wave. It should be noted that the dielectric constant can also be estimated through the physical properties of soil and rock materials.

[0077] It should be noted that each voxel already contains smoothed compressive strength and elastic modulus. Organizing these parameters together with the spatial coordinates of the voxels forms a complete 3D digital twin model with property annotations. This model not only displays the geometric information of the surrounding rock surface but also integrates mechanical properties such as underground dielectric constant, compressive strength, and elastic modulus, providing the necessary numerical basis for subsequent generation of stability scoring functions, zoning, and optimization of drilling schemes.

[0078] S3: Calculate voxel physical property parameters, classify voxels, evaluate voxel stability using stability scores, and divide different drilling areas according to voxel category and stability.

[0079] Furthermore, the voxel classification includes calculating the local curvature of the surface geometry points of each voxel in the three-dimensional digital twin model; simultaneously mapping the uniaxial compressive strength and elastic modulus from the three-dimensional digital twin model; the three indices of uniaxial compressive strength, elastic modulus, and local curvature correspond to three dimensions and are normalized according to a pre-set value range; each voxel corresponds to a three-dimensional feature vector.

[0080] Construct a similarity network to obtain an adjacency matrix, where nodes represent voxels and edge weights represent the similarity between two voxels in the feature space, resulting in a similarity graph. Based on the adjacency matrix of the similarity graph, construct a degree matrix. The degree value of each node in the degree matrix is ​​the sum of its similarity to all other nodes. Combine the degree matrix and the adjacency matrix to generate a Laplacian matrix.

[0081] Calculate the difference between the feature vectors of any two voxels, convert the distance into similarity, and use the k-nearest neighbor method to filter edges; for each voxel, only keep the similarity between the k closest voxels as edges, and set the other edges to zero.

[0082] For the Laplace matrix, before calculation The eigenvectors corresponding to the smallest eigenvalues, each eigenvector representing a scalar set with a length equal to the total number of voxels; take the first... Individual elements are spliced ​​together to form voxels in spectral space. Dimensional embedded coordinates.

[0083] get After embedding coordinates, the embedding vector corresponding to each voxel is normalized, retaining only the orientation information; after spectral embedding... In the 3D space, each voxel is treated as a point, and K initial cluster centers are randomly selected; the embedding vector of each voxel is assigned to the cluster corresponding to the nearest center; based on the assignment result, the center position of each cluster is updated to the average value of all embedding vectors of the cluster; the assignment and update steps are repeated until the center positions converge; the voxels are divided into K classes.

[0084] Furthermore, the step of dividing different punched regions according to voxel type and stability includes: performing connectivity retrieval between voxels according to the improved orientation rules of the three-dimensional mesh in the three-dimensional digital twin model to determine whether each voxel is connected to any adjacent voxel; calculating a stability score for each voxel using the finite element analysis method and setting a local stability threshold; if two voxels belong to the same cluster, the difference in stability between the two voxels is less than the local stability threshold, and they are neighbors on the mesh, then they are classified as the same connected component.

[0085] The finite element analysis method includes applying external loads based on the constructed three-dimensional digital twin model, such as the mechanical effects during tunnel boring machine (TBM) excavation (e.g., forward pressure, TBM cutterhead friction), the effect of the surrounding rock's self-weight on each voxel, and the effects of other possible external pressures or forces during construction. Appropriate boundary conditions are applied according to the actual situation. The equilibrium equations for the rock wall region are established using the finite element method. The displacement, stress, and force fields of the system are linked using the principle of virtual work or the principle of minimum potential energy. Stress and strain are calculated based on the obtained displacements. The stress of each voxel is checked, especially in critical areas (e.g., fractures, weak zones). The maximum principal stress and maximum shear stress of the rock mass are calculated.

[0086] By analyzing stress and strain, and combining the material's compressive strength, shear strength, and yield strength parameters, the stability score of each voxel is calculated. The stability score is obtained by dividing the maximum principal stress by the compressive strength and summing it with the maximum shear stress divided by the shear strength.

[0087] Traverse all voxels and extract interconnected sub-regions from the same cluster; voxels within each sub-region are semantically and spatially consistent.

[0088] For each connected sub-region, the sum of the stability scores of all voxels within the sub-region is calculated to obtain the regional stability score of the sub-region.

[0089] Based on the comparison results of the regional stability score and two step thresholds, all voxels are divided into high-stability, medium-stability, and low-stability regions, and the partition representation of the 3D digital twin model is output. Each region has a regional stability score, a list of candidate holes, regional center coordinates, and spatial topological relationships with adjacent regions. Among them, the high-stability region is equipped with low-density and conventional hole placement, the medium-stability region is equipped with conventional density hole placement, and the low-stability region needs to be reinforced first.

[0090] First, for each voxel Calculate stability score Set a local stability threshold. Only when adjacent voxels Satisfies: Both belong to the same spectral cluster; voxels and The difference in stability Spatially, two objects are considered connected if their coordinates differ by 1 on only one axis or if they are diagonally oriented (differing by 1 on each of the two axes). Thus, two objects belonging to the same cluster index, with similar stability, and a score difference not exceeding [a certain threshold] are considered connected. Even if they are adjacent along the diagonal direction of the volume, they can be included in the same connected component; conversely, if the stability difference is too large, even if the meshes are adjacent, they will not be forcibly connected, thus avoiding forcibly merging voxels with large stability differences into one region.

[0091] Under the same cluster label, for voxels Search space range All voxel candidates within .

[0092] For each candidate voxel pair If their center points are Euclidean distance And poor stability If they are connected, then they are determined to be connected.

[0093] The voxel labels obtained from spectral clustering are combined with a 3D mesh space. An improved connectivity method is used to treat voxels with the same label and spatially adjacent voxels as the same connected region. This yields voxels belonging to the label... Several connected subregions.

[0094] Each connected component These are all spatially connected blocks within the same spectral cluster. For each connected block... If its voxel count If the number is less than the minimum threshold, the component is merged with the nearest connected component with the same label and the closest stability, until the minimum size is met; regional stability statistics are performed on each connected region. By weighted average voxel stability score To define The overall stability of the region can be expressed by the formula:

[0095]

[0096] in, This represents the number of voxels contained in the region. For the region Individual stability score. High / medium / low stability regions are defined using pre-set thresholds. and (like ), dividing each connected region into:

[0097] High stability region: if Stable region: if Low stability region: if .

[0098] By traversing all voxels, several interconnected sub-regions can be extracted from the same cluster label; voxels within each sub-region maintain semantic and spatial consistency.

[0099] For each connected sub-region, the stability scores of all voxels within the sub-region are first calculated to obtain the overall stability index of the sub-region.

[0100] Based on the comparison results of the overall stability score and the high and low thresholds, all voxels are divided into high stability, medium stability and low stability regions, and a three-dimensional partition model is output. The high stability region is equipped with low density and conventional porosity, the medium stability region is equipped with conventional density porosity, and the low stability region needs to be reinforced first.

[0101] It should be noted that, based on grid connectivity, real spatially connected sub-regions under the same labels are extracted; the overall stability of each sub-region is calculated and divided into high, medium, and low categories, facilitating subsequent optimization of borehole layout strategies by region. This avoids isolated processing at the individual voxel level, instead first analyzing in the feature space and then combining spatial connectivity to extract regions. This ensures that the final partitioning reflects both the similarity of rock mass physical characteristics and the spatial connectivity structure, laying a more solid and reasonable foundation for borehole layout optimization.

[0102] S4: Construct a comprehensive objective function to optimize the punching scheme, use a multi-agent algorithm to update the 3D digital twin model, and iteratively optimize to obtain the final punching scheme.

[0103] Furthermore, the construction of the comprehensive objective function includes the comprehensive objective value of the region being a weighted synthesis of stability penalty, drilling movement cost, and cross-regional constraint penalty. The lower the value, the better the scheme. The cross-regional constraint penalty is calculated based on the degree to which the regional drilling scheme violates the cross-regional constraints.

[0104] The stability penalty consists of the overall stability of the region and the micro-stability of the location of the hole; the cross-region constraints include the minimum distance constraint between holes, the upper limit constraint of the total number of holes, and the region priority constraint.

[0105] Furthermore, the method of using a multi-agent algorithm to update the three-dimensional digital twin model includes treating each connected region obtained by spectral clustering as an agent, and independently maintaining a list of candidate holes and a hole-drilling location scheme within the region.

[0106] Each agent selects several candidate perforation schemes within the region and checks whether the candidate perforation schemes in this region meet the stability penalty requirements. If there is a conflict with the fixed perforation in the neighboring region, cross-regional constraint and key region priority constraint are added for recalculation to obtain local candidate schemes. The comprehensive objective function of this region is constructed to obtain the comprehensive objective value of the region.

[0107] The solution with the lowest comprehensive objective score is selected from the local candidate solutions as the current optimal solution, and the current optimal solution is sent to the neighboring Agents. Each Agent continuously receives the current optimal solution from its neighbors. When it finds that there is a conflict between the current optimal solution of the neighbor and the local current optimal solution in terms of cross-regional hole distance or cross-regional priority order, it automatically adjusts the local candidate solution according to the cross-regional constraints.

[0108] Every few rounds, the coordinator feeds back the current optimal solution to the 3D digital twin model, updates the digital twin model, and collects the updated current optimal solution information from all agents. It then checks whether the solution complies with cross-regional constraints and issues local candidate solution adjustment suggestions to agents that violate cross-regional constraints.

[0109] When the preset number of iterations is reached, the iteration stops; each Agent submits the final optimal drilling plan; after the coordinator checks the plans output by all regions, it ensures that the cross-regional hole positions meet the cross-regional constraints, the comprehensive target value of the entire region is minimized, and the overall hole layout plan is generated.

[0110] Based on what has been obtained Connected subregions Each sub-region Possesses: Regional voxel average stability index (A larger value indicates greater stability). Region centroid coordinates This is used to associate with the drilling rig's movement path. It represents the set of selectable candidate drilling points within the region. These points were selected from the previous gridded candidate set based on the connected regions.

[0111] Each region The decision variables for punching are represented by a vector. It means that each of them Indicates whether it is at the candidate point Punch holes and sort them (add candidate positions to the current position set). .

[0112] First, determine the number of holes needed for each area. , and then from Select specific Points. If the regional stability is higher, it allows... The smaller the value; the lower the stability, the more it requires. Increase the size and reinforce it beforehand.

[0113] The overall constraints include: minimum distance between holes; regardless of whether the holes are adjacent or in the same area, any two drilling points must maintain a minimum distance. Total number of holes: The total number of holes in all areas, based on the project budget or timeframe. Cannot exceed .in This indicates the area number. Drilling rig movement restrictions: the tunnel boring machine (TBM) platform can move laterally / longitudinally within a certain range, but the cost of moving from the current position to the next hole location before each drilling operation must be considered and reflected in the objective function. Area priority constraints: if certain areas are defined as high-risk / low-stability, drilling schemes on these areas need to be prioritized or combined with reinforcement work, affecting the stability penalty of the overall objective value.

[0114] Taking into account safety, economy, and efficiency, the comprehensive objective function should include at least the following types of indicators:

[0115] For regions with high stability This allows for a reduction in drilling density or a decrease in risk penalty in that area. In other words, when drilling in the same location, if... A larger value means lower risk and a smaller penalty to the stability of the objective function.

[0116] If the stability is low, the stability penalty needs to be increased to avoid disorderly large-scale drilling.

[0117] Drilling costs depend on the hole depth and rock parameters. Generally, the deeper the hole and the harder the rock, the greater the drilling costs. This can be measured by multiplying the cost per unit depth by the depth; alternatively, it can be simply tiered according to the average rock strength.

[0118] Equipment relocation cost: After each drilling operation, the drill frame / boom needs to be moved from the current hole location to the next. The greater the relocation distance and the more tortuous the path, the higher the cost. Therefore, it is necessary to perform weighted calculations on the Euclidean distance in the 3D digital twin model between two adjacent holes or the mileage along the tunnel boring machine's movement path.

[0119] For low stability regions Drilling at unsuitable locations could cause rockfalls or displacements, necessitating a heavier stability penalty. A common practice is to increase the regional stability penalty. Combined with pore stability or voxel local stability:

[0120] If the single-point stability of the voxel corresponding to the hole is low, a penalty coefficient is applied to the hole; or the overall stability of the region is directly affected. Set a threshold; for each hole drilled in an area below the threshold, add a fixed penalty coefficient.

[0121] For regional priority constraints, consider priority and time windows. Some regions may require pre-placement of holes or be prioritized due to safety or structural requirements. Introduce a priority reward, i.e., if a low-stability region... Completing drilling in the first half of the phase earns an efficiency bonus. A time window is planned, specifying that certain areas must be completed before the tunnel boring machine reaches a certain milestone.

[0122] For each region First, calculate the regional stability penalty for the perforation set in the region: this depends on the regional average stability. If stability The higher the value, the lower the penalty; conversely, the lower the value, the higher the penalty. Simultaneously, the local stability of specific aperture locations within the region is further weighted: if the stability of a candidate point's corresponding voxel is also low, an additional penalty is applied. The combined results yield a regional risk score. Calculate the drilling and relocation costs for this area. The drilling cost can be mapped to a scalar based on the expected depth of each hole and the rock hardness. Then, the drilling costs of all holes in the area are summed to obtain the final cost. The cost of moving the equipment is estimated based on the drilling order: if the drilling order is... Then, the three-dimensional distance (or the actual drivable path length of the tunnel boring machine platform) between each pair of adjacent holes is calculated, weighted, and summed to obtain the movement cost. .

[0123] Furthermore, each connected region obtained from spectral clustering is regarded as an agent, which independently maintains the candidate hole positions and drilling schemes for its region.

[0124] Each agent selects several drilling schemes within the region; checks whether the minimum distance between holes in the region is met; calculates local risk penalties, drilling costs, and movement costs; if there is a conflict with existing holes in neighboring regions, cross-regional penalties are added.

[0125] The agent performs a neighborhood search based on its current optimal solution:

[0126] Based on the existing optimal solution, try replacing a hole point in the current solution with an unused candidate point within the region; fine-tune the order by locally swapping two adjacent positions in the drilling sequence; adjust the quantity by vertically adjusting the total number of holes. For example, try adding or removing a point within a certain range. Generate several (e.g., n candidate) new solutions, and also perform feasibility corrections to ensure that the minimum distance between holes meets the constraints.

[0127] The new and old optimal solutions are submitted together to the local target evaluator: the local risk score, drilling cost, and movement cost are recalculated; cross-regional penalties are introduced with neighbor information, increasing the penalty if a point selected in the new solution does not maintain the minimum distance from the neighbor's optimal hole position. The best local solution for this round is selected from these solutions, and the updated summary information, including hole position, order, and score, is shared with the neighbors.

[0128] The lowest-scoring candidate solution is selected as the current optimal solution, and a summary of the solution is sent to neighboring agents. Each agent continuously receives the latest optimal solution from its neighbors, and automatically adjusts its local candidates when it finds a conflict with the local optimal solution in terms of hole distance, sequence order, etc.

[0129] Every few rounds, the coordinator collects the optimal solution, gathering the current best drilling plan from all agents. Every few rounds, the agents need to send feedback signals to the 3D digital twin model to update the current model state. This process ensures that each agent's optimal solution is consistent with the model's real-time data, and provides a data foundation for subsequent optimization adjustments. Each agent sends its current best drilling plan to the digital twin model, including the spatial coordinates of the hole positions, drilling sequence, stability score, and other information. The updated digital twin model state includes the stress state of the surrounding rock, deformation, and changes in adjacent areas.

[0130] The coordinator's role is to ensure that all agents' updated solutions satisfy global constraints and minimize the overall comprehensive objective value. It receives the optimal solutions updated by all agents, forming an updated global solution set. Cross-regional constraint checks are then performed.

[0131] The distance constraint between holes checks whether the hole positions between adjacent regions meet the preset distance constraints. If the hole positions in adjacent regions are too close, violating the distance constraint, adjustments are needed. The sequence order constraint checks whether there are conflicts in the drilling order of holes across regions. If the drilling order of adjacent regions is inconsistent, adjustments are needed. The coordinator checks the optimal solution for each pair of adjacent regions to determine whether the cross-region constraints are met and issues constraint adjustment suggestions to the relevant Agents.

[0132] Each Agent, during initialization, records its spatial neighbor list based on the partition topology obtained from spectral clustering. These neighboring regions refer to areas in the 3D mesh where at least one voxel face is adjacent to a voxel face in this region.

[0133] During the iteration process, the Agent will periodically or as needed send information such as the current candidate best local solution fragment, the best score, and the set of points that have been locked or are tentatively scheduled to be punched to its neighbors.

[0134] Similarly, the Agent will also receive similar information from its neighbors to adjust its own strategy and avoid conflicts between holes in different regions.

[0135] To take into account the overall goal, every few iterations, a coordinator can be introduced to collect the local optimal solutions and their scores from all agents, synthesize and broadcast the global optimal score, or a cross-regional conflict list.

[0136] After receiving information from the coordinator, each agent dynamically weights and merges its local and global objectives. For example, if a region experiences excessively high penalties due to conflicts, the coordinator may instruct that region to reduce the number or order of newly added holes, or to schedule them earlier; or to allow neighboring regions to make minor adjustments. Each agent can explore and update its local solutions according to its own pace and rhythm, communicating asynchronously with each other.

[0137] When the preset number of iterations is reached, the iteration stops; each agent submits the final optimal drilling plan; after the coordinator checks the plans output by all regions, it ensures that the cross-regional hole positions meet the safety distance and budget requirements, and generates the overall hole layout plan.

[0138] It should be noted that every few rounds, or when an Agent's score fails to improve consecutively, a logical coordinator collects the current optimal solutions from all Agents. It calculates whether there are irreparable conflicts between regions (e.g., two regions with severely overlapping hole positions and a distance less than a certain threshold). If a conflict is found, constraint adjustments are directly sent to the Agents involved in the conflict, either lowering their priority or increasing their local penalties. If the overall conflict is not severe, but the scores of certain regions are unsatisfactory, appropriate global guidance can be given to these regions, such as hinting that they should try increasing or decreasing the number of holes in the next iteration. The coordinator does not participate in decision-making; it only summarizes and provides hints based on horizontal information.

[0139] Example 2, an embodiment of the present invention, provides a drilling positioning system based on spatial scanning and surrounding rock characteristic survey, including a multimodal data acquisition module, a digital twin module, a stability assessment module, and a hole layout scheme generation module.

[0140] The multimodal data acquisition module is used for the automatic triggering, acquisition, registration, and fusion of multi-sensor data during the tunnel boring machine's advancement, and outputs a global point cloud.

[0141] The digital twin module is used to map the fused point cloud and GPR signal to voxels, and through mapping, generate a three-dimensional digital twin model with mechanical parameter annotations.

[0142] The stability assessment module is used to calculate voxel-level stability features on a three-dimensional digital twin model and obtain connected punched partitions through spectral clustering.

[0143] The aperture layout generation module is used to construct a comprehensive objective function based on the stability and candidate aperture positions of each region, and output the optimal aperture layout scheme for each region through multi-agent collaborative search.

[0144] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0145] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0146] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0147] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using a combination of any of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. 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, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A drilling positioning method based on spatial scanning and surrounding rock characteristic survey, characterized in that, include: Multimodal 3D scanning of the rock wall was performed, and the point cloud data was fused and registered to obtain a 3D model; The three-dimensional model is divided into voxels, and the real-time acquired geological signals are input into the three-dimensional model to extract features and calculate mechanical parameters, thus obtaining a three-dimensional digital twin model. Calculate voxel physical property parameters, classify voxels, evaluate voxel stability using stability scores, and divide different drilling areas according to voxel category and stability. A comprehensive objective function is constructed to optimize the drilling scheme. A multi-agent algorithm is used to update the 3D digital twin model, and the final drilling scheme is obtained through iterative optimization.

2. The drilling positioning method based on spatial scanning and surrounding rock characteristic survey as described in claim 1, characterized in that: The multimodal 3D scanning includes using a LiDAR sensor to scan the rock wall along the trailing edge of the drilling rig to generate a point cloud set, and a structured light projection module to simultaneously supplement the point cloud set of the shadow area that was not scanned, and then merging them into a global point cloud set; at the same time, ground-penetrating radar (GPR) is used to record the reflected geological signals. Based on the fused global point cloud set and the initial attitude estimation provided by the inertial measurement unit (IMU), a rigid body transformation is performed on the newly acquired point cloud set to complete coarse registration; then, the ICP iterative nearest neighbor algorithm is used to find the nearest neighbor corresponding point of each point in the new point cloud set in the global point cloud, and the rigid body transformation matrix is ​​iteratively updated to perform fine registration. The registered point cloud set and the shadow region point cloud are merged into the historical point cloud to obtain a temporary fusion set; the merged point cloud set is voxelized and resampled according to a fixed voxel size, retaining the point closest to the center in each voxel, and generating an updated global point cloud as a 3D model. The tunnel boring machine advances a fixed distance along a predetermined trajectory, then enters the next scanning position, triggering a multimodal data acquisition and registration process to update the 3D model.

3. The drilling positioning method based on spatial scanning and surrounding rock characteristic survey as described in claim 2, characterized in that: The three-dimensional digital twin model includes inputting the real-time geological signal data acquisition results of GPR into the three-dimensional model; for each voxel in the three-dimensional model, collecting the amplitude and arrival time information of all intersecting rays, and calculating the four signal characteristics of the voxel: average amplitude, amplitude variance, arrival time skewness, and spectral centroid. Four signal features are input into a multilayer perceptron regression model pre-trained on historical data to obtain the dielectric constant estimate for each voxel. Using the linear and power mapping coefficients calibrated by geological engineering, the dielectric constant estimate is successively converted into uniaxial compressive strength and elastic modulus. The uniaxial compressive strength and elastic modulus results of the voxels are smoothed based on neighborhood averaging to eliminate isolated outliers and output a three-dimensional digital twin model.

4. The drilling positioning method based on spatial scanning and surrounding rock characteristic survey as described in claim 3, characterized in that: The voxel classification process includes calculating the local curvature of each voxel's surface geometry point in the three-dimensional digital twin model; simultaneously mapping the uniaxial compressive strength and elastic modulus from the three-dimensional digital twin model; the three indices of uniaxial compressive strength, elastic modulus, and local curvature correspond to three dimensions and are normalized according to a pre-defined value range; each voxel corresponds to a three-dimensional feature vector; Construct a similarity network to obtain an adjacency matrix. Nodes represent voxels, and edge weights represent the similarity between two voxels in the feature space, resulting in a similarity graph. Based on the adjacency matrix of the similarity graph, construct a degree matrix. The degree value of each node in the degree matrix is ​​the sum of its similarity to all other nodes. Combine the degree matrix and the adjacency matrix to generate the Laplacian matrix; Calculate the difference between the feature vectors corresponding to any two voxels, convert the distance into similarity, and use the k-nearest neighbor method to filter edges; for each voxel, only keep the similarity between the k closest voxels as edges, and set the other edges to zero; For the Laplace matrix, before calculation The eigenvectors corresponding to the smallest eigenvalues, each eigenvector representing a scalar set with a length equal to the total number of voxels; take the first... Individual elements are spliced ​​together to form voxels in spectral space. 3D embedded coordinates; get After embedding coordinates, the embedding vector corresponding to each voxel is normalized, retaining only the orientation information; after spectral embedding... In the 3D space, each voxel is treated as a point, and K initial cluster centers are randomly selected; the embedding vector of each voxel is assigned to the cluster corresponding to the nearest center; based on the assignment result, the center position of each cluster is updated to the average value of all embedding vectors of the cluster; the assignment and update steps are repeated until the center positions converge; the voxels are divided into K classes.

5. The drilling positioning method based on spatial scanning and surrounding rock characteristic survey as described in claim 4, characterized in that: The process of dividing different punched regions according to voxel type and stability includes: performing connectivity retrieval between voxels according to the improved orientation rules of the three-dimensional mesh in the three-dimensional digital twin model to determine whether each voxel is connected to any adjacent voxel; calculating a stability score for each voxel using the finite element analysis method and setting a local stability threshold; if two voxels belong to the same cluster, the difference in stability between the two voxels is less than the local stability threshold, and they are neighbors on the mesh, then they are classified as the same connected component. Traverse all voxels and extract interconnected sub-regions from the same cluster; voxels within each sub-region are semantically and spatially consistent. For each connected sub-region, the sum of the stability scores of all voxels within the sub-region is calculated to obtain the region stability score of the sub-region. Based on the comparison results of the regional stability score and two step thresholds, all voxels are divided into high-stability, medium-stability, and low-stability regions, and the partition representation of the 3D digital twin model is output. Each region has a regional stability score, a list of candidate holes, regional center coordinates, and spatial topological relationships with adjacent regions. Among them, the high-stability region is equipped with low-density and conventional hole placement, the medium-stability region is equipped with conventional density hole placement, and the low-stability region needs to be reinforced first.

6. The drilling positioning method based on spatial scanning and surrounding rock characteristic survey as described in claim 5, characterized in that: The construction of the comprehensive objective function includes the comprehensive objective value of the region being a weighted synthesis of stability penalty, drilling movement cost and cross-regional constraint penalty, with the optimization objective being to minimize the comprehensive objective value. The cross-region constraint penalty is calculated based on the degree of violation of the cross-region constraint by the region punching scheme; The stability penalty consists of the overall stability of the region and the micro-stability of the location of the hole; the cross-region constraints include the minimum distance constraint between holes, the upper limit constraint of the total number of holes, and the region priority constraint.

7. The drilling positioning method based on spatial scanning and surrounding rock characteristic survey as described in claim 6, characterized in that: The method of using a multi-agent algorithm to update the three-dimensional digital twin model includes treating each connected region obtained by spectral clustering as an agent, and independently maintaining a list of candidate holes and a hole location scheme within the region; Each Agent selects several candidate perforation schemes within the region and checks whether the candidate perforation schemes in this region meet the stability penalty requirements. If there is a conflict with the existing holes in the neighboring area, cross-regional constraint key area priority constraint is added for recalculation to obtain local candidate solutions; construct the comprehensive objective function of this area to obtain the comprehensive objective value of the area; The solution with the lowest comprehensive objective score is selected from the local candidate solutions as the current optimal solution, and the current optimal solution is sent to the neighboring Agents. Each Agent continuously receives the current optimal solution from its neighbors. When it finds that there is a conflict between the current optimal solution of the neighbor and the local current optimal solution in terms of cross-regional hole distance or cross-regional priority order, it automatically adjusts the local candidate solution according to the cross-regional constraints. Every few rounds, the coordinator feeds back the current optimal solution to the 3D digital twin model, updates the digital twin model, and collects the updated current optimal solution information of all agents. It checks whether the solution complies with cross-regional constraints and issues local candidate solution adjustment suggestions to agents that violate cross-regional constraints. When the preset number of iterations is reached, the iteration stops; each Agent submits the final optimal drilling plan; after the coordinator checks the plans output by all regions, it ensures that the cross-regional hole positions meet the cross-regional constraints, the comprehensive target value of the entire region is minimized, and the overall hole layout plan is generated.

8. A system employing the drilling positioning method based on spatial scanning and surrounding rock characteristic survey as described in any one of claims 1 to 7, characterized in that: It includes a multimodal data acquisition module, a digital twin module, a stability assessment module, and a via generation module; The multimodal data acquisition module is used for the automatic triggering, acquisition, registration, and fusion of multi-sensor data during the tunnel boring machine's advancement, and outputs a global point cloud. The digital twin module is used to map the fused point cloud and GPR signal to voxels, and through mapping, generate a three-dimensional digital twin model with mechanical parameter annotations. The stability assessment module is used to calculate the stability score of a region on a three-dimensional digital twin model and obtain connected punched partitions through spectral clustering. The aperture layout generation module is used to construct a comprehensive objective function based on the stability and candidate aperture positions of each region, and output the optimal aperture layout scheme for each region through multi-agent collaborative search.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the drilling and positioning method based on spatial scanning and surrounding rock characteristic survey as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the drilling and positioning method based on spatial scanning and surrounding rock characteristic survey as described in any one of claims 1 to 7.

Images