Rock mass structure plane parameter acquisition and analysis method and system based on three-dimensional scanning

Abstract

This invention provides a method and system for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning, belonging to the field of information technology for rock mass engineering. Using the surface three-dimensional scanning point cloud as the main reference coordinate system, exposed structural surface parameters are extracted through curvature mutation tensor constraint segmentation. The method integrates the hidden attitude and in-situ stress tensor obtained from 4D borehole television image analysis, and uses the discrete mesh method to deduce the discontinuous stress shading field and initially construct the probability density evolution operator for stress sensitivity in the blind zone. Through indoor true triaxial perturbation experiments and acoustic emission full-waveform moment tensor inversion, the crack energy dissipation scalar field is analyzed. Finally, spatiotemporal flow damping boundary conditions are introduced to inversely demodulate the blind zone structural surface evolution attitude, endogenous roughness, and shear displacement parameters. This achieves dynamic inverse solution of blind zone parameters and energy field boundary correction under high stress and strong perturbation environments, and seamlessly reconstructs a three-dimensional digital twin model across scales, improving the accuracy and predictive reliability of deep engineering parameter acquisition and analysis.

Description

Technical Field

[0001] This invention relates to the field of information technology in rock mass engineering, specifically to a method and system for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning. Background Technology

[0002] As global deep underground engineering evolves towards ultra-deep burial depths, ultra-high stress levels, and extremely complex geological environments, higher demands are placed on the acquisition and analysis of rock mass structural parameters at hard rock excavation faces. In particular, real-time monitoring, forward-looking prediction, and the compatibility of deep mechanical responses under dynamic construction environments have become a focus of industry attention. For example, deep-buried inter-basin water diversion tunnels require stable geometric parameters of surrounding rock discontinuities for support design; high-slope underground powerhouse excavation requires wide-range coverage to identify complex joint network characteristics; and deep mining requires blind-spot-free discontinuous surface fields to achieve early warning of dynamic disasters. Currently, 3D scanning and digital television technologies are widely used in underground engineering geometric measurement and surrounding rock logging.

[0003] However, existing rock mass structural surface parameter acquisition and analysis technologies face the following key technical challenges in meeting the above requirements: 1. Insufficient Detection Depth: Traditional rock mass parameter acquisition relies heavily on surface visible light measurements. However, due to the three-dimensional spatial shading effect of the excavation face, only the exposed surface structural surfaces of outcrops and excavation faces can be collected. This results in the inability of current technology to determine the true extension length, spatial connectivity, and three-dimensional closed topology of hidden fractures within deep blind zones of the rock mass. Detection resolution and penetration are severely lacking, making it easy to miss deep safety hazards and affecting the integrity and accuracy of the initial structural surface network. 2. The disconnect between dynamic and static characteristics leads to limited environmental adaptability: Current scanning analysis is essentially a static geometric measurement, which is insufficiently adaptable to deep high-stress, excavation unloading, and strong dynamic disturbance environments. In actual construction environments with strong disturbances, hidden structural surfaces are prone to shear slip, compression closure, or new micro-fracture damage. Existing static geometric parameters and evolutionary models lack a dynamic inverse solution mechanism for mechanical-physical signals under changing loading conditions, making it impossible to characterize and predict the dynamic damage evolution process of structural surfaces. This results in high deviations in the fit of mechanical models in extremely complex geological operations, limiting their reliability and applicability in industrial safety early warning environments.

[0004] To address the aforementioned issues, there is an urgent need for a method and system for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning. This system would organically couple on-site static multi-source detection with dynamic demodulation of true triaxial load-damaged elastic waves in the laboratory, eliminate the illusion of blind zone theoretical deduction, complete the rigid spatial stitching of cross-scale topological networks, and ultimately construct a three-dimensional digital twin model of rock mass structural surfaces that highly conforms to the response of deep mechanics. Summary of the Invention

[0005] Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning, thus resolving the issues mentioned in the background section.

[0007] Technical solution To achieve the above objectives, the present invention provides the following technical solution: a method for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning, the method comprising the following steps: Sp1: Use a three-dimensional digital measurement and scanning terminal to acquire macroscopic geometric point cloud data of the target rock mass excavation surface, and use curvature change tensor constraints to perform anisotropic boundary mesh segmentation on the geometric point cloud data, and automatically compile and extract the first attitude parameters and trace length characteristic functions of the exposed structural surface. Sp2: A 360° panoramic unfolded image of the borehole wall was obtained using a 4D drilling ultra-high-definition television system. The one-dimensional intersecting boundary line of the internal hidden structural surface at the borehole wall truncation point was derived through image semantic analysis and mapping. The second attitude parameter of the internal hidden structural surface was obtained through analysis. Sp3: The in-situ three-dimensional geostress tensor is used as the initial field and injected into the spatial topology network constrained by the first and second attitude parameters. The discontinuous stress shielding field inside the target rock mass is generated by deduction, and then the stress sensitivity probability density evolution operator of the blind zone structural surface inside the rock mass is constructed. Sp4: Under the true triaxial servo loading and unloading environment, the shear slip and micro-fracture damage of the hidden structural plane inside the target rock mass are activated. The elastic wave group signal of the sample endogenous with the multi-channel full dynamic acoustic emission sensor array is captured in real time. The nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field are analyzed by the multi-level wave velocity anisotropic spatiotemporal demodulation algorithm. Sp5: By utilizing the nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field, the evolution of the blind zone structural surface under load, the endogenous roughness coefficient, and the actual shear displacement parameters are dynamically and inversely demodulated. This completes the acquisition of core structural surface parameters of the target rock mass, and based on this, the energy field boundary is corrected for the stress sensitivity probability density evolution operator. Based on the acquired structural surface parameters, a three-dimensional digital twin model of the rock mass structural surface that matches the deep mechanical response is constructed.

[0008] Preferably, Sp1 specifically includes the following sub-steps: Sp1.1: Multi-view image pairs of the target rock mass excavation face are acquired by a high-resolution industrial digital camera, and high-density multi-dimensional macro-geometric point cloud data is generated by a motion recovery structure algorithm based on feature point pair matching. Sp1.2: Calculate the local covariance matrix at each point in the geometric point cloud, extract eigenvalues ​​and calculate Gaussian curvature and mean curvature, and construct curvature mutation tensor constraints; Sp1.3: Based on curvature mutation tensor constraints, anisotropic boundary mesh segmentation is performed on geometric point cloud data. The region growing method is used to segment the geometric point cloud data into independent fitted flat patches, and then the normal vector of each patch is calculated. The first attitude parameter and trace length feature function of exposed structural surfaces are automatically recorded and extracted. The first attitude parameter specifically includes the first dip angle and the first dip value representing the exposed structural surface. The trace length feature function is obtained by statistically analyzing the trace length of the outcrop.

[0009] Preferably, Sp2 specifically includes the following sub-steps: Sp2.1: Insert the 4D drilling ultra-high-definition television probe into a horizontal or inclined hole to acquire full-view video images, and record the current spatial coordinates and time axis deflection angle of the probe in real time through the hole attitude sensor; Sp2.2: Using video stream keyframe extraction and real-time image stitching algorithms, the full-view video images are unfolded and stitched into a single-frame full-cylindrical 2D panoramic image of the hole wall, resulting in a 360° full-cylindrical panoramic unfolded image of the hole wall. Sp2.3: Input the 360° full cylindrical panoramic image of the hole wall into the trained convolutional neural network for semantic segmentation. Through image semantic parsing, the one-dimensional intersecting boundary line of the internal hidden structure surface at the hole wall truncation point is derived. The one-dimensional intersecting boundary line is represented by a sinusoidal curve trajectory line representing the hidden structure surface. Sp2.4: Extract the peaks, troughs, and phase difference parameters of the sinusoidal trajectory line. Combine the probe's spatial coordinates and time axis deflection angle, and use a spatial coordinate analytical transformation algorithm to obtain the second attitude parameters of the internal hidden structural surface. The second attitude parameters specifically include the second dip angle and the second dip value of the internal hidden structural surface.

[0010] Preferably, the stress sensitivity probability density evolution operator for constructing the blind zone structural surface inside the rock mass in Sp3 specifically includes the following sub-steps: Sp3.1: On-site stress measurement using digital hollow inclusion stress gauges is performed to directly obtain the in-situ three-dimensional geostress tensor. At the same time, the three-dimensional spatial coordinate system of the excavation face is used as the main reference coordinate system. The first and second attitude parameters are linearly mapped to the main reference coordinate system to construct a spatial topology network. Sp3.2: Exposed structural surfaces and internally concealed structural surfaces are used as boundary sources. The in-situ three-dimensional geostress tensor is injected into the spatial topology network as the initial field. The discrete grid method is used to construct a mechanical model of the discontinuous stress field of the rock mass. The normal stress distribution and shear stress distribution of any grid unit in the unknown blind zone due to the shading of structural surfaces are calculated. The discontinuous stress shading field inside the target rock mass is deduced and generated. Sp3.3: Establish a stress sensitivity function related to the deformation of structural surfaces and normal and shear stresses. Use the geometric parameters of exposed and concealed structural surfaces as planar and line segment constraints, respectively. Multiply the stress sensitivity function as a weighting factor by the geometric probability distribution function based on Bayesian probabilistic grid estimation. Construct a composite operator to generate the stress sensitivity probability density evolution operator of concealed structural surfaces in the unknown blind zone. Use this operator to quantitatively describe the spatial existence probability, attitude change probability, and connectivity evolution trend of structural surfaces in the blind zone of the rock mass under stress disturbance.

[0011] Preferably, Sp4 specifically includes the following sub-steps: Sp4.1: The rock mass sample prepared according to the excavation face is placed in the self-balancing pressure chamber of the true triaxial mechanical loading frame. The initial first principal stress, second principal stress and third principal stress are set according to the in-situ three-dimensional geostress tensor, and dynamic actuator disturbance load is applied to construct a true triaxial servo loading and unloading environment to activate the shear slip and micro-fracture damage of the hidden structural surface inside the target rock mass. Sp4.2: Using a multi-channel, fully dynamic acoustic emission sensor array arranged around the pressure chamber, the elastic wave group signals generated by structural surface slippage and micro-fractures during the sample loading and damage process are captured in real time, and the arrival time and waveform characteristics of the first arrival wave of each channel are recorded. Sp4.3: The three-dimensional spatial coordinates of the acoustic emission fracture source are calculated based on the time difference positioning algorithm to form acoustic emission spatiotemporal point cluster data. The moment tensor of the micro fracture is inverted using the waveform amplitude. Then, the nonlinear acoustic emission spatiotemporal flow waveform tensor and the crack energy dissipation scalar field are analyzed by the multi-level wave velocity anisotropic spatiotemporal demodulation algorithm.

[0012] Preferably, Sp5 specifically includes the following sub-steps: Sp5.1: The nonlinear acoustic emission space-time flow waveform tensor resolved in Sp4 is mapped to the main reference coordinate system. A density-based noise space clustering algorithm is used to filter out scattered unstructured noise points and identify the macroscopic nonlinear envelope surface of the acoustic emission point cloud cluster. Sp5.2: Introducing the space-flow damping boundary condition for acoustic emission elastic wave energy dissipation, by calculating the joint distribution of the central fitting attitude of the macroscopic nonlinear envelope surface of the acoustic emission point cloud and the crack energy dissipation scalar field, the evolution attitude, endogenous roughness coefficient and real displacement shear displacement parameters of the blind zone structural surface as the loading state change are dynamically and inversely demodulated, thus completing the acquisition of core structural surface parameters of the target rock mass. Among them, the evolution attitude specifically includes the evolution dip angle and evolution tendency. Sp5.3: Using the collected core structural surface parameters, the energy field boundary correction is performed on the stress sensitivity probability density evolution operator in Sp3 to update the spatial existence probability, attitude change probability and connectivity evolution trend in a linked manner. In the blind zone grid range, the Monte Carlo random sampling algorithm is used to generate a continuous topological grid structure that can match the energy dissipation characteristics of the elastic wave. Sp5.4: Spatially closes and stitches continuous topological mesh structures with exposed and concealed structural surfaces, and constructs a three-dimensional digital twin model of rock mass structural surfaces that matches the deep mechanical response based on the collected structural surface parameters.

[0013] Preferably, the system includes a macroscopic geometry acquisition module, a borehole topology acquisition module, a discontinuous field evolution module, a dynamic field analysis module, and a parameter output and twinning module, wherein: The macro-geometry acquisition module is used to acquire macro-geometry point cloud data of the target rock mass excavation surface using a three-dimensional digital measurement and scanning terminal. It performs anisotropic boundary mesh segmentation on the geometric point cloud data through curvature change tensor constraints, and automatically compiles and extracts the first attitude parameters and trace length characteristic functions of the exposed structural surfaces. The borehole topology acquisition module is used to acquire a 360° full cylindrical panoramic unfolded image of the borehole wall using a 4D drilling ultra-high-definition television system. Through image semantic parsing, it maps and evolves the one-dimensional intersecting boundary line of the internal hidden structural surface at the borehole wall truncation point, and obtains the second attitude parameter of the internal hidden structural surface through parsing. The discontinuity field evolution module is used to inject the in-situ three-dimensional geostress tensor as the initial field into the spatial topology network constrained by the first and second attitude parameters, and to deduce and generate the discontinuous stress shielding field inside the target rock mass, thereby constructing the stress sensitivity probability density evolution operator of the blind zone structural surface inside the rock mass. The dynamic field analysis module is used to activate the shear slip and micro-fracture damage of the hidden structural plane inside the target rock mass under the true triaxial servo loading and unloading environment. It uses a multi-channel full dynamic acoustic emission sensor array to capture the elastic wave group signal of the sample in real time, and uses a multi-level wave velocity anisotropic spatiotemporal demodulation algorithm to analyze the nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field. The parameter output and twin module is used to dynamically demodulate and output the evolution attitude, endogenous roughness coefficient and actual fault shear displacement parameters of the blind zone structural surface using the nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field. This completes the acquisition of core structural surface parameters of the target rock mass, and performs energy field boundary correction on the stress sensitivity probability density evolution operator accordingly. Based on the acquired structural surface parameters, a three-dimensional digital twin model of the rock mass structural surface that matches the deep mechanical response is constructed.

[0014] Preferably, the dynamic field analysis module is connected to a true triaxial test subsystem and an acoustic emission subsystem. The true triaxial test subsystem includes a loading frame, a self-balancing pressure chamber, a hydraulic servo system, and a dynamic actuator, used to apply triaxial independent and non-interfering high-stress loads and dynamic disturbance loads. The acoustic emission subsystem includes a high-pressure and high-temperature resistant acoustic emission sensor array arranged inside the pressure chamber and a multi-channel digital signal acquisition instrument.

[0015] Beneficial effects This invention provides a method and system for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning. It has the following beneficial effects: 1. This invention breaks through the physical bottleneck of traditional visible light measurement being limited to the surface by deeply integrating static multi-source detection with mechanical fields. The system automatically records the first orientation parameter and trace length characteristic function of the exposed surface from the three-dimensional scan point cloud by constructing curvature mutation tensor constraints. It further links the neural network semantic analysis of 4D borehole television images to obtain the second orientation parameter, and injects the measured in-situ three-dimensional geostress tensor as the initial mechanical boundary into the spatial topology network. It uses the discrete grid method to quantitatively calculate the normal stress and shear stress distribution in the blind zone, and deduce the discontinuous stress shading field and stress sensitivity probability density evolution operator. This mechanism can penetrate the visible blind zone, scientifically deduce and quantitatively describe the spatial existence probability and connectivity evolution trend of hidden fractures in the deep interior, effectively solve the technical problems of unknown true extension length, spatial connectivity and three-dimensional closed shape caused by the shading effect of traditional methods, greatly improve the integrity and accuracy of the initial network of the structural surface, and avoid missing deep safety hazards.

[0016] 2. This invention actively activates the shear slip and micro-fracture damage of the hidden structural surfaces inside the sample under highly realistic triaxial high stress and dynamic actuator disturbance loads. It utilizes a multi-channel, fully dynamic acoustic emission sensor array to capture elastic wave group signals in real time. Through time-difference positioning with velocity anisotropy correction and moment tensor inversion algorithms, it calculates and analyzes the nonlinear acoustic emission spatiotemporal flow waveform tensor and the crack energy dissipation scalar field. Based on this, it introduces elastic wave energy dissipation spatiotemporal flow damping boundary conditions and dynamically demodulates the evolution of the blind zone structural surfaces. The process of obtaining parameters such as the surface condition, intrinsic roughness coefficient, and actual displacement shear displacement transforms purely static geometric measurement into real-time dynamic inverse solution of mechanical-physical signals under changing loading conditions. This process can accurately characterize the dynamic damage evolution process of the structural surface and perform energy field boundary correction on the stress sensitivity probability density evolution operator, eliminating the probability illusion of purely theoretical deduction. It significantly improves the system's adaptability to extreme environments such as deep high stress, excavation unloading, and strong dynamic disturbances, ensuring a high degree of consistency between the digital twin model and the actual mechanical response of deep engineering. Attached Figure Description

[0017] Figure 1This is a flowchart of the parameter acquisition and analysis method of the present invention; Figure 2 This is a structural diagram of the parameter acquisition and analysis system of the present invention; Figure 3 This is a diagram showing the main interface navigation of the system of the present invention; Figure 4 This is the interface for displaying the evolution parameters of the blind zone structural surface of the system of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Specific Implementation Example 1: This paper presents a method and system for acquiring and analyzing rock mass structural surface parameters based on 3D scanning. Taking deep rock excavation projects (such as the excavation of large underground hydraulic tunnels and the advancement of mining faces in deep, high-stress mines) as a typical implementation background, it addresses the issue that in deep, high-stress environments, the excavation face often contains numerous unknown, discontinuous structural surfaces. These surfaces are highly susceptible to shear instability under excavation unloading and external dynamic disturbances, posing a significant threat to engineering safety. To achieve accurate acquisition and mechanical twin reconstruction of the core structural surface parameters of these blind areas, a hardware and software collaborative analysis system was constructed, encompassing macroscopic scanning of the working face, static detection of deep boreholes, in-situ determination of geostress, and indoor true triaxial dynamic inversion. The overall architecture and specific application methods of the system in the implementation field are as follows: 3D Digital Measurement and Scanning Terminal: Physically, this terminal mainly consists of a high-precision ground-based 3D laser scanner (LiDAR), a high-resolution industrial digital camera (45 million pixels or more), an integrated and stable tripod, a data transmission optical cable, and an external mobile workstation. Its core function is to realize the digital reconstruction of the geometric morphology of exposed rock masses that are visible to the naked eye and can be directly contacted. Through non-contact optical ranging and image capture, it can acquire a high-density point cloud containing millions of three-dimensional spatial coordinates (X,Y,Z) and reflectivity information within seconds, so as to capture the macroscopic morphology of exposed rock surface structures such as joints, fissures, and bedding caused by excavation and unloading without blind spots.

[0020] Specific application method: Within a safe step distance after the excavation of deep underground long tunnels, large underground caverns, or mined-out areas, engineers set up terminals on stable ground and acquire macroscopic geometric point cloud data of the target rock mass excavation surface through multi-view scanning at mutually intersecting angles. This data serves as the geometric boundary source for the entire method and system operation and is called by the background control system. By calculating the local covariance matrix at each point, eigenvalues ​​are extracted and Gaussian curvature and mean curvature are calculated, thereby constructing curvature mutation tensor constraints. This provides a unique and true initial geometric boundary for subsequent anisotropic boundary mesh segmentation, automatic recording and extraction of the first attitude parameters and trace length characteristic functions of exposed structural surfaces.

[0021] 4D Drilling Ultra-High-Definition Television System: Physically, this system includes a panoramic cylindrical imaging probe, a high-strength cable counting winch, an in-hole attitude sensor (integrating a high-precision spatial three-axis electronic compass and a microelectromechanical system MEMS gyroscope), and a ground-based digital image acquisition controller. The probe front end integrates a high-brightness LED cold light source and a panoramic fisheye optical lens. This system aims to extend human vision and solve the blind spot problem that external scanning cannot see the hidden fractures deep in the rock mass. Its core function is to obtain the one-dimensional intersecting boundary line of the hidden internal structural surface at the cut-off point of the borehole wall in the deep rock mass. By advancing the probe at a constant speed, the microscopic morphology of the 360° full cylindrical surface of the borehole wall is converted into a distortion-free continuous video stream, and the spatial coordinates and angles are bound in real time through the attitude sensor.

[0022] Specific application method: Observation holes are drilled obliquely or horizontally towards the deep edge of the rock mass at the tunnel face or the edge of the excavation face. Imaging probes are sent into the holes to collect full-view video images. During the movement, especially during the advancement of horizontal and oblique holes, the attitude sensor inside the hole records the spatial coordinates and time axis deflection of the probe in real time to correct geometric errors caused by borehole bending or probe spin. The ground controller outputs a 360° full-cylindrical panoramic image of the borehole wall through a real-time image stitching algorithm and inputs it into a pre-trained convolutional neural network for semantic segmentation to identify the sinusoidal curve trajectory line representing the hidden structural surface. Then, through a spatial coordinate analysis and transformation algorithm, the second attitude parameters (i.e., the second dip angle and the second dip value) of the internal hidden structural surface are obtained.

[0023] Digital hollow inclusion stress gauge: This structure belongs to the in-situ geostress measurement component. It mainly consists of a hollow inclusion stress gauge cylinder (with at least 12 or 12 channels of high-precision resistance strain gauges symmetrically bonded to its inner or outer walls in the circumferential, longitudinal, and oblique directions), a special casing release drilling tool, a multi-channel static digital strain gauge, and a high-precision signal transmission shielded cable. Geostress is the dynamic source of all deformation and fracture in deep rock masses. The core function of this structure is to obtain the complete in-situ three-dimensional geostress tensor before excavation disturbance by direct measurement, thereby providing the key initial stress dynamic field boundary conditions for this invention and avoiding the drawbacks of traditional technology that relies on pure geometry and blind guessing of blind joints.

[0024] Specific application method: Drill a geostress measurement borehole on site, insert a hollow inclusion stress gauge to a specified depth, and bond it tightly to the borehole wall with special epoxy resin. After the resin has completely cured, use a large-diameter casing drill to unload the casing. During the unloading process, the rock mass releases in-situ stress, causing the inclusion cylinder to undergo small elastic strain. A static digital strain gauge records the strain curves of multiple strain gauges in real time. By inversely solving the elastic-plastic theory formula, the in-situ three-dimensional geostress tensor is directly obtained. This tensor has two roles in the system application: first, it is injected into the spatial topology network as an initial field to construct a mechanical model with the discrete mesh method, calculate the stress distribution in the unknown blind zone, and thus construct a stress sensitivity probability density evolution operator; second, it is used as the initial control parameter for pressure holding in the laboratory true triaxial test.

[0025] True Triaxial Test Subsystem: This subsystem is an indoor dynamic inversion and feature activation hardware. Its physical structure includes a high-rigidity true triaxial loading frame, a high-pressure self-balancing pressure chamber, a hydraulic servo system (containing multiple independently controlled electro-hydraulic servo valves), and dynamic actuators. The pressure chamber is equipped with three completely independent and non-interfering rigid pressure heads. Its core function is to perfectly reproduce the real loading environment of deep rock under high underground stress and strong excavation disturbance in the test chamber. Since the deep rock mass is in a triaxial asymmetric stress state where the first principal stress, second principal stress, and third principal stress are not equal, this subsystem artificially activates and induces shear slip, instability, and micro-fracture damage of the hidden structural surfaces inside the rock mass by applying asymmetric static and dynamic loads, causing it to radiate elastic waves and provide a physical signal source for subsequent reverse parameter acquisition.

[0026] Specific application method: Standard cubic rock samples are prepared according to the on-site tunneling conditions and placed in a self-balancing pressure chamber. The system calls the in-situ three-dimensional geostress tensor measured on-site and drives the rigid pressure head through a hydraulic servo system to apply the corresponding first, second, and third principal stresses respectively. On the basis of maintaining triaxial static load balance, the dynamic actuator is activated through the background controller to apply dynamic disturbance loads of specific frequency and amplitude (simulating engineering blasting or mechanical tunneling impact) to the sample, thereby actively stimulating the deformation response of the blind zone structural surface.

[0027] Acoustic emission subsystem: This subsystem is a high-sensitivity signal acquisition and analysis hardware. Physically, it mainly consists of a multi-channel fully dynamic acoustic emission sensor array (containing at least 8 high-pressure and high-temperature resistant piezoelectric ceramic AE sensors), a preamplifier (with adjustable gain and wideband filtering function), a multi-channel high-speed digital signal acquisition instrument, and a full-waveform moment tensor analysis chip. The sensors are embedded around the pressure head of the true triaxial pressure chamber through specially designed pressure-bearing sealed channels. Its function is similar to a micro seismic network. When the internal structural surface of the rock mass undergoes shear displacement or micro-crack initiation due to disturbance, energy is released instantaneously in the form of elastic waves (acoustic emission). This subsystem is used to capture the series of elastic wave group signals generated by the sample in real time and in full waveform, and convert them into digital electrical signals to locate the fracture location, identify the fracture properties, and quantitatively calculate the energy.

[0028] Specific application method: During the entire true triaxial disturbance test, the sensor array monitors the elastic wave in real time at a sampling rate of megahertz (MHz). The arrival time of the first wave and the full waveform tensor captured by each channel are stored in real time. The processor calls the wave velocity anisotropic spatiotemporal demodulation algorithm, combined with time difference positioning and full waveform moment tensor inversion, to analyze the nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field. Subsequently, the data is transmitted to the core algorithm layer for DBSCAN noise clustering and macroscopic nonlinear envelope fitting in the principal reference coordinate system. Then, it dynamically inversely demodulates the evolution of the blind zone structural surface, the endogenous roughness coefficient and the actual shear displacement parameters, and finally drives the construction of a three-dimensional digital twin model of the rock mass structural surface.

[0029] First, static data acquisition is performed on-site using a 3D digital measurement scanning terminal and a 4D drilling ultra-high-definition television system to obtain the planar and linear geometric boundaries of the visible area. Next, the in-situ stress field measured by a digital hollow inclusion stress gauge is introduced. Within the software system (discontinuous field evolution module), a stress sensitivity probability density evolution operator capable of predicting cracking trends in unknown blind zones is constructed. To eliminate the unreliability of purely probabilistic inference, the system further integrates a true triaxial testing subsystem and an acoustic emission subsystem. In the laboratory, real ground stress data measured on-site is used as input, and under disturbance... The system activates the mechanical damage of the structural surface under the condition, captures the elastic waves radiated endogenously during the slip of the structural surface, and finally transforms the elastic wave waveform tensor and energy field back to the engineering site coordinates. It dynamically demodulates the precise mechanical characteristic parameters of the blind zone structural surface and uses them to correct the energy field boundary of the previous probability density evolution operator. It completes the Monte Carlo topology network reconstruction and spatial mesh stitching. Through such a highly collaborative physical structure and application logic, the system finally outputs a three-dimensional digital twin model of the rock mass structural surface, realizing a technological leap from pure geometric guessing to conforming to the deep mechanical response.

[0030] For further details, please refer to Figure 1 As shown, the specific operation process of the rock mass structural surface parameter acquisition and analysis method based on three-dimensional scanning is as follows: First, the system initiates macroscopic morphological capture of the exposed, directly visible tunnel face, executing Sp1: based on 3D scanning and curvature mutation tensor constraints, extracting the first attitude parameter and trace length feature function of the exposed structural surface. This step achieves pixel-level accurate recording of the geometric morphology of the exposed area of ​​the excavation face, specifically through the following sub-steps: Sp1.1: Acquisition of Macroscopic Geometric Point Cloud Data: Within the safe step distance of the excavation face after completion, the aforementioned 3D digital measurement and scanning terminal is set up. A high-precision ground-based 3D laser scanner within the terminal emits laser pulses towards the working face. The spatial relative coordinates of each laser reflection point are calculated in real time using the Time-of-Flight (TOF) method. Simultaneously, a high-resolution industrial digital camera in the 3D digital measurement and scanning terminal captures multi-view image pairs of the working face from three mutually 60-degree angles. After receiving the multi-view image pairs, the external mobile workstation formally runs a motion reconstruction structure algorithm based on feature point pair matching to generate high-density, multi-dimensional macroscopic geometric point cloud data. The underlying mathematical implementation and calculation process for obtaining the point cloud is as follows: The first step, feature point extraction and matching: A scale-space Gaussian pyramid is built using the Scale Invariant Feature Transform (SIFT) operator. By detecting the extreme points in the Difference Gaussian space (DoG), cell-level feature points with scale and rotation invariance are extracted from each image. A one-dimensional high-dimensional feature descriptor vector is constructed for the feature points of any two images. Preliminary matching of feature point pairs is performed by calculating the ratio of the nearest neighbor to the second nearest neighbor in Euclidean distance. The Two-Way Random Sampling Consensus (RANSAC) algorithm is introduced, and a single-point reprojection error threshold for the homography matrix is ​​set to eliminate mismatched point pairs caused by illumination and noise, thereby obtaining precise pairs of identically named feature points.

[0031] The second step is camera pose recovery and sparse reconstruction: Let the homogeneous coordinates of a pair of corresponding feature points in the two views be respectively... and According to the principle of epipolar geometry, it satisfies the following matrix relationship formula: ; In the formula, for The system solves for the fundamental matrix using the eight-point method. The essential matrix is ​​calculated by combining the camera intrinsic parameter matrix. Using Singular Value Decomposition (SVD) to transform the essential matrix Decompose into rotation matrices Translation vector This allows the spatial relative pose of the multi-view camera at the moment of shooting to be recovered. Then, using the principle of triangulation, ray spatial intersection is performed on all pairs of feature points with the same name to calculate the three-dimensional coordinates of these feature points in space, forming a sparse point cloud.

[0032] The third step is multi-view stereo dense matching and point cloud fusion: using sparse point cloud and camera pose. As the initial seed, the Multi-View Stereo Geometric Surface Mesh Dense Matching (PMVS) algorithm is run. Pre-block growth and depth map estimation are performed around each spatial seed point. The Normalized Cross-Correlation (NCC) coefficient of adjacent pixels is calculated. When this coefficient is greater than a preset threshold (e.g., 0.85), dense matching is considered successful. Finally, this dense point cloud is rigidly transformed and fused with the spatial relative coordinate mesh directly measured by a terrestrial 3D laser scanner using the Time-of-Flight (TOF) method. After conventional filtering processes such as dust and blind zone noise removal in the principal reference coordinate system, the final result containing three-dimensional spatial coordinates (…) is calculated. High-density, multi-dimensional macroscopic geometric point cloud data with reflectivity information.

[0033] Sp1.2: Constructing the Curvature Abrupt Tensor Constraint Field: The system's background processor performs global spatial indexing on the generated high-density, multi-dimensional macroscopic geometric point cloud data. To automatically locate the physical boundaries of joints and fracture surfaces in the massive point cloud, the system constructs the curvature abrupt tensor constraint field through the following specific discrete differential geometric calculations: For any spatial control point in the point cloud... The system utilizes spatial boundary constraints The neighborhood search algorithm extracts the nearest Euclidean distance. Neighboring control points (in this embodiment) (The value is 25), using this Calculate the spatial covariance matrix of the local point cloud cluster using the spatial coordinates of its neighboring points. Its mathematical expression formula is: ; In the formula, For this The geometric center coordinates of each neighboring control point, the processor for a symmetric positive definite matrix Eigenvalue decomposition yields three non-negative eigenvalues. These three eigenvalues ​​quantitatively characterize the spatial distribution of the neighborhood point set along three orthogonal principal directions. The system uses these three eigenvalues ​​and the local quadratic surface fitting equation to calculate the Gaussian curvature at that point. With mean curvature The specific formula for calculating the governing equation is as follows: ; ; Inside a perfectly flat structural surface Approaching zero, at the intersection boundary or trace break of two structural planes with different attitudes, due to the drastic change in spatial geometric convexity and concavity, the Gaussian curvature and mean curvature undergo discontinuous abrupt changes. The system constructs a second-order symmetric tensor at each control point by composing the first-order spatial gradient vector of the Gaussian curvature and the mean curvature with the second-order Hessian matrix as differential tensors, thus ultimately constructing a curvature abrupt change tensor constraint. The maximum principal eigenvalue of this tensor represents the maximum intensity of the curvature abrupt change, and its corresponding eigenvector points to the normal direction where the curvature jump is most intense, thus providing a rigid differential geometric constraint for boundary segmentation.

[0034] Sp1.3: Automatically segmenting facets and fitting to extract the first attitude parameter and trace length feature function: After obtaining the curvature change tensor constraint field, the system calculates the final first attitude parameter and trace length through the following grid clustering and geometric inverse solution steps: The first step is region growth and segmentation based on curvature constraints: The system selects the point with the smallest rate of curvature change in the point cloud as the seed point for growth and spreads it to the surrounding neighborhood. During the spread, the system calculates the difference between the largest eigenvalue of the curvature change tensor between the current point and the neighboring points in real time, and sets a preset flatness threshold. (In this embodiment, the value is set to 0.02 based on the hard rock morphology characteristics). If the difference is less than... If the difference is greater than or equal to 1, then the neighboring point is merged into the current patch point set; (Indicating that the edge of a fissure or a change in orientation has been reached), growth stops and the edge is locked. After traversing the point cloud, the high-density point cloud is automatically divided into several flat patch point sets representing independent exposed structural surfaces.

[0035] The second step is the geometric inverse solution calculation of the attitude parameters: For each segmented independent flat surface point set, the least squares method is used to fit its spatial plane equation. According to the first-order optimality condition, the standard unit normal vector of the fitted plane in the principal reference coordinate system is... for: ; Given the standard unit normal vector, the system directly calculates the first attitude parameter, i.e., the first dip angle, of the exposed structural surface based on the analytical geometric projection mapping relationship. and first tendency value The specific calculation formula is as follows: ; ; The third step involves statistical fitting of the trace length characteristic function: For each flat surface locked by a curvature abrupt boundary, the system extracts the convex hull mesh of its boundary polygon, calculates the Euclidean distance of the maximum fracture trace in the three-dimensional space of the face of the exposed surface, and denoted as the trace length of a single crack. The system performs clustered statistical analysis on the trace length samples of all structural surfaces segmented from the entire wall, introduces the maximum likelihood estimation method (MLE), and fits them to a probability density function of a negative exponential distribution or a log-normal distribution, thereby calculating the trace length characteristic function that characterizes the scale distribution law of the exposed structural surfaces in the current region. : ; In the formula, Because the trace of the head is long, The crack line density parameters are estimated by maximum likelihood estimation from the inverse solution.

[0036] After completing the acquisition of the surface geometric features (first attitude parameter and trace length characteristic function) of the visible area of ​​the tunnel face using the aforementioned precise mathematical operators and 3D scanning, due to the existence of an information black box at the deep edge of the rock mass that cannot be penetrated by the naked eye or laser scanning, in order to extend the geometric detection grid to the deep topology of the rock mass, the system further integrates in-hole optical detection and deep learning methods to execute Sp2: acquiring panoramic images of the borehole wall through a 4D borehole ultra-high-definition television system, segmenting intersecting boundaries through a convolutional neural network, and inversely solving for the second attitude parameter, specifically achieved through the following sub-steps: Sp2.1 and Sp2.2: A 4D drilling ultra-high-definition television probe is inserted into a horizontal or angled hole to acquire full-view video images. An in-hole attitude sensor records the probe's spatial coordinates and time axis deflection in real time. Using video stream keyframe extraction and real-time image stitching algorithms, the full-view video images are unfolded and stitched into a single 2D panoramic image of the hole wall's full cylindrical surface, resulting in a 360° panoramic image of the hole wall's full cylindrical surface. Specifically: The first step is the synchronous acquisition of video and attitude data: In a pre-drilled observation blind hole (such as a horizontal or inclined hole extending into the depths), a cable counting winch is used to uniformly advance the panoramic cylindrical imaging probe of the 4D borehole ultra-high-definition television system. Under cold light illumination, the fisheye lens integrated at the probe's front end records a full-view video stream of the borehole wall at a frequency of thirty frames per second. During this process, the winch encoder outputs the probe's drilling depth in real time along the current time axis. Simultaneously, the borehole attitude sensor integrated inside the probe (a spatial three-axis electronic compass and a microelectromechanical system gyroscope) reads and records the probe's spatial coordinates in the main reference coordinate system at a real-time high frequency of fifty Hz. And the time axis deflection (including azimuth angle) that fluctuates with the time axis. and tilt angle .

[0037] The second step is the stitching and reconstruction of the 2D panoramic image: After receiving the video stream, the external workstation first runs the video stream keyframe extraction algorithm to calculate the chi-square distance of the grayscale histogram of the preceding and following frames. When the distance is greater than a set threshold, it is retained as a keyframe, thereby eliminating a large number of repetitive and redundant frames during the constant speed advance of the probe. Then, a real-time image stitching algorithm based on cylindrical projection transformation is adopted to project the keyframe pixels of the 3D cylindrical video onto the latitude and longitude grid of the 2D plane. According to the time axis deflection angle and depth coordinates, the homography matrix of adjacent image grids is calculated. Through cylindrical multi-scale image fusion and overlapping edge pixel fade-in and fade-out smoothing processing, the spin and bending errors caused by the unevenness of the probe hole wall are eliminated. Finally, the continuous cylindrical image stream is seamlessly stitched into a single, undistorted 360° full cylindrical panoramic unfolded image of the hole wall.

[0038] Sp2.3: The 360° full-cylindrical panoramic image of the borehole wall is input into a pre-trained convolutional neural network for semantic segmentation. Through image semantic parsing and mapping, a one-dimensional intersection boundary line of the internal hidden structural surfaces at the borehole wall truncation point is derived. This one-dimensional intersection boundary line is represented by a sinusoidal curve trajectory line representing the hidden structural surfaces. Specifically: since the spatial truncation intersection line between the planar hidden structural surfaces and the cylindrical borehole is mathematically presented as a standard sine curve in the two-dimensional unfolded image of the cylindrical surface, in order to automatically remove this topological boundary in complex rock textures and scratches, the system performs image semantic parsing and mapping: the system uses the generated 360° full-cylindrical panoramic image of the borehole wall as the full input tensor and inputs it into a pre-trained deep learning convolutional neural network (a dedicated network for semantic segmentation of rock joints and fractures based on the MobileNet-Unet architecture). The image passes through multiple continuous layers in the network's contraction path. Convolutional kernels and max pooling layers are used to extract high-dimensional semantic features, capturing low-contrast and discontinuous texture features of the edges of hidden structural surfaces. Then, in the expansion path, upsampling is performed through transposed convolution, and the spatial geometric position information of shallow layers is fused through skip connections. Finally, the network performs pixel-level binarization probability classification at the output, automatically filtering out unstructured background noise such as drill bit scratches and mud contamination on the borehole wall, and accurately segmenting and extracting the one-dimensional intersecting boundary lines of the internal hidden structural surfaces at the borehole wall truncation point. This set of intersecting boundary lines is represented in the image topology as a spatial sine curve trajectory representing the hidden structural surfaces.

[0039] Sp2.4: Extracts the peaks, troughs, and phase difference parameters of the sinusoidal trajectory line. Combining the probe's spatial coordinates and time axis deflection, a spatial coordinate analytical transformation algorithm is used to analytically obtain the second attitude parameters of the internal hidden structural surface. Specifically, the second attitude parameters include the second dip angle and second dip value of the internal hidden structural surface. In detail: the system extracts contour pixels on the target sinusoidal trajectory line of the binarized segmentation image, and through mathematical extreme value search, accurately identifies and extracts the peak coordinates of the sinusoidal trajectory line. Valley coordinates According to the panoramic unfolded diagram, the horizontal axis represents the hole wall. The circumferential angle and the vertical axis represent the geometric mapping relationship of the current actual borehole depth, and the amplitude of the sine curve is statistically calculated. (That is, half the difference between the ordinates of the peaks and troughs:) The initial phase difference parameter is the angle of the x-coordinate corresponding to the lowest point of the curve. Subsequently, the system uses a sinusoidal spatial coordinate analytical transformation algorithm to solve the image features into the true mechanical boundary parameters. Let the actual radius of the observed borehole be... The current truncation position is located at the probe spatial coordinates recorded and bound by the attitude sensor. The borehole axis deflection angle specifically includes the azimuth angle of the borehole axis. Inclination angle of the hole axis The normal vector of the hidden structural plane in the hole-axis coordinate system can be obtained by inverse solving the amplitude and phase of the sine curve. The system transforms the hole-axis coordinate system back to the principal reference coordinate system through a coordinate rotation matrix, thereby deriving the three-dimensional analytical equation of the hidden structural plane in the principal reference coordinate system. Ultimately, based on the plane normal vector The analytical solution directly calculates and analyzes the second attitude parameters of the internal concealed structural surface, which includes the second dip angle of the internal concealed structural surface. Compared with the second tendency value The specific mathematical transformation formula is as follows: ; ; At this point, the system has completed the precise inverse solution acquisition of static topological boundary parameters that are invisible to the naked eye and hidden at the truncated section of the borehole wall inside the rock mass.

[0040] Up to Sp2, the system acquired the planar geometric features of the excavation face surface (first attitude parameter and trace length characteristic function) through surface 3D scanning, and the local linear intersecting topological features of the deep rock mass (second attitude parameter) through borehole ultra-high-definition television. However, the data collected in the above two steps are only discrete visual islands in spatial distribution. In the vast unknown hard rock blind box that the scanner cannot "see" and the borehole cannot "penetrate", the distribution and evolution of fractures still lack solid information support.

[0041] Therefore, the actual in-situ three-dimensional geostress measured on-site is used as the deep mechanical power source. The mechanical causality brought about by stress shielding is used to predict the development probability of blind zone fractures. Sp3 is executed: the in-situ three-dimensional geostress tensor is used as the initial field and injected into the spatial topology network constrained by the first and second attitude parameters. The discontinuous stress shielding field inside the target rock mass is generated by deduction, and then the stress sensitivity probability density evolution operator of the blind zone structural surface inside the rock mass is constructed. This is achieved through the following sub-steps: Sp3.1: In-situ stress measurement using a digital hollow inclusion stress gauge is performed to directly obtain the in-situ three-dimensional geostress tensor. Simultaneously, using the three-dimensional spatial coordinate system of the excavation face as the primary reference coordinate system, the first and second attitude parameters are linearly mapped and transformed to this primary reference coordinate system, constructing a spatial topology network. Specifically: The first step is to determine and obtain the in-situ stress tensor: Stress measurement holes are drilled in the vicinity of the excavation site. A digital hollow inclusion stress gauge is inserted to a specified depth and tightly bonded to the hole wall with epoxy resin. In-situ measurements are performed using the casing unloading method. The release of in-situ stress in the rock mass causes minute elastic strain in the inclusion cylinder. A multi-channel static digital strain gauge records the entire strain curve of twelve high-precision resistance strain gauges in real time at high frequency. Based on the spatial elastoplastic mechanics theory equations, the strain values ​​are inversely solved to directly obtain the in-situ three-dimensional stress tensor. Mathematically, it is represented as a second-order symmetric tensor matrix constructed from the magnitudes of three orthogonal principal stresses and their azimuth and inclination angles: .

[0042] The second step is the unified transformation of multi-source spatial coordinates and the initial network construction: taking the lower left corner of the tunnel face as the spatial reference origin, a three-dimensional independent spatial rectangular coordinate system is established as the main reference coordinate system. The system calls the spatial three-dimensional homogeneous coordinate rotation and translation matrix to map and linearly transform the exposed structural surface fitting mesh, trace length characteristic function (corresponding to the first attitude parameter) obtained in Sp1, and the borehole intersection mesh and sine curve spatial analytical equation (corresponding to the second attitude parameter) obtained in Sp2 to this main reference coordinate system. The system uses the transformed known surface structural surface mesh and the internal hidden structural surface intersection segment mesh as discrete mechanical boundary sources, and performs seamless edge stitching and topological linking in the main reference coordinate space to initially construct a spatial topological network strongly constrained by known geometric boundaries.

[0043] Sp3.2: Exposed and concealed structural surfaces are used as boundary sources. The in-situ three-dimensional geostress tensor is injected into the spatial topology network as the initial field. A mechanical model of the discontinuous stress field of the rock mass is constructed using the discrete mesh method. The normal stress distribution and shear stress distribution of any grid cell in the unknown blind zone due to structural surface shading are calculated. The discontinuous stress shading field inside the target rock mass is deduced and generated. Specifically: Since exposed and concealed structural surfaces are mechanically fractured media, their normal tensile strength is extremely low, and the shear fracture surface cannot completely transmit shear force. Therefore, under high stress environment, they will have a significant mechanical shading effect on the surrounding hard rock mesh. The system uses the discrete grid method to divide the vast, unknown hard rock blind zone within the spatial topological network. A spatial micro-finite element grid is used to construct a mechanical model of the discontinuous stress field in the rock mass, and the system incorporates the in-situ three-dimensional geostress tensor. The initial field is introduced into the mechanical model as an infinite boundary load, and the exposed structural surfaces and the internally concealed structural surfaces are solidified into internal discontinuous mechanical contact work sources. The computational engine iteratively solves the spatial stress equilibrium equations that satisfy the boundary discontinuous displacement constraints, and calculates the normal stress distribution of any mesh element in the unknown blind zone due to the shading of the structural surfaces. and shear stress distribution The governing equations of the force matrix of its element nodes are: ; In the formula, For including the contact nonlinear stiffness matrix, Let be the displacement vector of the blind zone mesh node. Given the initial equilibrium load vector of the ground stress, after solving for the nodal displacements, the normal stress at each point within the mesh is calculated using the strain geometry equation and the constitutive equation. With shear stress After traversing the stress calculation results of all blind zone grid cells, the system generates the discontinuous stress shielding field inside the target rock mass in memory space.

[0044] Sp3.3: A stress sensitivity function is established to relate structural surface deformation to normal and shear stresses. The geometric parameters of exposed and concealed structural surfaces are used as planar and line segment constraints, respectively. The stress sensitivity function is multiplied by a geometric probability distribution function based on Bayesian probabilistic grid estimation as a weighting factor. This composite construction generates a stress sensitivity probability density evolution operator for concealed structural surfaces within unknown blind zones. This operator is used to quantitatively describe the spatial existence probability, attitude change probability, and connectivity evolution trend of structural surfaces in blind zones of rock masses under stress disturbance. Specifically, to eliminate the inherent flaws of traditional techniques that rely on purely geometric and blindly random guesses about blind zone fractures, this invention, based on the discontinuous rock mechanics cracking and failure mechanism, derives the core operator through the following mathematical composite calculations: First, a stress sensitivity function is systematically established to characterize shear slip or compressive closure deformation of structural surfaces under different stress states. It is based on the Mohr-Coulomb fracture criterion and the change in Coulomb energy, and its specific mathematical form is as follows: ; In the formula, The internal friction angle of hard rock. For cohesion, To determine the uniaxial compressive strength of the rock, the system uses the geometric parameters of exposed structural surfaces (planar mesh) and the geometric parameters of internally concealed structural surfaces (line segment mesh) as planar and line segment constraints for spatial topological mesh growth, respectively. A Bayesian probabilistic mesh estimation algorithm is then run to calculate the geometric probability distribution function of the unknown blind zone mesh under pure geometric topological evolution. Next, the system performs nonlinear composite calculations, multiplying the stress sensitivity function as a weighting factor by the geometric probability distribution function, thereby constructing a stress sensitivity probability density evolution operator for hidden structural surfaces in the unknown blind zone within the blind zone grid space. The mathematical expression of this operator is as follows: ; In the evolution operator calculated above, if a certain grid cell within the blind zone... In a discontinuous stress shielding field, the stress sensitivity weight value is located in a high shear stress zone or a low tensile stress zone (i.e., a mechanically vulnerable part of the rock mass that is highly susceptible to stress-induced cracking or slippage instability). Nonlinear amplification occurs, significantly increasing the probability of structural surface development at that grid location. Conversely, if the grid is in a strong compressive stress shielded zone (where cracks are tightly compacted and cannot develop), the weight value approaches zero. Thus, this operator is no longer a purely random geometric guess, but a four-dimensional spatiotemporal tensor that can be dynamically updated as the rock mass's loading state changes. By running this operator, the system can quantitatively describe the spatial existence probability, attitude change probability, and spatial trend of the structural surface in the blind zone of the rock mass under future stress disturbances at each micro-grid unit. This provides underlying mechanical causal logic support for the next stage of accurate dynamic reverse measurement and twin model reconstruction of blind zone parameters using mechanical and physical signals in true triaxial acoustic emission indoor experiments.

[0045] Up to Sp3, the system has initially constructed a stress sensitivity probability density evolution operator for the hidden structural surface in the blind zone at the software algorithm layer through static geometric scanning, deep hole optical detection, and in-situ stress field injection in actual engineering sites. However, at this time, the operator model is still a product of geological static estimation and physical mechanics theory deduction. Whether the core structural surface parameters (such as the actual displacement and endogenous roughness coefficient) inside the blind zone have undergone activation changes still lacks verification by real physical living signals.

[0046] To eliminate the illusions of theoretical deductions and achieve accurate measurement and dynamic correction of parameters in deep discontinuous fields, this invention further integrates a true triaxial mechanical loading framework with a multi-channel fully dynamic acoustic emission monitoring system. In a highly simulated indoor environment, the rock load damage response is activated, and the elastic wave group signal generated by the fracture is used to perform inverse solution and boundary correction of the blind zone parameters.

[0047] Sp4: The purpose of this step is to reproduce the deep in-situ loading environment with high fidelity in the laboratory, and to activate and capture the activation response of hidden cracks in the unknown blind zone through physical means. The specific implementation process is as follows: Sp4.1: The system first calls the in-situ three-dimensional geostress tensor measured on the engineering site using a digital hollow inclusion stress gauge in Sp3.1. Based on the geological loading conditions of the actual excavation face, the original rock sample is prepared into a standard cubic rock mass sample with a side length of 100 mm using a precision double-end stone grinding machine. The sample is then placed in the true triaxial test subsystem, which is physically composed of a loading frame, a self-balancing pressure chamber, a hydraulic servo system, and a dynamic actuator.

[0048] The system activates the hydraulic servo system to drive the rigid pressure head. Based on the in-situ three-dimensional geostress tensor, it applies the initial first principal stress, second principal stress, and third principal stress to the three orthogonal directions of the rock sample (the maximum output load can reach 2,000 kN, 4,000 kN, and 120 MPa, respectively), placing it in a true triaxial high-stress state. While maintaining triaxial static load balance, the system's background controller officially activates the dynamic actuator, applying dynamic disturbance loads of specific frequency and amplitude (response frequency can reach 20 Hz) to the sample axially. This formally constructs a highly realistic true triaxial servo loading and unloading environment. Under this strong excavation disturbance dynamic field, the local shear stress at the hidden structural surface inside the sample breaks through the Mohr-Coulomb barrier, thereby successfully activating the hidden structural surface inside the target rock mass to undergo shear slip and micro-fracture damage, causing it to actively radiate endogenous elastic waves outward.

[0049] Sp4.2: During the entire operation of the above-mentioned true triaxial servo loading and unloading environment, when the hidden crack surface undergoes shear slip or micro-crack damage, it will instantly release the strain energy accumulated inside. The acoustic emission subsystem connected to the dynamic field analysis module is used for full life cycle monitoring. The multi-channel full dynamic acoustic emission sensor array, which contains at least eight high-pressure and high-temperature acoustic emission sensors arranged around the pressure head inside the self-balancing pressure chamber, senses and captures the elastic wave group signal generated by the structural surface displacement and micro-fracture during the loading and damage process of the sample in real time and with full waveform at an ultra-high sampling rate of megahertz. After filtering and amplitude modulation of the signal by the pre-amplifier, the signal is input into the multi-channel digital signal acquisition instrument for high-speed analog-to-digital conversion without loss. The system processor extracts and accurately stores the arrival time of the first wave of the array elastic wave captured by each independent channel in real time (i.e., the first jump axis position of the P wave), and simultaneously records the core waveform characteristics such as the maximum amplitude, rise time and duration of the full waveform of the entire cycle.

[0050] SP4.3: After obtaining the above-mentioned digitized full waveform and first-wave clock data, the system derives the core data source through the following specific mechanical analysis algorithm: The first step is to obtain the acoustic emission spatiotemporal point cluster data: the system extracts the arrival time of the first arrival wave from each channel. Because true triaxial loading causes strong anisotropy in the rock mass, the system operates a time-difference positioning algorithm with constant anisotropic velocity correction. Let the first... The known coordinates of the acoustic emission sensors within the pressure chamber are: The three-dimensional spatial coordinates of the micro-fracture source (acoustic emission fracture source) to be solved are: The rupture occurred at the time when And the nonlinear wave velocity in this direction is The specific steps for constructing the nonlinear equation system for time difference positioning are as follows: ; The processor uses the Gauss-Newton iterative method to solve the problem and calculate the precise spatial coordinates and time of each micro-fracture event. The system combines all the successfully located spatial points into a streaming matrix group according to the time flow, thereby calculating and constructing the acoustic emission spatiotemporal point cluster data in the main memory.

[0051] The second step involves the analytical derivation of the nonlinear acoustic emission spatiotemporal wave waveform tensor: For each successfully located event, the system extracts the maximum amplitude characteristics of the multi-channel waveform and runs a multi-level wave velocity anisotropic spatiotemporal solution algorithm to perform moment tensor inversion. Let the first... The amplitude of the first arrival displacement of the far-field P-wave received by each sensor is Its relationship with the second-order symmetric tensor of microfractures (It contains six independent components) The following elastic wave dynamics radiation equation formula is satisfied: ; In the formula, For rock density, For the corrected space wave velocity, Let be the Euclidean distance from the seismic source to the sensor. and The epicenter points to the first The direction cosines of each sensor are used to obtain the moment tensor through linear inversion of singular value decomposition (SVD). The spatial location, timestamp, maximum waveform amplitude, and corresponding moment tensor components of each rupture source are fused into a four-dimensional spatiotemporal flow matrix, thereby finally calculating and analyzing the nonlinear acoustic emission spatiotemporal flow waveform tensor.

[0052] The third step involves calculating the scalar field of crack energy dissipation: based on the inverted moment tensor components, the system calculates the scalar seismic moment energy released by a single microfracture by solving its principal eigenvalues. The system divides the sample into discrete three-dimensional meshes within a finite three-dimensional space, and distributes the energy released by all events to the corresponding spatial mesh cells through a Gaussian smoothing kernel function for accumulation and superposition. The specific calculation expression is as follows: ; In the formula, For the first The coordinates of the rupture source, To control the bandwidth parameter of the spatial smoothing range, the crack energy dissipation scalar field that is continuously distributed in space is finally calculated and analyzed accurately.

[0053] Sp5: The purpose of this step is to dynamically demodulate and finally stitch together the geometric and mechanical network of the blind zone using the dynamic physical signals obtained above. The specific implementation process is detailed below: SP5.1: Since the indoor test data is calculated based on the physical boundaries of the specimen, the system first initiates a transformation of the rigid spatial coordinate system without rotation or translation, completely mapping all event coordinates in the nonlinear acoustic emission spatiotemporal waveform tensor to the principal reference coordinate system (i.e., the Cartesian coordinate system of the excavation face). In the principal reference coordinate system, the system calls and runs a density-based noise spatial clustering algorithm (DBSCAN algorithm) for any point. Its governing equation is determined by constructing the Eps neighborhood, and the specific mathematical expression is as follows: ; In the formula, For the mapped set of acoustic emission points, For Euclidean distance, if (In this embodiment, the neighborhood radius is set) Minimum number of dots: 3 mm If the value is 5, then the decision point is... Using the core point as the center, the algorithm clusters and grows points with achievable density in the surrounding area. It automatically identifies and forcibly filters out scattered unstructured noise points with extremely low spatial density caused by random grain fracture. At the same time, it clusters strong aggregated point clouds generated by the overall slippage and cracking of the hidden structural surface in the internal blind zone. The system extracts the external nonlinear convex hull mesh of these continuously distributed clusters, thereby automatically identifying the macroscopic nonlinear envelope surface of the acoustic emission point cloud aggregation.

[0054] Sp5.2: To achieve quantitative inversion of deep microscopic mechanical parameters in the blind zone, this invention introduces a spaceflow damping boundary condition for acoustic emission elastic wave energy dissipation in the calculation. This condition sets the energy attenuation coefficient due to viscous damping when the elastic wave passes through the concealed structural surface as follows: Under this constraint, the system performs multivariate nonlinear inverse demodulation by calculating the joint distribution of the central fitting orientation of the macroscopic nonlinear envelope surface and the crack energy dissipation scalar field: The first step, the inverse process of evolutionary attitude, is to use the maximum likelihood estimation method to fit the above macroscopic nonlinear envelope surface to a spatial surface and extract the best-fit unit normal vector at the center point of the surface. The system directly calculates the evolutionary attitude that changes with the loading state based on the spatial projection geometric mapping equation, specifically including the evolutionary dip angle. and evolutionary tendency : ; ; The second step involves the inverse calculation of the intrinsic roughness coefficient: based on the joint distribution, the spatial three-dimensional geometric fluctuation variance of the point cloud inside the macroscopic nonlinear envelope surface is systematically calculated. Based on the theory of elastic wave damping attenuation, the intrinsic roughness coefficient (JRC) of the structural surface and the attenuation energy flux in the energy dissipation field are related. and volatility variance The following multi-nonlinear mapping control equations are satisfied: ; In the formula, and Based on the empirical constants pre-calibrated for the current hard rock lithology, the system solves the equation to reverse-calculate and reconstruct the endogenous roughness coefficient (JRC) of the blind zone structural surface.

[0055] The third step involves the inverse derivation of the true shear displacement parameters: the system demodulates the source mechanism of the core full waveform of the dense event cluster, and inversely derives the source time function. Using this function, based on the principles of shear fracture mechanics, the actual shear displacement parameters at the shear slip point can be determined. The second time integral of the source time function and the scalar seismic moment satisfy the following governing equation formula: ; In the formula, For rock shear modulus, The micro-fracture area is obtained by fitting the envelope surface. To determine the fracture duration, the system directly and accurately extracts the true shear displacement parameters by running this formula. The core structural parameters of the target rock mass were collected.

[0056] Please see Figure 4 As shown, it provides a core results display interface after the system dynamically inversely demodulates the parameters of the discontinuous surface in the unknown blind zone. This interface is set up with "post-evolution" state monitoring cards for each hidden fracture surface (such as "Blind Zone Structural Surface 1", "Blind Zone Structural Surface 2", and "Blind Zone Structural Surface 3") in the unknown blind zone after undergoing indoor true triaxial servo loading and unloading disturbance tests. Each card intuitively and accurately digitally outputs and displays the core mechanical morphology parameters obtained by inversely solving the spatiotemporal flow damping boundary conditions through acoustic emission elastic wave energy dissipation. Specifically, these parameters include evolution tilt angle (°), evolution tendency (°), roughness coefficient (JRC), and actual shear displacement (m). This interface is used to transparently and quantitatively present the evolution morphology inside the deep blind zone, which is originally impossible to observe directly with the naked eye. It also provides the most direct and physically constrained dynamic measured data source for the energy field boundary correction of the stress sensitivity probability density evolution operator and the Monte Carlo topological mesh growth. Sp5.3: The system will substitute the core parameters obtained from the experimental inversion solution in Sp5.2 into the stress sensitivity probability density evolution operator obtained from the purely theoretical derivation in Sp3. Energy field boundary correction is performed, and the underlying governing equation for the correction is to introduce an energy field correction factor. Composite mapping function: ; In the formula, The function takes the real acoustic emission energy dissipation scalar field as a physical hard constraint. If the theoretical probability is high but there is no acoustic emission energy release in the region ( ),but Approaching 0, the probability quickly returns to zero after correction. Therefore, the spatial existence probability, attitude change probability, and connectivity evolution trend of the unknown blind zone structural surface are updated and corrected spatially. Subsequently, within the target's rigid unknown blind zone grid, the system runs a Monte Carlo random sampling algorithm. For any candidate grid point coordinates in the grid, the system generates a... Uniformly distributed random numbers within the interval The probability of rejection is satisfied if and only if the following random rejection condition equation is met: ; The system only accepts and retains the grid point as the growth skeleton point of the structural surface; otherwise, it rejects it. Through thousands of iterations of sampling and growth, a continuous topological grid structure that can 100% match the energy dissipation characteristics of the elastic wave is finally generated within the blind zone.

[0057] Sp5.4: This step is the ultimate destination and data closure point of the technical chain. In the background computing space, the system performs overall spatial closure stitching and cross-scale rigid fusion of the above-generated blind zone continuous topological mesh structure with the exposed structural surfaces (planar mesh) obtained in Sp1 and the internal hidden structural surfaces (one-dimensional line segment intersecting mesh) obtained in Sp2. During stitching, the system constructs a global geometric boundary coordination control equation, assuming the mesh misalignment topological error on the stitching boundary is... The system optimizes energy by running the following least-squares functional energy equation: By solving the Lagrange multiplier matrix, the nodes at the stitching edge are finely aligned to eliminate geometric faults, and the exposed geometric plane, the topological line segment inside the hole, and the blind zone prediction mesh are completely fused into a topologically continuous, physically tightly closed discontinuous surface full-element three-dimensional assembly.

[0058] Ultimately, based on the core structural parameters collected throughout the entire process, the system successfully constructed a three-dimensional digital twin model of the rock mass structure in the main reference coordinate system. This model perfectly matches the actual mechanical response of deep engineering and integrates geometry and mechanics. It can realistically reproduce the discontinuous failure evolution mechanism of deep rock mass under loading, disturbance and excavation conditions, providing a highly fully disclosed and accurate scientific digital foundation for the stability evaluation of underground engineering.

[0059] Please see Figure 3 As shown, the system provides a main interface navigation diagram, which includes core function menus such as homepage dashboard, point cloud data acquisition, borehole image acquisition, structural surface analysis, stress field analysis, acoustic emission monitoring, digital twin, analysis report, and system settings. This interface is used to construct an interactive bus for the hardware control and multi-source data flow of the entire system. Through this main interface, operators can send static and dynamic detection and test control commands to the lower-level hardware with one click, realizing high-frequency synchronous interception and unified convergence storage of discrete data streams such as on-site 3D scanning point cloud, borehole TV cylindrical panoramic image, on-site geostress tensor, and indoor acoustic emission spatiotemporal flow waveform tensor, forming the digital control base of the entire rock mass structural surface parameter acquisition and analysis system.

[0060] For further details, please refer to Figure 2 As shown, the module configuration, data flow and operation mode, and interaction logic of the system hardware entities of this software system are as follows: The macroscopic geometry acquisition module is physically connected to the control bus of the 3D digital measurement scanning terminal, high-resolution industrial digital camera, and external mobile workstation deployed at the construction site. This module is used to acquire macroscopic geometric point cloud data of the target rock excavation face using the 3D digital measurement scanning terminal. The data flow and implementation path is as follows: The control console first issues optical measurement commands to drive the high-resolution industrial digital camera to perform multi-angle overlapping shooting of the target rock excavation face, such as the tunnel face, to acquire multi-view image pairs. The module's internal computing engine calls the built-in Structure for Motion Restoration (SfM) algorithm based on feature point pair matching to perform SIFT feature point extraction, RANSAC mismatch removal, and epipolar geometric constraint matrix (fundamental matrix). The algorithm calculates and generates a sparse point cloud. Then, it calls the PMVS dense stereo matching algorithm to generate high-density, multi-dimensional macroscopic geometric point cloud data in main memory. Next, the module performs discrete geometric topology analysis on the point cloud, establishing a [data structure] at each control point. Neighborhood 3D local covariance matrix Non-negative eigenvalues ​​are extracted through eigenvalue decomposition. Substituting the values ​​into the curvature solution equation, the local Gaussian curvature at each point is calculated. and mean curvature Thus, a curvature mutation tensor constraint is constructed. Finally, the module uses this tensor as the boundary criterion for anisotropy, runs a region growing algorithm for flatness threshold segmentation, and automatically divides the massive point cloud into independent sets of fitted planar patch points. It then calculates the standard unit normal vector of the spatial plane for each patch. The first attitude parameter (first dip angle) of the exposed structural surface is automatically recorded, calculated, and extracted. and first tendency value The trace length characteristic function of the negative exponential statistical distribution of the outcrop trace length This static multidimensional feature matrix is ​​then encapsulated and packaged, and sent to the system mainline data bus.

[0061] The borehole topology acquisition module engages in bidirectional data interaction with the hardware entity of the 4D drilling ultra-high-definition television system deployed inside horizontal or inclined boreholes. This system hardware includes a cable counting winch, a panoramic cylindrical imaging television probe, and an in-hole attitude sensor (a spatial three-axis electronic compass and a microelectromechanical system gyroscope) integrated within the probe. This module is used to acquire a 360° panoramic image of the borehole wall using the 4D drilling ultra-high-definition television system. The data flow and implementation path is as follows: when the panoramic cylindrical imaging probe is driven by the winch to advance at a constant speed within the blind borehole, this module intercepts and reads the video image stream in real time, while simultaneously and seamlessly recording the probe's spatial coordinates output at high frequency by the in-hole attitude sensor. With time deflection (azimuth) and tilt angle The module first runs a video stream keyframe extraction algorithm, filtering out redundant frames by calculating the chi-square distance of the grayscale histograms of adjacent frames. Next, it calls a cylindrical projection transformation stitching algorithm to calculate the unity matrix of adjacent image grids based on pose and depth coordinates. Multi-scale image fusion is used to eliminate spin and bending errors, stitching together a single, distorted 360° full cylindrical panoramic image of the hole wall. Subsequently, the module inputs this panoramic image into a pre-trained deep learning convolutional neural network based on the MobileNet-Unet architecture. The network extracts keyframes through cascaded convolutions and skip connections in the contraction and expansion paths. High-dimensional texture features are used to perform pixel-level binarization probability classification at the output. Through image semantic parsing and mapping, the internal hidden structural surface at the hole wall truncation point is automatically evolved on the topology—the 3D intersecting boundary line (i.e., the standard spatial sine curve trajectory line). Finally, the module accurately extracts the peak, trough, and phase difference parameters of the sine curve through mathematical extreme value search. Combined with the probe spatial coordinates and time axis deflection bound to its time axis, a spatial coordinate analytical transformation algorithm is run to inversely solve the 3D geometric analytical equation of the hidden plane in the principal reference coordinate system. The second attitude parameter (second dip angle) of the internal hidden structural surface is successfully analyzed and calculated. Compared with the second tendency value ).

[0062] The discontinuous field evolution module is a core component of the large-scale numerical calculation and probabilistic inference in the background. Its input end receives measurement data from digital hollow inclusion stress gauges at the construction site and is seamlessly connected to the data output ends of the macroscopic geometry acquisition module and the borehole topology acquisition module. The data flow and implementation path is as follows: This module first reads the strain curve of the entire process actually measured at the construction site using digital hollow inclusion stress gauges through the borehole unloading method. Based on the inverse solution of the spatial elastoplastic mechanics theory equations, it directly obtains the in-situ three-dimensional geostress tensor. The module uses a second-order symmetric matrix and simultaneously calls homogeneous spatial coordinate rotation and translation matrices to seamlessly map and linearly transform the previously extracted first attitude parameters (planar mesh) and second attitude parameters (one-dimensional intersection mesh) to a spatial principal reference coordinate system with the excavation face as the main body, thus initially constructing a spatial topology network. Next, the module uses exposed structural surfaces and internal concealed structural surfaces as known mechanical contact boundary sources, and calculates the in-situ three-dimensional geostress tensor. As the initial field of the infinitely far boundary load is injected into the spatial topology network, a discontinuous stress field mechanical model of the rock mass is constructed using the discrete mesh method. The computational engine iteratively solves the equilibrium equations that satisfy the boundary displacement discontinuity constraints. The normal stress distribution in any discrete mesh element within the unknown blind zone due to structural surface obstruction is calculated. and shear stress distribution This allows for the derivation and generation of discontinuous stress traces within the target rock mass in main memory; finally, the module establishes a stress sensitivity function based on the Mohr-Coulomb fracture criterion. The module uses the known geometric networks of exposed and concealed structural surfaces as planar constraints and line segment constraints, respectively, and runs a Bayesian probabilistic network estimation algorithm to calculate the geometric probability distribution function of the blind zone mesh under pure geometric topology evolution. Finally, the module uses the stress sensitivity function as a weighting factor to multiply by the geometric probability distribution function for nonlinear composite calculation, thereby constructing a stress sensitivity probability density evolution operator for hidden structural surfaces within the unknown blind zone. It is used to quantitatively describe the spatial existence probability, attitude change probability, and connectivity evolution trend of unknown blind zone structural surfaces under stress disturbance.

[0063] The dynamic field analysis module serves as a bridge connecting the system's internal algorithms with indoor physical experiments. This module connects the true triaxial testing subsystem and the acoustic emission subsystem. The true triaxial testing subsystem includes a loading frame, a self-balancing pressure chamber, a hydraulic servo system, and dynamic actuators. The acoustic emission subsystem includes a high-pressure, high-temperature resistant acoustic emission sensor array (at least eight channels) and a multi-channel digital signal acquisition instrument arranged inside the pressure chamber. The data flow and implementation path is as follows: During indoor experiments, the loading control bus calls the in-situ three-dimensional geostress tensor parameters sent from the discontinuous field evolution module to control the hydraulic servo system to drive the rigid pressure head, precisely applying initial pressure to the cubic specimen placed in the self-balancing pressure chamber. The first, second, and third principal stresses, while maintaining static load equilibrium, are used to apply dynamic disturbance loads of specific frequency and amplitude by the dynamic actuator activated by the control system. This artificially constructs a true triaxial servo unloading environment to strongly activate the hidden structural surfaces inside the sample, causing shear slip and micro-fracture damage. Throughout the entire loading and damage process, the hidden fracture surfaces radiate high-frequency elastic waves. A high-pressure, high-temperature acoustic emission sensor array arranged inside the pressure chamber captures these elastic wave signals in real time and with full waveform. After filtering and amplitude modulation by a preamplifier, the signals are losslessly input into a multi-channel digital signal acquisition instrument. The processor of this module intercepts the memory buffer of the acquisition instrument in real time, accurately recording the arrival time of the first arrival wave of each channel. The module then calculates the three-dimensional spatial coordinates and occurrence time of each acoustic emission fracture source by iteratively solving a set of nonlinear distance equations and obtaining the full waveform characteristic parameters. This results in an acoustic emission spatiotemporal point cluster dataset in main memory. Based on this dataset, the module extracts the waveform amplitude and, by solving the far-field shift amplitude equation, linearly inverts the second-order symmetric tensor of each micro-fracture. Finally, the processor runs a multi-level wave velocity anisotropic spatiotemporal demodulation algorithm to integrate the spatial points, generation clock, short tensor components, and energy dissipation of a single micro-fracture source calculated based on principal eigenvalues. By performing matrix confluence, the nonlinear acoustic emission spacetime flow waveform tensor and the crack energy dissipation scalar field continuously distributed in the three-dimensional mesh space are finally calculated analytically. .

[0064] The parameter output and twin module is the ultimate data aggregation and graphical rendering center for the entire system. Its input end maintains high-frequency synchronization with the data bus of the dynamic field analysis module and the discontinuous field evolution module. This module is used to perform inverse parameter demodulation using the nonlinear acoustic emission spatiotemporal flow waveform tensor and the crack energy dissipation scalar field and output a twin model. The data flow and implementation path is as follows: After receiving the nonlinear acoustic emission spatiotemporal flow waveform tensor, the module first performs a spatial rigid coordinate system transformation, completely mapping and linearly transforming the spatial coordinates of all elastic micro-fractures to the main reference coordinate system. Subsequently, the module... The module invokes and runs a density-based noise spatial clustering algorithm (DBSCAN algorithm) to construct a neighborhood Eps range determination equation. This automatically filters out spatially scattered unstructured tearing noise points, clusters strongly clustered slip events into clusters, and extracts their external nonlinear convex hull mesh. This automatically identifies the macroscopic nonlinear envelope surface of the acoustic emission point cloud cluster. Next, the module introduces the spatiotemporal flow damping boundary condition for acoustic emission elastic wave energy dissipation. By calculating the joint distribution of the central fitted attitude of the macroscopic nonlinear envelope surface and the crack energy dissipation scalar field, multivariate nonlinear inverse demodulation is performed. The envelope surface is fitted using the maximum likelihood estimation method, its central normal vector is extracted, and the evolutionary attitude (evolutionary dip angle) that changes with loading state is calculated and demodulated. and evolutionary tendency ); Statistical variance of spatial three-dimensional geometric fluctuation of point cloud within envelope surface With decaying energy flux Running the damped topography inversion empirical control equation The intrinsic roughness coefficient (JRC) of the blind zone structure surface is obtained by reverse reconstruction calculation. Extracting the source time function from the core full waveform A second-order time integration was performed to calculate and extract the actual shear displacement parameters during the crack shear instability process. This completes the acquisition of core structural parameters of the target rock mass and outputs them to the terminal device. Next, the module will output the acquired evolutionary attitude, JRC, and other parameters. Converted into real-time energy field correction factor The stress-sensitive probability density evolution operator obtained from static derivation Perform energy field boundary correction, and then perform composite update to generate the corrected probability field. This eliminates the probability illusion caused by the inhomogeneity of the material in the blind zone. Subsequently, the module uses a Monte Carlo random sampling algorithm within the blind zone grid, with the corrected probability number as the rejection criterion equation. The system performs multi-path evolutionary growth in space, generating a continuous topological mesh structure within the blind zone that perfectly matches the energy dissipation characteristics of the elastic wave. Finally, the module runs the global geometric boundary coordination control equations to perform cross-scale spatial closure stitching and boundary node fine-tuning alignment of the generated continuous topological mesh structure in the blind zone with the planar mesh of the exposed structural surface and the one-dimensional line segment mesh of the hidden structural surface inside the hole, eliminating geometric gaps. Based on the above full-element parameters, the system comprehensively and completely constructs a three-dimensional digital twin model of the rock mass structural surface on the graphics workstation, which fully matches the actual mechanical response of deep engineering and integrates geometry and mechanics. This provides a digital foundation with the highest accuracy, full openness, and entity information support for the stability evaluation, rock mass cracking evolution early warning, and support decision design of underground hard rock engineering.

[0065] In summary, the control logic of the overall scheme of a method and system for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning is summarized as follows: it constructs a closed-loop technical chain for parameter acquisition and analysis that integrates surface and core, dynamic and static coupling, and geometry and mechanics.

[0066] In the stage of static detection and operator initial construction, the system uses the point cloud data collected by the three-dimensional digital measurement and scanning terminal as the main spatial reference coordinate system. It constructs curvature mutation tensor constraints by calculating the eigenvalues ​​of the local covariance matrix, automatically segments and records the first attitude parameter and trace length characteristic function of the exposed structural surface. On this basis, the system injects the in-situ three-dimensional geostress tensor measured in the field and the sinusoidal curve trajectory line (second attitude parameter) of the internal hidden structural surface (semantic segmentation of 4D borehole television) as known boundary sources into the spatial topology network. It uses the discrete grid method to deduce and generate the discontinuous stress shading field, and finally constructs the stress sensitivity probability density evolution operator of the blind zone structural surface.

[0067] In the indoor dynamic test and inverse parameter demodulation stage, the system activates the shear slip of the hidden structural surfaces inside the rock mass under true triaxial servo loading and unloading and dynamic disturbance environment. It captures elastic wave group signals using a multi-channel full-dynamic acoustic emission sensor array, and analyzes the nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field through time difference positioning algorithm and moment tensor inversion. Then, the system introduces the elastic wave energy dissipation spatiotemporal flow damping boundary condition. By calculating the joint distribution of the macroscopic nonlinear envelope surface center fitting attitude and energy dissipation field identified by DBSCAN clustering, the system dynamically inversely demodulates the evolution attitude (including evolution dip angle and evolution tendency), endogenous roughness coefficient and real fault shear displacement parameters of the blind zone structural surface.

[0068] Finally, in the data closure and twin reconstruction stage, the system uses the core parameters acquired dynamically to perform energy field boundary correction on the aforementioned stress sensitivity probability density evolution operator, and runs the Monte Carlo random sampling algorithm to generate a continuous topological mesh structure matching this feature within the blind zone. Finally, through the least squares functional energy minimization equation, the topological mesh in the blind zone is spatially closed and stitched with the exposed and hidden structural surfaces across scales. Thus, a three-dimensional digital twin model of the rock mass structural surface that perfectly matches the deep mechanical response is successfully constructed and output in the main reference coordinate system, realizing high-precision, full-element accurate acquisition and analysis of parameters of high-stress discontinuous rock mass.

[0069] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0070] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning, characterized in that: The method includes the following steps: Sp1: Use a three-dimensional digital measurement and scanning terminal to acquire macroscopic geometric point cloud data of the target rock mass excavation surface, and use curvature change tensor constraints to perform anisotropic boundary mesh segmentation on the geometric point cloud data, and automatically compile and extract the first attitude parameters and trace length characteristic functions of the exposed structural surface. Sp2: A 360° panoramic unfolded image of the borehole wall was obtained using a 4D drilling ultra-high-definition television system. The one-dimensional intersecting boundary line of the internal hidden structural surface at the borehole wall truncation point was derived through image semantic analysis and mapping. The second attitude parameter of the internal hidden structural surface was obtained through analysis. Sp3: The in-situ three-dimensional geostress tensor is used as the initial field and injected into the spatial topology network constrained by the first and second attitude parameters. The discontinuous stress shielding field inside the target rock mass is generated by deduction, and then the stress sensitivity probability density evolution operator of the blind zone structural surface inside the rock mass is constructed. Sp4: Under the true triaxial servo loading and unloading environment, the shear slip and micro-fracture damage of the hidden structural plane inside the target rock mass are activated. The elastic wave group signal of the sample endogenous with the multi-channel full dynamic acoustic emission sensor array is captured in real time. The nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field are analyzed by the multi-level wave velocity anisotropic spatiotemporal demodulation algorithm. Sp5: By utilizing the nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field, the evolution of the blind zone structural surface under load, the endogenous roughness coefficient, and the actual displacement shear displacement parameters are dynamically and inversely demodulated. This completes the acquisition of core structural surface parameters of the target rock mass, and based on this, the energy field boundary is corrected for the stress sensitivity probability density evolution operator. Based on the acquired structural surface parameters, a three-dimensional digital twin model of the rock mass structural surface that matches the deep mechanical response is constructed.

2. The method for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning according to claim 1, characterized in that: Sp1 specifically includes the following sub-steps: Sp1.1: Multi-view image pairs of the target rock mass excavation face are acquired by a high-resolution industrial digital camera, and high-density multi-dimensional macro-geometric point cloud data is generated by a motion recovery structure algorithm based on feature point pair matching. Sp1.2: Calculate the local covariance matrix at each point in the geometric point cloud, extract eigenvalues ​​and calculate Gaussian curvature and mean curvature, and construct curvature mutation tensor constraints; Sp1.3: Based on curvature mutation tensor constraints, anisotropic boundary mesh segmentation is performed on geometric point cloud data. The region growing method is used to segment the geometric point cloud data into independent fitted flat patches, and then the normal vector of each patch is calculated. The first attitude parameter and trace length feature function of exposed structural surfaces are automatically recorded and extracted. The first attitude parameter specifically includes the first dip angle and the first dip value representing the exposed structural surface. The trace length feature function is obtained by statistically analyzing the trace length of the outcrop.

3. The method for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning according to claim 1, characterized in that: Sp2 specifically includes the following sub-steps: Sp2.1: Insert the 4D drilling ultra-high-definition television probe into a horizontal or inclined hole to acquire full-view video images, and record the current spatial coordinates and time axis deflection angle of the probe in real time through the hole attitude sensor; Sp2.2: Using video stream keyframe extraction and real-time image stitching algorithms, the full-view video images are unfolded and stitched into a single-frame full-cylindrical 2D panoramic image of the hole wall, resulting in a 360° full-cylindrical panoramic unfolded image of the hole wall. Sp2.3: Input the 360° full cylindrical panoramic image of the hole wall into the trained convolutional neural network for semantic segmentation. Through image semantic parsing, the one-dimensional intersecting boundary line of the internal hidden structure surface at the hole wall truncation point is derived. The one-dimensional intersecting boundary line is represented by a sinusoidal curve trajectory line representing the hidden structure surface. Sp2.4: Extract the peaks, troughs, and phase difference parameters of the sinusoidal trajectory line. Combine the probe's spatial coordinates and time axis deflection angle, and use a spatial coordinate analytical transformation algorithm to obtain the second attitude parameters of the internal hidden structural surface. The second attitude parameters specifically include the second dip angle and the second dip value of the internal hidden structural surface.

4. The method for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning according to claim 1, characterized in that: The stress sensitivity probability density evolution operator for constructing the blind zone structural surface inside the rock mass in Sp3 specifically includes the following sub-steps: Sp3.1: On-site stress measurement using digital hollow inclusion stress gauges is performed to directly obtain the in-situ three-dimensional geostress tensor. At the same time, the three-dimensional spatial coordinate system of the excavation face is used as the main reference coordinate system. The first and second attitude parameters are linearly mapped to the main reference coordinate system to construct a spatial topology network. Sp3.2: Exposed structural surfaces and internally concealed structural surfaces are used as boundary sources. The in-situ three-dimensional geostress tensor is injected into the spatial topology network as the initial field. The discrete grid method is used to construct a mechanical model of the discontinuous stress field of the rock mass. The normal stress distribution and shear stress distribution of any grid unit in the unknown blind zone due to the shading of structural surfaces are calculated. The discontinuous stress shading field inside the target rock mass is deduced and generated. Sp3.3: Establish a stress sensitivity function related to the deformation of structural surfaces and normal and shear stresses. Use the geometric parameters of exposed and concealed structural surfaces as planar and line segment constraints, respectively. Multiply the stress sensitivity function as a weighting factor by the geometric probability distribution function based on Bayesian probabilistic grid estimation. Construct a composite operator to generate the stress sensitivity probability density evolution operator of concealed structural surfaces in the unknown blind zone. Use this operator to quantitatively describe the spatial existence probability, attitude change probability, and connectivity evolution trend of structural surfaces in the blind zone of the rock mass under stress disturbance.

5. The method for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning according to claim 1, characterized in that: Sp4 specifically includes the following sub-steps: Sp4.1: The rock mass sample prepared according to the excavation face is placed in the self-balancing pressure chamber of the true triaxial mechanical loading frame. The initial first principal stress, second principal stress and third principal stress are set according to the in-situ three-dimensional geostress tensor, and dynamic actuator disturbance load is applied to construct a true triaxial servo loading and unloading environment to activate the shear slip and micro-fracture damage of the hidden structural surface inside the target rock mass. Sp4.2: Using a multi-channel, fully dynamic acoustic emission sensor array arranged around the pressure chamber, the elastic wave group signals generated by structural surface slippage and micro-fractures during the sample loading and damage process are captured in real time, and the arrival time and waveform characteristics of the first arrival wave of each channel are recorded. Sp4.3: The three-dimensional spatial coordinates of the acoustic emission fracture source are calculated based on the time difference positioning algorithm to form acoustic emission spatiotemporal point cluster data. The moment tensor of the micro fracture is inverted using the waveform amplitude. Then, the nonlinear acoustic emission spatiotemporal flow waveform tensor and the crack energy dissipation scalar field are analyzed by the multi-level wave velocity anisotropic spatiotemporal demodulation algorithm.

6. The method for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning according to claim 1, characterized in that: Sp5 specifically includes the following sub-steps: Sp5.1: The nonlinear acoustic emission space-time flow waveform tensor resolved in Sp4 is mapped to the main reference coordinate system. A density-based noise space clustering algorithm is used to filter out scattered unstructured noise points and identify the macroscopic nonlinear envelope surface of the acoustic emission point cloud cluster. Sp5.2: Introducing the space-flow damping boundary condition for acoustic emission elastic wave energy dissipation, by calculating the joint distribution of the central fitting attitude of the macroscopic nonlinear envelope surface of the acoustic emission point cloud and the crack energy dissipation scalar field, the evolution attitude, endogenous roughness coefficient and real displacement shear displacement parameters of the blind zone structural surface as the loading state change are dynamically and inversely demodulated, thus completing the acquisition of core structural surface parameters of the target rock mass. Among them, the evolution attitude specifically includes the evolution dip angle and evolution tendency. Sp5.3: Using the collected core structural surface parameters, the energy field boundary correction is performed on the stress sensitivity probability density evolution operator in Sp3 to update the spatial existence probability, attitude change probability and connectivity evolution trend in a linked manner. In the blind zone grid range, the Monte Carlo random sampling algorithm is used to generate a continuous topological grid structure that can match the energy dissipation characteristics of the elastic wave. Sp5.4: Spatially closes and stitches continuous topological mesh structures with exposed and concealed structural surfaces, and constructs a three-dimensional digital twin model of rock mass structural surfaces that matches the deep mechanical response based on the collected structural surface parameters.

7. The system corresponding to the method for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning according to claim 1, characterized in that: The system includes a macroscopic geometry acquisition module, a borehole topology acquisition module, a discontinuous field evolution module, a dynamic field analysis module, and a parameter output and twinning module, wherein: The macro-geometry acquisition module is used to acquire macro-geometry point cloud data of the target rock mass excavation surface using a three-dimensional digital measurement and scanning terminal. It performs anisotropic boundary mesh segmentation on the geometric point cloud data through curvature change tensor constraints, and automatically compiles and extracts the first attitude parameters and trace length characteristic functions of the exposed structural surfaces. The borehole topology acquisition module is used to acquire a 360° full cylindrical panoramic unfolded image of the borehole wall using a 4D drilling ultra-high-definition television system. Through image semantic parsing, it maps and evolves the one-dimensional intersecting boundary line of the internal hidden structural surface at the borehole wall truncation point, and obtains the second attitude parameter of the internal hidden structural surface through parsing. The discontinuity field evolution module is used to inject the in-situ three-dimensional geostress tensor as the initial field into the spatial topology network constrained by the first and second attitude parameters, and to deduce and generate the discontinuous stress shielding field inside the target rock mass, thereby constructing the stress sensitivity probability density evolution operator of the blind zone structural surface inside the rock mass. The dynamic field analysis module is used to activate the shear slip and micro-fracture damage of the hidden structural plane inside the target rock mass under the true triaxial servo loading and unloading environment. It uses a multi-channel full dynamic acoustic emission sensor array to capture the elastic wave group signal of the sample in real time, and uses a multi-level wave velocity anisotropic spatiotemporal demodulation algorithm to analyze the nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field. The parameter output and twin module is used to dynamically demodulate and output the evolution attitude, endogenous roughness coefficient and actual fault shear displacement parameters of the blind zone structural surface using the nonlinear acoustic emission spatiotemporal flow waveform tensor and crack energy dissipation scalar field. This completes the acquisition of core structural surface parameters of the target rock mass, and performs energy field boundary correction on the stress sensitivity probability density evolution operator accordingly. Based on the acquired structural surface parameters, a three-dimensional digital twin model of the rock mass structural surface that matches the deep mechanical response is constructed.

8. The method and system for acquiring and analyzing rock mass structural surface parameters based on three-dimensional scanning according to claim 7, characterized in that: The dynamic field analysis module is connected to a true triaxial test subsystem and an acoustic emission subsystem. The true triaxial test subsystem includes a loading frame, a self-balancing pressure chamber, a hydraulic servo system, and a dynamic actuator, which are used to apply triaxial independent and non-interfering high-stress loads and dynamic disturbance loads. The acoustic emission subsystem includes a high-pressure and high-temperature acoustic emission sensor array arranged inside the pressure chamber and a multi-channel digital signal acquisition instrument.

Images