Information-based monitoring system for tunnel construction based on 3D scanning point clouds

Abstract

This invention relates to the field of tunnel construction monitoring technology, specifically to an information-based monitoring system for tunnel construction based on three-dimensional scanning point clouds. The system includes: collecting and registering full-section point cloud data of the tunnel construction area to form a standardized point cloud dataset; dividing the tunnel space into three-dimensional units using voxel-based mesh segmentation technology; quantitatively calculating the surface deviation between the measured point cloud and the design model within each unit to achieve precise localization of deformation anomaly areas; verifying the anomaly areas using ground-penetrating radar data; and constructing a deformation evolution trajectory based on multi-time-series point cloud data; and using modal decomposition technology to separate elastic deformation caused by construction loads and plastic deformation caused by structural damage from the evolution trajectory. This invention improves the sensitivity and quantification level of local tunnel deformation monitoring and effectively identifies the physical causes of deformation, thus providing a more accurate and reliable basis for assessing the structural safety status during construction.

Description

Technical Field

[0001] This invention relates to the field of tunnel construction monitoring technology, and in particular to an information-based monitoring system for tunnel construction based on three-dimensional scanning point clouds. Background Technology

[0002] During tunnel construction, accurate and efficient monitoring of the deformation of the surrounding rock and support structure is crucial to ensuring construction safety and quality. Current technologies primarily rely on single-point contact measurements using total stations and convergence meters, or on using 3D laser scanning to acquire point clouds for overall deformation trend analysis. These methods have limitations in comprehensiveness, analytical depth, and automation. Conventional single-point measurements struggle to capture continuous deformation information across the entire tunnel cross-section, resulting in sparse and inefficient monitoring data. Existing point cloud data processing methods typically involve comparing measured point clouds with design models or performing simple distance calculations. This approach only yields macroscopic discrepancies and fails to systematically quantify localized, subtle, irregular deformation areas on the tunnel surface. Consequently, it lacks sufficient sensitivity to identify initial local instability or defects, making it difficult to meet the demands of dynamic and refined monitoring during construction.

[0003] In the structural condition assessment phase, most existing technologies rely on total deformation or rate of change for early warning, lacking in-depth analysis of the physical causes of deformation. Tunnel monitoring data contains a mixture of recoverable elastic deformation caused by temporary construction loads and irreversible plastic deformation resulting from the accumulation of structural damage. Traditional methods cannot effectively separate these two distinctly different deformation components, leading to insufficient evidence for structural safety diagnosis. Early warning thresholds often rely on experience, potentially resulting in false alarms or missed alarms due to the inability to distinguish the nature of deformation. This makes it difficult to accurately assess the true extent and development stage of internal structural damage, thus hindering precise guidance for construction decisions and maintenance interventions. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an information-based monitoring system for tunnel construction based on three-dimensional scanning point clouds.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a tunnel construction information monitoring system based on three-dimensional scanning point clouds, comprising:

[0006] The acquisition and processing module acquires full-section point cloud data of the tunnel construction area through a multi-view 3D laser scanning device, performs point cloud registration and noise filtering on the full-section point cloud data, and forms a standardized point cloud dataset.

[0007] The deformation analysis module performs spatial overlay analysis on the standardized point cloud dataset and the design model point cloud. It divides the tunnel space into three-dimensional units of equal size using voxelized mesh segmentation technology and calculates the surface deviation between the measured point cloud and the design model point cloud in each three-dimensional unit.

[0008] The anomaly identification and verification module identifies anomalous deviation areas based on the distribution characteristics of the surface deviation, and verifies the geological defect correlation of the anomalous deviation areas by combining ground-penetrating radar detection data.

[0009] The evolution and reconstruction module establishes a mapping relationship between the abnormal deviation areas and the construction progress time sequence, constructs the deformation evolution trajectory through multi-frame point cloud sequence analysis, and uses spatiotemporal interpolation methods to complete the deformation data of missing monitoring points;

[0010] The structural diagnosis module, based on the complete deformation evolution trajectory, uses modal decomposition technology to separate elastic deformation under construction loads from plastic deformation caused by structural damage, generating a tunnel structural condition diagnosis report.

[0011] As a further aspect of the present invention, the step of performing spatial overlay analysis on the standardized point cloud dataset and the design model point cloud, dividing the tunnel space into equal-sized three-dimensional units using voxelized mesh segmentation technology, and calculating the surface deviation between the measured point cloud and the design model point cloud within each three-dimensional unit includes:

[0012] The standardized point cloud dataset and the design model point cloud are registered using the iterative nearest point algorithm to establish a spatial correspondence under a unified coordinate system.

[0013] The tunnel space is divided into three-dimensional meshes using an octree structure to generate cubic voxel units with equal side lengths, and the density distribution characteristics of the measured point cloud within each voxel unit are statistically analyzed.

[0014] Calculate the set of shortest distances from the measured point cloud surface to the design model point cloud surface within each voxel unit, and take the root mean square value of the set of shortest distances as the surface deviation of the voxel unit.

[0015] Gaussian smoothing is applied to the surface deviation of adjacent voxel units to eliminate local fluctuations caused by measurement noise and form a continuous spatial deviation distribution map.

[0016] As a further aspect of the present invention, the step of calculating the set of shortest distances from the measured point cloud surface to the design model point cloud surface within each voxel unit, and taking the root mean square value of the set of shortest distances as the surface deviation of the voxel unit, includes:

[0017] The measured point cloud within each voxel unit is reconstructed using triangular meshes to generate a continuous surface model. The normal distance from each vertex of the continuous surface model to the surface of the point cloud of the design model is then calculated.

[0018] The kd-tree acceleration algorithm is used to quickly find the nearest neighbor of each vertex to the surface of the point cloud of the design model, and the minimum distance value of all vertices is recorded to form a distance set;

[0019] Distance values ​​exceeding three standard deviations in the distance set are removed, and the remaining valid distance values ​​are subjected to a normality test.

[0020] The root mean square value of the valid distance that has passed the test is calculated, and the resulting value is used as the final surface deviation of the voxel unit.

[0021] As a further aspect of the present invention, the step of identifying abnormal deviation areas based on the distribution characteristics of the surface deviation, and verifying the geological defect correlation of the abnormal deviation areas in conjunction with ground-penetrating radar detection data, includes:

[0022] A surface deviation threshold is set, and voxel units that exceed the threshold are marked as initial anomalous points. Region growth is then performed from the initial anomalous points to adjacent voxel units.

[0023] The region growth termination condition is determined based on the gradient change of surface deviation, and spatially continuous anomalous voxel units are merged to form an anomalous region set.

[0024] Spatial matching is performed between the three-dimensional coordinates of the set of anomalous regions and the results of ground-penetrating radar detection, and the overlap between each anomalous region and the geological anomaly is calculated.

[0025] Based on the degree of overlap, the abnormal areas are divided into geologically related anomalies and construction-induced anomalies, and an anomaly area classification map is established.

[0026] As a further aspect of the present invention, the step of spatially matching the three-dimensional coordinates of the set of anomalous regions with the ground-penetrating radar detection results, and calculating the overlap between each anomalous region and the geological anomaly, includes:

[0027] The ground-penetrating radar data is converted into a three-dimensional point cloud format and then registered to the tunnel's unified coordinate system through coordinate transformation.

[0028] Extract anomalous reflection signal regions from ground-penetrating radar point clouds to generate a three-dimensional boundary model of the geological anomaly.

[0029] Calculate the volume intersection of each anomalous region with the three-dimensional boundary model of the geological anomaly, and divide it by the total volume of the anomalous region to obtain the spatial overlap.

[0030] An overlap threshold is set, and when the overlap exceeds the threshold, it is determined that the abnormal area is related to the geological defect.

[0031] As a further aspect of the present invention, the step of establishing the mapping relationship between the deviation anomaly region and the construction progress time sequence, constructing the deformation evolution trajectory through multi-frame point cloud sequence analysis, and using spatiotemporal interpolation methods to complete the deformation data of missing monitoring points includes:

[0032] Record the construction progress marker when each deviation abnormal area first appears, and establish the time correspondence between abnormal areas and construction stages;

[0033] Extract point cloud data of the same abnormal area within a continuous monitoring period, and calculate its deformation vector and deformation rate;

[0034] The deformation data during the monitoring interval were reconstructed using cubic spline interpolation to generate a continuous deformation time series.

[0035] Principal component analysis was used to extract characteristic parameters of the deformation evolution trajectory, including deformation direction angle, deformation acceleration, and deformation fluctuation amplitude.

[0036] As a further aspect of the present invention, the method of reconstructing deformation data during the monitoring interval using cubic spline interpolation to generate a continuous deformation time series includes:

[0037] Construct a piecewise cubic polynomial function with monitoring time points as nodes and deformation variables as node values;

[0038] This ensures that adjacent polynomials have continuous first and second derivatives at the nodes, forming a smooth deformation curve;

[0039] Different interpolation step sizes are set according to the intensity of construction activities. Small step size interpolation is used during periods of intensive construction, and large step size interpolation is used during periods of respite.

[0040] The interpolated deformation time series is subjected to low-pass filtering to eliminate high-frequency measurement noise and preserve the true deformation trend.

[0041] As a further aspect of the present invention, the separation of elastic deformation and plastic deformation caused by structural damage under construction loads based on the complete deformation evolution trajectory and using modal decomposition technology includes:

[0042] Empirical mode decomposition was performed on the complete deformation evolution trajectory to obtain eigenmode functions at different time scales;

[0043] Analyze the correlation coefficients between each intrinsic mode function and the construction load time sequence to identify the elastic deformation components that are synchronized with construction activities;

[0044] The trend term in the residual component is extracted and combined with the material creep model to identify plastic deformation related to structural damage;

[0045] Calculate the energy ratio of elastic deformation to plastic deformation, and assess the contribution of construction load and structural damage to the overall deformation.

[0046] As a further aspect of the present invention, the analysis of the correlation coefficients between each intrinsic modal function and the construction load time sequence, and the identification of elastic deformation components synchronized with construction activities, includes:

[0047] Establish a construction load intensity time sequence curve and record the time distribution of the main construction procedures of tunneling and support.

[0048] Calculate the cross-correlation function between each intrinsic mode function and the construction load intensity time series, and determine the maximum cross-correlation coefficient;

[0049] The eigenmode functions with cross-correlation coefficients exceeding the threshold and phase differences less than the allowable value are selected as elastic deformation components.

[0050] The selected intrinsic mode functions are superimposed and reconstructed to obtain the time series of elastic deformation caused by construction load.

[0051] As a further aspect of the present invention, the step of acquiring full-section point cloud data of the tunnel construction area using a multi-view three-dimensional laser scanning device, performing point cloud registration and noise filtering on the full-section point cloud data to form a standardized point cloud dataset includes:

[0052] Multiple fixed scanning stations are deployed, and cloud data from each station is automatically stitched together using a target sphere to generate a complete tunnel point cloud.

[0053] A statistical outlier removal algorithm is used to filter out flying points and noise points in the point cloud, while preserving the true geometric features of the tunnel surface.

[0054] Point cloud density is reduced by using voxelization downsampling method, thereby reducing the amount of data while maintaining shape features;

[0055] The processed point cloud data is uniformly converted to the tunnel design coordinate system to generate a standardized point cloud dataset.

[0056] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0057] A voxel-based mesh segmentation technique is employed to discretize the three-dimensional space of the tunnel into uniformly sized solid units, and the surface deviation between the measured point cloud and the design model point cloud is calculated within each unit. This method transforms continuous, massive point cloud data into quantitative indicators within a regular mesh, realizing a shift from macroscopic qualitative to microscopic quantitative analysis of tunnel surface deformation. By independently calculating and statistically analyzing the deviation of each voxel unit, abnormal areas with deviations significantly higher than the background value can be systematically identified, even if these areas are small in size and exhibit weak deformation. This mesh-based analysis method improves the detection accuracy and identification efficiency of local deformation and potential defects, overcoming the insensitivity of traditional overall comparison methods to local details, and providing high-resolution deformation distribution maps for refined management during the construction phase.

[0058] Modal decomposition (MDD) technology was used to analyze the complete deformation evolution trajectory, separating the elastic deformation component under construction loads from the plastic deformation component caused by structural damage. Based on the fundamental principles of structural dynamic response, this technique uses an algorithm to decompose the monitored total displacement time series into modes with different physical meanings. The elastic deformation component reflects the normal response of the structure under load, and its magnitude is related to load changes; the plastic deformation component reveals the cumulative process of irreversible damage such as material yielding and crack propagation. This separation allows the assessment of structural condition to go beyond simple displacement exceedance judgments, enabling a more accurate identification of the timing, location, and severity of damage. This provides a direct and mechanistic scientific basis for assessing structural safety margins and developing differentiated repair and reinforcement strategies. Attached Figure Description

[0059] Figure 1 This is a timing diagram of the tunnel construction information monitoring system based on three-dimensional scanning point cloud as described in this invention;

[0060] Figure 2 This is a flowchart for spatial overlay analysis and surface deviation calculation.

[0061] Figure 3 A flowchart for identifying and verifying abnormal deviation regions;

[0062] Figure 4 Deformation analysis diagram for tunnel construction monitoring;

[0063] Figure 5 This is a comparative diagram showing the deformation evolution trends of multiple abnormal areas during tunnel construction. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0065] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0066] See Figure 1 The acquisition and processing module uses a multi-view 3D laser scanning device to acquire full-section point cloud data of the tunnel construction area, and performs point cloud registration and noise filtering on this data to form a standardized point cloud dataset. Subsequently, the deformation analysis module performs spatial overlay analysis on this standardized point cloud dataset and the tunnel design model point cloud. It uses voxelized mesh segmentation technology to divide the tunnel space into equal-sized three-dimensional units and calculates the surface deviation between the measured point cloud and the design model point cloud within each three-dimensional unit. The anomaly identification and verification module identifies anomalous deviation areas based on the distribution characteristics of the surface deviation and verifies the correlation of these anomalous areas with geological defects using ground-penetrating radar data. The evolution and reconstruction module establishes a mapping relationship between anomalous deviation areas and the construction progress timeline. It constructs a deformation evolution trajectory by analyzing multi-frame point cloud sequences and uses spatiotemporal interpolation methods to complete the deformation data of missing monitoring points. Finally, the structural diagnosis module, based on the complete deformation evolution trajectory, uses modal decomposition technology to separate the elastic deformation caused by construction loads from the plastic deformation caused by structural damage, thereby generating a tunnel structural condition diagnosis report.

[0067] See Figure 2 In one embodiment of the present invention, taking the monitoring of a circular tunnel cross-section as an example, the design model point cloud provides the theoretical three-dimensional coordinates of the inner surface of the tunnel lining, while the standardized point cloud dataset obtained through three-dimensional laser scanning contains the actual three-dimensional coordinates of the lining surface after construction. The iterative nearest-point algorithm is used to register the two point clouds. The algorithm iteratively searches for the optimal rigid body transformation between point pairs, minimizing the sum of squared average distances between corresponding points, thereby establishing a precise spatial correspondence between the measured points and the design points in a unified coordinate system. In specific implementation, the registration process constrains the large degree of freedom variation in the tunnel axis direction, mainly optimizing the translation and rotation parameters around the axis to adapt to the directionality of tunnel excavation.

[0068] In practical implementation, after point cloud registration, the tunnel space is divided into three-dimensional meshes using an octree structure. An octree structure is a spatial partitioning data structure that recursively divides the entire tunnel space into eight sub-cubes until the side length of each sub-cube reaches a preset voxel size. In some embodiments, the side length of a voxel unit can be set to 5 cm according to the monitoring accuracy requirements. After partitioning, the system counts the number of measured point clouds falling within each cubic voxel unit and their spatial distribution, forming the density distribution characteristics of the measured point clouds. In practical implementation, the set of shortest distances from the surface of the measured point cloud within each voxel unit to the surface of the design model point cloud is calculated, and the root mean square value of this shortest distance set is taken as the surface deviation of this voxel unit. Optionally, during calculation, the measured point cloud within each voxel unit is first reconstructed using a triangular mesh. The Delaunay triangulation algorithm is used to generate a continuous surface model, and then the normal distance from the vertex of each triangular facet on the continuous surface model to the surface of the design model point cloud is calculated.

[0069] In the specific implementation, a kd-tree acceleration algorithm is used to quickly find the nearest neighbor of each vertex to the surface of the design model point cloud. A kd-tree is a data structure for multidimensional spatial retrieval; it organizes the design model point cloud into a binary tree, significantly accelerating the nearest neighbor search process. The system records the minimum distance value of all vertices, forming an initial distance set. It is understood that due to point cloud noise or the presence of isolated points, the initial distance set may contain outliers. In the specific implementation, distance values ​​exceeding three standard deviations in the distance set need to be removed. Then, the remaining valid distance values ​​are subjected to a Shapiro-Wilke test to verify whether they conform to a normal distribution. The valid distance values ​​that pass the normal distribution test are used to calculate the final surface deviation of the voxel unit. In some embodiments, the surface deviation... The calculation formula is as follows:

[0070]

[0071] in: This represents the i-th valid normal distance value that passed the test. Represents the total number of valid distance values. This refers to the calculated surface deviation. In practice, after the initial calculation of the surface deviation for all voxel elements, the surface deviations of adjacent voxel elements are Gaussian smoothed. Gaussian smoothing uses a two-dimensional Gaussian convolution kernel to convolve the spatial deviation matrix, thereby eliminating local numerical fluctuations caused by single-point measurement noise or small registration errors. The result is a continuous, smooth spatial deviation distribution map, which visually reflects the construction deviations of various points on the tunnel lining surface relative to the design model. Optionally, the standard deviation parameter of the Gaussian convolution kernel can be adjusted according to the voxel size and the expected noise level.

[0072] See Figure 3 In one embodiment of the present invention, taking a local lining heave scene discovered during the construction of a subway tunnel as an example, based on the spatial deviation distribution map, the system sets a surface deviation threshold, for example, marking voxel units exceeding ±50 mm as initial anomaly points. In specific implementation, with each initial anomaly point as the center, the algorithm performs region growth to its six spatially adjacent voxel units. It can be understood that during the region growth process, it will determine whether the surface deviation of adjacent voxel units also exceeds the threshold or whether its gradient change is less than the set termination gradient. After merging all spatially continuous abnormal voxel units, an abnormal region set is formed. For example, an ellipsoidal abnormal region set consisting of more than one hundred voxel units is identified at the tunnel arch position.

[0073] In practical implementation, the three-dimensional coordinates of the set of anomalous regions are spatially matched with the results of ground-penetrating radar (GPR) detection. In some embodiments, GPR data is acquired in the form of a series of two-dimensional profile images, which are converted into a three-dimensional point cloud format through inversion and three-dimensional reconstruction algorithms. This GPR point cloud data is then registered to a construction coordinate system unified with the tunnel point cloud through common control point coordinate transformation. In practical implementation, regions with signal amplitudes significantly higher than the background value are extracted from the GPR point cloud. These regions correspond to interfaces with abrupt changes in dielectric constant, such as cavities or water-bearing fissures. Based on this, the system generates a three-dimensional boundary model of the geological anomaly, typically represented by a triangular mesh surface. Optionally, for the identified set of ellipsoidal anomalous regions, the system calculates the volume intersection between its three-dimensional space and the space enclosed by the three-dimensional boundary model of the geological anomaly.

[0074] It is understandable that calculating the spatial overlap between each anomalous region and the geological anomaly is the core of verifying the correlation. In specific implementation, the system uses voxelization methods or polyhedral Boolean operations in computational geometry to calculate the volume intersection and spatial overlap between the set of anomalous regions and the three-dimensional boundary model of the geological anomaly. The calculation formula is as follows:

[0075]

[0076] in: The numerical value representing the volume intersection of the set of anomalous regions and the three-dimensional boundary model of the geological anomaly. A value representing the total volume of the set of anomalous regions. This refers to the calculated spatial overlap. In some embodiments, an overlap threshold is set, such as 60%. When the spatial overlap of a set of anomalous regions exceeds 60%, the anomalous region is determined to have a strong correlation with geological defects and is classified as a geologically related anomaly. When the spatial overlap is less than 60%, the deformation is considered to be mainly caused by construction techniques or temporary loads and is classified as a construction-induced anomaly. In specific implementation, the system performs the above calculation and judgment on all identified sets of anomalous regions, and finally generates a tunnel anomalous region classification map that marks the spatial distribution of different types of anomalies. In the map, geologically related anomaly regions can be rendered in red, and construction-induced anomaly regions can be rendered in yellow. Optionally, the calculation of volume intersection needs to consider the difference in accuracy between the two spatial models. If necessary, the three-dimensional boundary model of the geological anomaly body should be appropriately dilated to match its detection uncertainty.

[0077] In one embodiment of the present invention, taking an identified tunnel arch settlement anomaly area as an example, the system records the construction progress marker corresponding to the anomaly area when it is first monitored. The construction progress marker can be ring number information, such as "the 3rd day after the completion of the lining pouring of the 150th ring". The system establishes a time correspondence between the anomaly area and the specific construction stage. In specific implementation, the system extracts the point cloud data of the anomaly area in the subsequent ten consecutive monitoring cycles, with each monitoring cycle being one day. The system calculates the change in the three-dimensional coordinates of the centroid of the point cloud of the anomaly area relative to the design model in each cycle, forming a deformation vector sequence. The system then calculates the deformation rate sequence by dividing the difference in deformation vectors between adjacent cycles by the time interval.

[0078] In practical implementation, since monitoring is not continuous and there are data acquisition intervals, cubic spline interpolation is used to reconstruct the deformation data during the monitoring intervals. It is understood that cubic spline interpolation ensures the smoothness of the curve. In some embodiments, a piecewise cubic polynomial function is constructed using the daily monitoring time point as a node and the observed subsidence of the abnormal area at that time as the node value. The system ensures that adjacent cubic polynomials have continuous first and second derivatives at common nodes, thus forming a smooth deformation curve describing the change of arch settlement over time. Optionally, different interpolation step sizes are set according to the intensity of construction activities. During the three days before and after the tunneling machine passes through the section, when construction activities are intensive, a small step size of half an hour is used for interpolation; during the intermittent periods without construction activities, a large step size of two hours is used for interpolation. In practice, the interpolated continuous deformation time series is subjected to low-pass filtering. A Butterworth low-pass filter with a cutoff frequency of 0.1 Hz is used to eliminate high-frequency measurement noise and retain the true deformation trend in days.

[0079] In practical implementation, principal component analysis (PCA) is used to extract feature parameters of the deformation evolution trajectory. The system treats the point cloud dataset of the abnormal area acquired over ten monitoring cycles as a multidimensional time series and performs PCA on the series. PCA can be understood as extracting the main patterns of data change. In some embodiments, the principal direction of the deformation evolution trajectory in three-dimensional space is determined by the first principal component vector, and the deformation direction angle is the angle between the projection of this vector onto the horizontal plane and the tunnel axis direction. Deformation acceleration is obtained by calculating the second difference of the deformation rate time series. Deformation fluctuation amplitude... The calculation formula is as follows:

[0080]

[0081] in: This represents the eigenvalue of the first principal component obtained in principal component analysis. Represents the eigenvalues ​​of the second principal component. Represents the eigenvalues ​​of the third principal component. These are the characteristic parameters that characterize the total fluctuation amplitude of deformation in three-dimensional space. Optionally, principal component analysis is performed based on the coordinate change matrix of all points in the point cloud of the anomaly region within a continuous monitoring period, and the extracted characteristic parameters are used to quantitatively describe the geometric and kinematic properties of the deformation evolution trajectory.

[0082] In one embodiment of the present invention, taking a complete deformation time series of an abnormal settlement area in a tunnel arch as an example, this series comes from the interpolation and filtering results of the embodiment and contains continuous deformation data for more than 30 days. Empirical Mode Decomposition (EMD) is performed on the complete deformation evolution trajectory. EMD is an adaptive signal processing method that decomposes a complex signal into a finite number of intrinsic mode functions (EMFs) and a residual component. In specific implementation, the deformation time series is input into the EMD algorithm. The algorithm extracts EMFs at different time scales through an iterative filtering process, for example, decomposing it into five EMFs and a monotonic residual component. Each EMF represents an oscillation mode at a specific frequency. The trend term in the residual component is extracted, and combined with a material creep model, plastic deformation related to structural damage is identified. The residual component exhibits a slowly changing trend line. This trend term is fitted with a Burgers model based on the creep law of concrete. It can be understood that the material creep model describes the characteristics of strain growth over time under constant stress. When there is a systematic deviation between the observed trend and the trend predicted by the creep model, the deviation is identified as plastic deformation caused by structural damage. The energy ratio of elastic deformation to plastic deformation is calculated to assess the contribution of construction load and structural damage to the overall deformation. The energy of elastic deformation is calculated by summing the squares of the elastic deformation time series, and the energy of plastic deformation is calculated by summing the squares of the plastic deformation time series.

[0083] In some embodiments, the correlation coefficients between each intrinsic mode function and the construction load time series are analyzed to identify the elastic deformation component, and a construction load intensity time series curve is established. This curve is constructed by recording the time distribution of the main construction procedures of tunneling and support and quantifying the daily construction activity intensity. It can be understood that the construction activity intensity is comprehensively assigned based on factors such as machinery operation time and concrete pouring volume. The cross-correlation function between each intrinsic mode function and the construction load intensity time series is calculated. This cross-correlation function measures the similarity between two time series at different lag times. Intrinsic mode functions with cross-correlation coefficients exceeding a threshold and phase differences less than an allowable value are selected as elastic deformation components. For example, if the cross-correlation coefficient threshold is set to 0.7 and the allowable phase difference is two hours, intrinsic mode functions IMF2 and IMF3 meet the conditions. The selected intrinsic mode functions are then superimposed and reconstructed to obtain the elastic deformation time series caused by the construction load. Table 1 shows the cross-correlation analysis results for the five intrinsic mode functions.

[0084] Table 1: Results of Cross-Correlation Analysis between Intrinsic Modal Functions and Construction Load Intensity

[0085]

[0086] In practical implementation, energy ratio The calculation formula is as follows:

[0087]

[0088] in: A numerical value representing the energy of elastic deformation. A numerical value representing the energy of plastic deformation. This is the energy ratio of elastic deformation to plastic deformation. In some embodiments, the energy ratio is... A value greater than 1 indicates that elastic deformation caused by construction loads is dominant, and the energy ratio is... A value less than 1 indicates that the plastic deformation caused by structural damage is significant. Optionally, energy calculation can be based on the variance of the deformation sequence to eliminate the influence of the DC component. Optionally, the allowable phase difference value can be adjusted according to the duration of the construction process. In practice, a tunnel structural condition diagnosis report is finally generated, which presents the elastic deformation component, plastic deformation component, and energy ratio as key indicators.

[0089] See Figure 4This is a deformation analysis chart from tunnel construction monitoring, showing the variation of crown settlement and deformation rate over a 10-day monitoring period. The settlement rate increases rapidly from days 1 to 4, peaking at approximately 7.3 mm on day 4, then gradually decreases and stabilizes, remaining stable at around 3.9 mm by day 10. A peak of approximately 2.6 mm / day is reached on day 3; while this value does not exceed the abnormal threshold, it is close to the danger warning line. A negative rate appears on day 6, indicating a brief rebound in settlement, after which the rate gradually decreases to near zero, indicating that settlement has entered a stable phase. The rapid settlement and high rate in the first four days correspond to the intensive construction period, representing the stage dominated by elastic deformation caused by construction loads. The fluctuations in deformation rate verify the necessity of cubic spline interpolation and low-pass filtering, providing a practical basis for the subsequent reconstruction of smooth deformation curves.

[0090] In one embodiment of the present invention, taking a tunnel construction section with a length of 200 meters as an example, in order to obtain complete full-section point cloud data, multiple fixed scanning stations need to be set up in the tunnel. In some embodiments, a scanning station is set up every 40 meters along the centerline of the tunnel, for a total of 6 scanning stations. Each scanning station scans the surrounding 360-degree horizontal field of view and 270-degree vertical field of view using a 3D laser scanner. In specific implementation, the scanning range of adjacent scanning stations has a 30% overlap area. At least three target spheres are arranged within the visible range of each scanning station. The target spheres are spherical markers with highly reflective surfaces. It can be understood that the target spheres appear as obvious spherical point clusters in the point cloud. By identifying the common target sphere point clusters in the point clouds of different stations, automatic stitching is performed using their center coordinates. The algorithm calculates the optimal rotation and translation matrix to transform the point clouds of each station to a unified temporary coordinate system, thereby generating a complete tunnel point cloud covering the 200-meter section.

[0091] In practical implementation, a statistical outlier removal algorithm is used to filter out flypoints and noise points in the point cloud. This algorithm analyzes the average distance between each point and its K nearest neighbors, assuming the average distance follows a Gaussian distribution. In practice, points whose mean distance exceeds three times the standard deviation are identified as outliers and removed. This operation effectively filters flypoints caused by dust, water mist, or temporary obstacles, while preserving the true geometric features of the tunnel lining surface, such as the texture and joint details of the concrete surface. In some embodiments, after noise filtering, the point cloud density is reduced using a voxelization downsampling method. Voxelization downsampling divides the three-dimensional space into a uniform cubic grid, and within each grid, the centroid or center point of all points represents the points within that grid. The downsampling rate is... The calculation formula is as follows:

[0092]

[0093] in: This represents the number of points in the original filtered point cloud. This represents the number of points in the point cloud after downsampling. This refers to the downsampling rate. In practice, the voxel grid side length is set to 5 mm. While maintaining the macroscopic and microscopic features of the tunnel cross-section, such as shape and contour, the point cloud data volume is reduced from billions of points to hundreds of millions of points. Optionally, the downsampling rate is usually set based on the accuracy requirements of subsequent analysis and computational resources. It can be understood that uniformly converting the processed point cloud data to the tunnel design coordinate system is the final step in generating a standardized point cloud dataset. The tunnel design coordinate system typically uses the starting point of the line as the origin and the tangent direction of the line as the X-axis. In practice, using at least three control points with known precise coordinates in the design coordinate system, a seven-parameter transformation model of the point cloud from the temporary coordinate system to the design coordinate system is calculated. This model includes three translation parameters, three rotation parameters, and one scale parameter. After converting all point cloud data to the tunnel design coordinate system, a standardized point cloud dataset with a unified coordinate reference, denoised, and optimized data volume is generated, providing input for subsequent deformation analysis. Optionally, the coordinate transformation residuals need to be checked to ensure that the overall accuracy of splicing and transformation meets monitoring requirements.

[0094] See Figure 5 This is a comparative chart of deformation evolution trends in multiple abnormal areas during tunnel construction. It visually presents the deformation changes of three abnormal areas over a 28-day monitoring period. Through multi-area comparison, risks can be directly classified: Abnormal area 1 is high-risk, requiring immediate structural damage investigation; abnormal area 2 is medium-risk, requiring optimized construction procedures; and abnormal area 3 is low-risk, allowing for routine monitoring. The deformation characteristics of different areas help distinguish between "elastic deformation caused by construction loads" and "plastic deformation caused by structural damage," providing intuitive phenomenological support for subsequent empirical mode decomposition (EMD). This chart can guide construction teams to prioritize monitoring and reinforcement resources in high-risk areas, avoiding resource waste and improving construction efficiency and safety. Long-term deformation data from multiple areas can be used to establish a spatial deformation evolution model of the tunnel structure, providing crucial information for operational maintenance.

[0095] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A tunnel construction information monitoring system based on three-dimensional scanning point clouds, characterized in that, include: The acquisition and processing module acquires full-section point cloud data of the tunnel construction area through a multi-view 3D laser scanning device, performs point cloud registration and noise filtering on the full-section point cloud data, and forms a standardized point cloud dataset. The deformation analysis module performs spatial overlay analysis on the standardized point cloud dataset and the design model point cloud. It divides the tunnel space into three-dimensional units of equal size using voxelized mesh segmentation technology and calculates the surface deviation between the measured point cloud and the design model point cloud in each three-dimensional unit. The anomaly identification and verification module identifies anomalous deviation areas based on the distribution characteristics of the surface deviation, and verifies the geological defect correlation of the anomalous deviation areas by combining ground-penetrating radar detection data. The evolution and reconstruction module establishes a mapping relationship between the abnormal deviation areas and the construction progress time sequence, constructs the deformation evolution trajectory through multi-frame point cloud sequence analysis, and uses spatiotemporal interpolation methods to complete the deformation data of missing monitoring points; The structural diagnosis module, based on the complete deformation evolution trajectory, uses modal decomposition technology to separate elastic deformation under construction loads from plastic deformation caused by structural damage, and generates a tunnel structural condition diagnosis report. The method of separating elastic deformation and plastic deformation caused by structural damage under construction loads based on the complete deformation evolution trajectory and using modal decomposition technology includes: Empirical mode decomposition (EMD) is performed on the complete deformation evolution trajectory to obtain intrinsic mode functions (EMFs) at different time scales. The correlation coefficients between each EMF and the construction load time series are analyzed to identify elastic deformation components that are synchronous with construction activities. Trend terms are extracted from the residual components, and plastic deformation related to structural damage is identified by combining the material creep model. The energy ratio of elastic deformation to plastic deformation is calculated to assess the contribution of construction load and structural damage to the overall deformation. The analysis of the correlation coefficients between each intrinsic modal function and the construction load time sequence, and the identification of elastic deformation components synchronized with construction activities, includes: Establish a construction load intensity time series curve and record the time distribution of the main construction procedures of tunneling and support; calculate the cross-correlation function between each intrinsic mode function and the construction load intensity time series, and determine the maximum cross-correlation coefficient; select intrinsic mode functions whose cross-correlation coefficient exceeds the threshold and whose phase difference is less than the allowable value as elastic deformation components; superimpose and reconstruct the selected intrinsic mode functions to obtain the elastic deformation time series caused by construction load.

2. The tunnel construction information monitoring system based on three-dimensional scanning point cloud as described in claim 1, characterized in that, The step involves spatially overlaying the standardized point cloud dataset with the design model point cloud, dividing the tunnel space into equal-sized three-dimensional units using voxelized mesh segmentation technology, and calculating the surface deviation between the measured point cloud and the design model point cloud within each three-dimensional unit. This includes: The standardized point cloud dataset and the design model point cloud are registered using the iterative nearest point algorithm to establish a spatial correspondence under a unified coordinate system. The tunnel space is divided into three-dimensional meshes using an octree structure to generate cubic voxel units with equal side lengths, and the density distribution characteristics of the measured point cloud within each voxel unit are statistically analyzed. Calculate the set of shortest distances from the measured point cloud surface to the design model point cloud surface within each voxel unit, and take the root mean square value of the set of shortest distances as the surface deviation of the voxel unit. Gaussian smoothing is applied to the surface deviation of adjacent voxel units to eliminate local fluctuations caused by measurement noise and form a continuous spatial deviation distribution map.

3. The tunnel construction information monitoring system based on three-dimensional scanning point cloud as described in claim 2, characterized in that, The calculation of the set of shortest distances from the measured point cloud surface to the design model point cloud surface within each voxel unit, and the taking of the root mean square value of the set of shortest distances as the surface deviation of the voxel unit, includes: The measured point cloud within each voxel unit is reconstructed using triangular meshes to generate a continuous surface model. The normal distance from each vertex of the continuous surface model to the surface of the point cloud of the design model is then calculated. The kd-tree acceleration algorithm is used to quickly find the nearest neighbor of each vertex to the surface of the point cloud of the design model, and the minimum distance value of all vertices is recorded to form a distance set; Distance values ​​exceeding three standard deviations in the distance set are removed, and the remaining valid distance values ​​are subjected to a normality test. The root mean square value of the valid distance that has passed the test is calculated, and the resulting value is used as the final surface deviation of the voxel unit.

4. The tunnel construction information monitoring system based on three-dimensional scanning point cloud as described in claim 1, characterized in that, The step of identifying anomalous deviation areas based on the distribution characteristics of the surface deviation, and verifying the correlation of geological defects in the anomalous deviation areas using ground-penetrating radar detection data, includes: A surface deviation threshold is set, and voxel units that exceed the threshold are marked as initial anomalous points. Region growth is then performed from the initial anomalous points to adjacent voxel units. The region growth termination condition is determined based on the gradient change of surface deviation, and spatially continuous anomalous voxel units are merged to form an anomalous region set. Spatial matching is performed between the three-dimensional coordinates of the set of anomalous regions and the results of ground-penetrating radar detection, and the overlap between each anomalous region and the geological anomaly is calculated. Based on the degree of overlap, the abnormal areas are divided into geologically related anomalies and construction-induced anomalies, and an anomaly area classification map is established.

5. The tunnel construction information monitoring system based on three-dimensional scanning point cloud as described in claim 4, characterized in that, The step of spatially matching the three-dimensional coordinates of the set of anomalous regions with the results of ground-penetrating radar detection, and calculating the overlap between each anomalous region and the geological anomaly, includes: The ground-penetrating radar data is converted into a three-dimensional point cloud format and then registered to the tunnel's unified coordinate system through coordinate transformation. Extract anomalous reflection signal regions from ground-penetrating radar point clouds to generate a three-dimensional boundary model of the geological anomaly. Calculate the volume intersection of each anomalous region with the three-dimensional boundary model of the geological anomaly, and divide it by the total volume of the anomalous region to obtain the spatial overlap. An overlap threshold is set, and when the overlap exceeds the threshold, it is determined that the abnormal area is related to the geological defect.

6. The tunnel construction information monitoring system based on three-dimensional scanning point cloud as described in claim 1, characterized in that, The process of establishing a mapping relationship between the abnormal deviation areas and the construction progress timeline, constructing deformation evolution trajectories through multi-frame point cloud sequence analysis, and using spatiotemporal interpolation methods to complete the deformation data of missing monitoring points includes: Record the construction progress marker when each deviation abnormal area first appears, and establish the time correspondence between abnormal areas and construction stages; Extract point cloud data of the same abnormal area within a continuous monitoring period and calculate its deformation vector and deformation rate; The deformation data during the monitoring interval were reconstructed using cubic spline interpolation to generate a continuous deformation time series. Principal component analysis was used to extract characteristic parameters of the deformation evolution trajectory, including deformation direction angle, deformation acceleration, and deformation fluctuation amplitude.

7. The tunnel construction information monitoring system based on three-dimensional scanning point cloud as described in claim 6, characterized in that, The method of reconstructing deformation data during the monitoring interval using cubic spline interpolation to generate a continuous deformation time series includes: Construct a piecewise cubic polynomial function with monitoring time points as nodes and deformation variables as node values; This ensures that adjacent polynomials have continuous first and second derivatives at the nodes, forming a smooth deformation curve; Different interpolation step sizes are set according to the intensity of construction activities. Small step size interpolation is used during periods of intensive construction, and large step size interpolation is used during periods of respite. The interpolated deformation time series is subjected to low-pass filtering to eliminate high-frequency measurement noise and preserve the true deformation trend.

8. The tunnel construction information monitoring system based on three-dimensional scanning point cloud as described in claim 1, characterized in that, The process involves acquiring full-section point cloud data of the tunnel construction area using a multi-view 3D laser scanning device, performing point cloud registration and noise filtering on the full-section point cloud data to form a standardized point cloud dataset, including: Multiple fixed scanning stations are deployed, and cloud data from each station is automatically stitched together using a target sphere to generate a complete tunnel point cloud. A statistical outlier removal algorithm is used to filter out flying points and noise points in the point cloud, while preserving the true geometric features of the tunnel surface. Point cloud density is reduced by using voxelization downsampling method, thereby reducing the amount of data while maintaining shape features; The processed point cloud data is uniformly converted to the tunnel design coordinate system to generate a standardized point cloud dataset.

Images