Surrounding rock over-excavation and under-excavation detection method and monitoring and early warning system for curved tunnel

Abstract

The invention relates to a surrounding rock over-break and under-break detection method and a monitoring and early warning system for a curved tunnel, and belongs to the technical field of tunnel engineering. The method comprises the following steps: firstly, based on supervision of a bounding rectangle and a design file, quickly extracting a median line of a tunnel; the median line is used as a framework of the tunnel form and can accurately reflect the bending degree and direction of the tunnel; then, the meso-line infinitesimal vertical line cluster technology is utilized to realize refined extraction of the tunnel section, and it is ensured that high-precision section data can be obtained even in a complex bent tunnel; on the basis, the quasi-Monte-Carlo method is introduced to calculate the back-break volume, the quasi-Monte-Carlo method serves as an efficient numerical integration method, the volume of a complex geometric body can be approximately solved through statistical analysis of random samples, and the method is particularly suitable for back-break calculation of irregular bodies such as tunnels and has the advantages of being high in calculation efficiency, high in calculation accuracy and high in calculation precision. According to the method, the calculation precision is improved, the calculation time is remarkably shortened, and powerful support is provided for instant decision making of a construction site.

Description

Technical Field

[0001] The invention belongs to the technical field of tunnel engineering, and relates to a surrounding rock over-excavation and under-excavation detection method and a monitoring and early warning system for a curved tunnel. Background Art

[0002] With the vigorous advancement of national infrastructure construction, highway engineering, as a link connecting various places, has become increasingly important. In this process, rivers are no longer obstacles, mountains have also been conquered, and tunnel engineering, with its unique charm, shuttles between mountains and becomes an indispensable part of the modern transportation network. With the surge in the number of tunnels, the construction technology and technical level have also continued to rise. However, how to effectively control the over-excavation and under-excavation problems in the tunnel construction process while ensuring construction safety has become a focus of widespread attention within and outside the industry.

[0003] In short, tunnel over-excavation and under-excavation means that there is a deviation between the actual excavation interface and the design requirements. This deviation is not only related to the quality of the project, but also directly related to the long-term stability and safety of the tunnel structure. Over-excavation will increase the amount of concrete lining and increase construction costs; while under-excavation may lead to insufficient tunnel clearance, affect traffic efficiency, and even cause serious safety hazards. Especially when the tunnel passes through complex geological conditions, such as weak surrounding rock, karst development areas, etc., the problem of over-excavation and under-excavation is particularly prominent, which may induce the instability of the tunnel structure, such as collapse, collapse and other catastrophic consequences, posing a serious threat to the natural environment around the tunnel, residents' lives and even life safety.

[0004] Traditionally, the detection of tunnel over-break and under-break relies on manual measurement, that is, after each blasting operation, technicians use precision measuring equipment such as laser rangefinders and total stations to measure the tunnel sections one by one, and calculate the specific values ​​of over-break and under-break by comparing with the design drawings. However, this method has many limitations, especially when dealing with curved tunnels. Due to the complexity of the tunnel direction, traditional measurement methods are difficult to accurately and efficiently capture the true shape of the tunnel, resulting in a significant reduction in the accuracy of over-break and under-break assessment.

[0005] In view of the above problems existing in the prior art, the present invention proposes an innovative tunnel over-break and under-break detection method, which is particularly suitable for accurate measurement and evaluation of curved tunnels. Summary of the invention

[0006] In view of this, the purpose of the present invention is to provide a method for detecting over-break and under-break of surrounding rock of curved tunnels and a monitoring and early warning system. The method firstly extracts the median line of the tunnel quickly based on the supervision of the outer rectangle and the design file. The median line, as the skeleton of the tunnel shape, can accurately reflect the curvature and direction of the tunnel; then, the technology of the median line microelement vertical line cluster is used to realize the refined extraction of the tunnel section. This step ensures that high-precision section data can be obtained even in complex curved tunnels; on this basis, the method introduces the quasi-Monto Carlo method to calculate the over-break volume. As an efficient numerical integration method, the quasi-Monto Carlo method can approximate the volume of complex geometric shapes through statistical analysis of random samples. It is particularly suitable for the over-break calculation of irregular shapes such as tunnels. This method not only improves the calculation accuracy, but also significantly shortens the calculation time, providing strong support for immediate decision-making at the construction site.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] A method for detecting over-excavation and under-excavation of surrounding rock in a curved tunnel, the method specifically comprising the following steps:

[0009] S1: After tunnel blasting, a robot is used to enter the tunnel and extract the excavation point cloud data of the inner wall surface through 3D laser scanning;

[0010] S2: Obtain the tunnel section based on the extracted point cloud data;

[0011] S3: Determine the over-break and under-break volume of the tunnel according to the tunnel section;

[0012] S4: Design monitoring thresholds based on the designed excavation section profile and conduct over-excavation and under-excavation monitoring of the tunnel.

[0013] Further, in step S1, it specifically includes:

[0014] S11. First, the excavation point cloud data of the inner wall surface is extracted by using robot 3D scanning, including arranging target balls in the curved tunnel;

[0015] S12. After the target sphere is arranged, the coordinates of the control points of each tunnel pile are measured by using a total station, so as to obtain the coordinates of the crosshair midpoint of the target sphere, and the segment point cloud is obtained by using a robot 3D laser scanner, from which the spherical target is extracted and its center is fitted by the principle of rear intersection;

[0016] S13. Output point cloud data under unified coordinates.

[0017] Further, in step S12, it includes:

[0018] 1) The resection principle is to measure the angle or distance from the unknown point to several known points (usually three or more), combine the coordinates of these known points, and calculate the coordinates of the unknown point through geometric or mathematical methods; the known point is the target sphere, which requires at least two (usually three or more) control points with known coordinates. The positions of these points are called known points;

[0019] 2) The robot performs a 360° scan in a direction perpendicular to the line direction or at a 45-degree angle to the line direction, extracts the spherical target and fits its center;

[0020] 3) The three-dimensional coordinates of each target sphere center point measured by the total station using back intersection (in the total station coordinate system) are called the reference point set P = {P1, P2, ..., P n}, the target sphere center coordinates extracted from the point cloud, these points are called the point set to be aligned Q = {q1,q2,...,q n}; According to the relative position of the target ball, establish the corresponding relationship between the center point of the ball in the point cloud and the center point of the ball measured by the total station to ensure that p i and q i One to one correspondence;

[0021] 4) Rigid body transformation model, that is, transforming the point set Q of the point cloud coordinate system to the total station coordinate system P, the formula is, p i =R×q i +T; R represents the rotation matrix (3*3), T represents the translation vector (3*1);

[0022] 5) Solve the rigid body change and use the least squares method to solve it. According to the definition of the objective function: Solve for the optimal R and T to minimize the error E, and use Horn to quickly solve the rigid body transformation parameters;

[0023] 6) Apply rigid body transformation and use the calculated R and T to transform the target sphere center coordinate q in the point cloud data i Transform to total station coordinate system: Verify and optimize the calculated results and calculate the error If the error is large, the iterative closest point (ICP) algorithm is used to further align the point cloud with the reference points.

[0024] Further, in step S2, it specifically includes:

[0025] S21. Obtain at least 3 pairs of target sphere center coordinates and obtain the absolute point cloud through inter-station coordinate conversion: Calculate the transformation parameters from the local coordinate system of the point cloud to the global coordinate system through at least 3 pairs of target sphere center coordinates, including the rotation matrix R and the translation vector T: p absolute =R*P local+T, convert point cloud data from the local coordinate system to the absolute coordinate system;

[0026] S22. Obtain the tunnel median line according to the processed point cloud data: extract the tunnel median line by projecting the point cloud on the XOY plane, extracting the outer rectangle of the point cloud on the XOY plane to determine the axial direction of the tunnel, calculating the centerline point in combination with the vault, and jointly designing the centerline to encrypt the tunnel centerline, and finally obtaining the tunnel median line;

[0027] S23. Based on the median line extracted from the tunnel, the cross section of the vertical line cluster of the tunnel microelement is extracted.

[0028] Further, in step S22, the maximum and minimum coordinates of the tunnel point cloud in the XOY plane are X max , X min , Y max , Y min , it is possible to determine an enclosing rectangle, and determine the direction of the tunnel and the approximate centerline position according to the direction of the major axis of the rectangle; the arch vertex, that is, the highest point in the tunnel, is used as the arch vertex in the cross-section point cloud; the arch vertex and the center point of the bottom of the cross-section (or the midpoint of the two side points) are connected to calculate the geometric center of the cross-section as the centerline point; the centerline of the tunnel is encrypted by combining the designed centerline, and the deviation between the centerline point of the point cloud and the designed centerline point is compared to evaluate the degree to which the actual tunnel deviates from the design. The centerline point of the point cloud is combined with the designed centerline point using the weighted fitting method to generate an encrypted and optimized tunnel centerline.

[0029] Further, in step S23, it includes:

[0030] 1) Using three-dimensional coordinates P i (x i ,y i ,z i ), the construction of the centerline microelement and the vertical line cluster includes: microelement tangent vector calculation, and sequential calculation of the tangent vector for the centerline data points (indicating the direction of the tunnel centerline at that point): T i Indicates that the midline is at P i The tangent vector at P i-1 ,P i+1 Indicates the points before and after the midline;

[0031] 2) At each midline point P i , construct the normal plane, the normal plane is perpendicular to the tangent vector T i , with an optional reference direction U, the normal vector is calculated by cross product: N i =T i *U, forming three directions of the plane: tangent vector T i , normal vector N i , and its orthogonal complement vector B i =Ti ×N i ; Vertical line cluster generation: At each point P i On the normal plane, P i Generate a vertical line cluster as the center, use radius R as the threshold, and limit the range of point cloud points on the normal plane;

[0032] 3) Filter the point cloud and generate cross-section points, including: calculation of the distance from the point cloud point to the normal plane, that is, the calculation formula is: d = N i *(Q j -P i ), and at the same time, it is necessary to screen the point cloud points Qj that satisfy |d|<∈; the point cloud is projected onto the plane, and the screened point cloud points Qj are projected onto the normal plane to calculate their two-dimensional coordinates. The calculation formula is: Q' j ={(B i *(Q j -P i ),N i *(Q i -P i ))}, then we get the two-dimensional point set of the tunnel section;

[0033] 4) Fitting the cross-sectional curve, using the least squares method to fit the point cloud data, that is, Thus, the geometric parameters of the section are obtained; the tunnel section is generated, the fitting curve on the normal plane is converted into a tunnel section geometric model, and the section points are interpolated, the sampling points are encrypted, and the smoothness of the section is ensured; the extracted section is compared with the designed section, and the deviation is calculated. If there is a deviation, it can be refitted, and then the filtering method is used to remove local anomalies to ensure the continuity of the section; the result is output, and all micro-element sections are integrated into a complete micro-element vertical line cluster of the section point cloud; according to the extracted section point cloud, the point cloud is non-uniformly down-sampled and optimized;

[0034] 5) Calculate the local density of the point cloud of the section using the following formula: Ni is the number of neighborhood points of point Pi within radius r, and Vi is the neighborhood search volume. The point cloud data is divided into sparse and dense areas according to the calculated threshold, and the basis for the division is: ρ i >ρ thresh For dense areas, the opposite is sparse areas; for the point cloud in dense areas, the voxel grid downsampling method is used to optimize, and the voxel grid side length d is set dense , usually 5-10 mm, only one point is retained in this area; a smaller voxel side length d is used for sparse areas sparse , usually 2-5mm, the steps are the same as those for dense areas;

[0035] 6) In the key geometric feature areas including the vault and the wall base, retain the high curvature point cloud, according to the curvature C of the point i , the screening point meets Ci >C thresh point, and Where λ1, λ2, λ3 are the eigenvalues ​​of the covariance matrix; the processed point cloud data are recombined to form an optimized cross-sectional point cloud D' = {P'1, P'2, ..., P' n}; Project the optimized cross-section point cloud and perform noise reduction;

[0036] 7) According to the tunnel design file or the extracted centerline, take the tunnel centerline point as the origin, project the point cloud vertically to the reference plane, and calculate the projection point P'(x', y', z') of the point P(x, y, z) on its reference plane; assuming the reference plane is a plane, the projection formula is: N is the normal vector of the reference plane, d is the offset of the reference plane, and the error is checked and corrected after projection. Let the distance from point P to the reference plane be d. If d>∈, ∈ is the error threshold, it is marked as an abnormal point, and the abnormal point is smoothly adjusted to ensure that it is projected within the range of the reference plane;

[0037] 8) Apply SOR filtering technology to the projected point cloud data to remove noise points, calculate the number of points k in the neighborhood of each point, identify dense areas and sparse areas, and if the density difference is too large, it is necessary to fill in the sparse area or downsample the dense area. The number of neighborhood points k needs to be adjusted according to the density of the cross-section point cloud; for each point, calculate the average distance from its neighboring points to the point, and remove outliers whose distance mean exceeds the standard deviation multiple threshold. The standard deviation multiple σ is usually set to 1-2; use a smoothing algorithm to remove local irregular fluctuations; compare the projected point cloud after denoising with the original point cloud to confirm that the feature points are not over-smoothed, and finally save the projected and denoised point cloud data in a standard format for subsequent analysis or modeling.

[0038] Further, in step S3, the simplified point cloud is obtained, and the over-excavation and under-excavation volumes are calculated using the quasi-Monte Carlo method, which specifically includes:

[0039] S31, constructing a minimum construction minimum circumscribed cube of the point cloud to provide a sampling space range;

[0040] S32. The random body generated in the bounding volume generates coordinates through uniformly distributed random numbers, and its coordinate range must satisfy: x min ≤x≤x max ,y min ≤y≤y max ,z min ≤z≤z max ; Based on the distance detection from point to point cloud surface, determine whether it falls within the cross section described by the point cloud;

[0041] S33, calculate the cross-sectional contour volume, the calculation formula is: Vpcd is the volume of the model to be tested, V aabb is the volume of the axisymmetric bounding box, Np all Np is the total number of generated sampling points. in is the number of sampling points in the model;

[0042] S34. Based on the cross-sectional contour volume obtained by the quasi-Monte Carlo method, the over-break and under-break volume after tunnel blasting is calculated using the sampling method. The calculation formula is: Where: V over is the tunnel over-excavation volume, V under is the undercut volume of the tunnel, V aabb is the volume of the bounding box, P over is the number of sampling points in the over-excavation area, P under is the number of sampling points in the under-excavation area, P all is the number of all sampling points.

[0043] Further, in step S4, it specifically includes:

[0044] S41. According to the Technical Specifications for Highway Tunnel Construction and tunnel construction conditions, the outer and inner envelopes of the design section are set to form a safety range;

[0045] S42, the outer and inner envelopes of the cross section are the designed cross section plus the allowable over-break range and the designed cross section minus the allowable under-break range, respectively;

[0046] S43. According to the risk level within the safety range, set a percentage and divide the warning level. Including: normal range (green): between the inner and outer lines, no adjustment required; minor warning (yellow): close to the outer or inner line, need attention; serious warning (red): exceeding the outer or inner line, need immediate rectification.

[0047] The invention also provides a surrounding rock over-excavation and under-excavation monitoring and early warning system for a curved tunnel.

[0048] The beneficial effects of the present invention are:

[0049] The method provided by the present invention firstly quickly extracts the median line of the tunnel based on the supervision of the outsourcing rectangle and the design document. The median line, as the skeleton of the tunnel shape, can accurately reflect the curvature and direction of the tunnel; then, the technology of the median line microelement vertical line cluster is used to realize the refined extraction of the tunnel section. This step ensures that even in complex curved tunnels, high-precision cross-sectional data can be obtained; on this basis, the method introduces the quasi-Monto Carlo method to calculate the over-excavation volume. As an efficient numerical integration method, the quasi-Monto Carlo method can approximate the volume of complex geometric shapes through statistical analysis of random samples. It is particularly suitable for the over-excavation and under-excavation calculation of irregular shapes such as tunnels. This method not only improves the calculation accuracy, but also significantly shortens the calculation time, providing strong support for immediate decision-making at the construction site.

[0050] In addition, in order to further improve the safety of tunnel construction, the present invention also proposes to install an over-excavation and under-excavation monitoring and early warning system inside the tunnel. The system can monitor the over-excavation and under-excavation during tunnel excavation in real time. Once an abnormality is found, an early warning signal will be issued immediately, providing construction workers with valuable escape time and also giving them the first opportunity to take remedial measures, thereby effectively reducing the safety risks during tunnel construction and ensuring the quality of the project and the safety of personnel.

[0051] In summary, the present invention not only solves the difficult problem of over-excavation and under-excavation measurement of curved tunnels through technological innovation, but also greatly improves the safety and efficiency of tunnel construction through the introduction of intelligent monitoring and early warning systems, making an important contribution to promoting the advancement of my country's tunnel engineering technology.

[0052] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0054] Figure 1 A flow chart of a tunnel over-break and under-break detection calculation method according to the present invention;

[0055] Figure 2 It is a flow chart of the tunnel section non-uniform downsampling method of the present invention;

[0056] Figure 3 A flow chart of obtaining a cross section based on tunnel point cloud data according to the present invention;

[0057] Figure 4A flow chart for the cross section of the microelement vertical line cluster based on the median line of the present invention;

[0058] Figure 5 The present invention is a flow chart of the method for calculating the over-break and under-break volumes of a tunnel. DETAILED DESCRIPTION

[0059] The technical solution of the present invention is described in detail below with reference to the accompanying drawings.

[0060] The present invention provides a method for detecting over-excavation and under-excavation of surrounding rock of a curved tunnel and a monitoring and early warning system. The method can extract a median line proposed for a curved tunnel, obtain the over-excavation and under-excavation section based on the extraction of the median line, and calculate its volume based on the quasi-Monte Carlo method. At the same time, monitoring and early warning are set for over-excavation and under-excavation, and surrounding rock damage is evaluated, so as to continuously optimize the construction plan and reduce over-excavation and under-excavation of the tunnel.

[0061] like Figure 1 As shown, the method for detecting over-excavation and under-excavation of surrounding rock of curved tunnels and the monitoring and early warning system include four steps: 1. Robot scanning to obtain point cloud data; 2. Obtaining tunnel section; 3. Determining over-excavation and under-excavation volume; 4. Over-excavation and under-excavation early warning monitoring. Specifically:

[0062] 1. Robot scanning to obtain point cloud data:

[0063] The target ball is arranged in the curved tunnel. After the target ball is arranged, the coordinates of the tunnel pile control points are measured by the total station to obtain the coordinates of the crosshair midpoint of the target ball. The robot is used to perform a 360° three-dimensional laser scanner in a direction perpendicular to the line direction or at a 45-degree angle to the line direction to obtain the section point cloud. The three-dimensional coordinates of the center point of each target ball measured by the total station using the back intersection are the reference point set P = {P1, P2, ..., P n}, the target sphere center coordinates extracted from the point cloud are the point set to be aligned Q = {q1, q2, ..., q n}, according to the relative position of the target ball, establish the corresponding relationship between the center point of the ball in the point cloud and the center point of the ball measured by the total station to ensure that p i and q i One-to-one correspondence; then the point set Q of the point cloud coordinate system is converted to the total station coordinate system P, the formula is, p i =R×q i +T.

[0064] According to the definition of objective function: Solve the optimal R and T to minimize the error E, use Horn to quickly solve the rigid body transformation parameters, use the calculated R and T values, and convert the target ball center coordinate q in the point cloud data into i Transform to total station coordinate system: Then calculate the error The mean and standard deviation of the result are used to verify and optimize the result. If the error is large, the iterative closest point (ICP) algorithm can be used to further align the point cloud with the reference point, and finally output the point cloud data under the unified coordinates, and then the rotation matrix R and the translation vector T: p absolute =R*P local +T, thereby converting the point cloud data from the local coordinate system to the absolute coordinate system, which is convenient for the subsequent splicing of the tunnel model.

[0065] 2. Get the tunnel section:

[0066] like Figure 2 , 3 As shown in ,4, the main steps include: (1) extracting the median line; (2) constructing the cross section of the microelement vertical line cluster; (3) performing non-uniform downsampling optimization on the cross section point cloud; (4) projection and noise reduction processing;

[0067] (1) Extract the median line:

[0068] Project the point cloud on the XOY plane, and use the maximum and minimum coordinates of the point cloud on the XOY plane as X max , X min , Y max , Y min , we can determine an enclosing rectangle, determine the tunnel direction and approximate centerline position according to the direction of the long axis of the rectangle, take the highest point in the point cloud as the arch vertex, then connect the arch vertex and the center point of the bottom of the section (or the midpoint of the two side points), calculate the geometric center of the section as the centerline point; then compare the center point of the point cloud with the design center point, use the weighted fitting method to combine the point cloud centerline point with the design centerline point, thereby generating an encrypted and optimized tunnel centerline.

[0069] (2) Construct the cross section of the microelement vertical line cluster:

[0070] Use P i (x i ,y i ,z i ) represents the tunnel centerline, and the centerline data points obtained above are calculated according to the formula Calculate the tangent vector (indicating the direction of the tunnel centerline at that point) and calculate it at each centerline point P i , construct the normal plane, the specific steps are: the normal plane is perpendicular to the tangent vector T i , with an optional reference direction U, the normal vector is calculated by cross product: N i =T i *U, forming three directions of the plane: tangent vector T i , normal vector N i , and its orthogonal complement vector B i =T i ×N i, and finally at each point P i On the normal plane, a cluster of vertical lines is generated with Pi as the center; according to the calculation formula: d = N i *(Q j -P i ), select the point cloud points Qj that satisfy |d|<∈, project these points onto the normal plane, and calculate the point cloud points Qj according to Q' j ={(B i *(Q j -P i ),N i *(Q i -P i ))} calculate its two-dimensional coordinates, obtain the two-dimensional point set of the tunnel section, and then use The cross-section curve is fitted to obtain the geometric parameters of the cross-section; the fitted cross-section curve is converted into a tunnel cross-section geometric model, and the cross-section points are interpolated and the sampling points are encrypted to ensure the smoothness of the cross-section. At the same time, the extracted cross-section is compared with the designed cross-section to calculate the deviation. If there is a deviation, it can be refitted, and then the filtering method is used to remove local anomalies to ensure the continuity of the cross-section. Finally, all micro-element cross-sections are integrated into a complete micro-element vertical line cluster to obtain a cross-section point cloud.

[0071] (3) The specific steps of non-uniform downsampling optimization of cross-section point cloud are as follows: Figure 2 As shown, it includes: 1) identifying the non-uniform distribution of point cloud and partitioning analysis; 2) using the voxel grid non-uniform downsampling method; 3) curvature detection; 4) outputting the cross-section point cloud;

[0072] 1) Identify the non-uniform distribution of point clouds and perform partition analysis:

[0073] In the tunnel cross-section point cloud, the tunnel wall surface area usually has dense point clouds, while the vault, bottom and other areas may have sparse point clouds. Calculate the local density of the point cloud of the section according to ρ i >ρ thresh It is divided into dense area, and vice versa, it is divided into sparse area.

[0074] 2) Using voxel grid non-uniform downsampling method

[0075] For the point cloud in dense areas, the voxel grid downsampling method is used for optimization, and the voxel grid side length d is set dense , usually 5-10 mm, only one point is retained in this area, and a smaller voxel side length d is used in sparse areas sparse , usually 2-5mm, and only one point is left in the area.

[0076] 3) Curvature detection:

[0077] In key geometric feature areas such as vaults and wall bases, high curvature point clouds are retained. The selected C i >C thresh points; then reassemble the points in different areas to form an optimized cross-section point cloud D' = {P'1', P'2, ..., P' n}.

[0078] 4) Output cross-section point cloud:

[0079] The integrity of the optimized point cloud is verified, and the cross-sectional point cloud is finally output.

[0080] (4) Projection and noise reduction processing

[0081] According to the tunnel design file or the extracted center line, determine the projection reference plane, take the tunnel center line as the origin, project the point cloud vertically to the reference plane, and use the projection formula Calculate the projection point P'(x', y', z') of point P(x, y, z) on its reference plane, and then check the projection point. Set the distance from point P to the reference plane to d. If d>∈(error threshold), it is marked as an abnormal point. Smooth the abnormal point to ensure that it is projected within the range of the reference plane. Then, for each point in the neighborhood of the point, the number of points k is used to identify dense and sparse areas. If the density difference is too large, it is necessary to fill in the sparse area or downsample the dense area. Then use the SOR filtering technology to remove noise points. First, k is set between 10 and 50, and the standard deviation multiple σ is usually Set it to 1~2, calculate the average distance from each point to its neighboring points, and then remove outliers whose distance mean exceeds the standard deviation multiple threshold; secondly, use bilateral filtering to retain the point cloud features while removing small noise, set the point distance and grayscale weight, control the smoothing intensity, and ensure that the key features of the cross-section are not over-smoothed during the denoising process; finally, compare the projected point cloud after denoising with the original point cloud, calculate the curvature distribution of the point cloud as above, check whether there are abnormal fluctuations, confirm that the feature points are not over-smoothed, and then save the projected and denoised point cloud data in a standard format.

[0082] 3. Determine the over-excavation and under-excavation volume:

[0083] Figure 5 The flowchart of the method for calculating the over-break and under-break volume of a tunnel of the present invention is as follows: first, the minimum construction minimum circumscribed cube of the point cloud is constructed to provide a sampling space range, and then the random volume generated in the bounding volume generates coordinates through uniformly distributed random numbers, and the coordinate range must satisfy: x min ≤x≤x max ,y min ≤y≤y max ,z min ≤z≤z max, and then use the distance detection method based on the point-to-point cloud surface to determine whether it falls into the cross section described by the point cloud: that is, for a given random point P (x, y, z) to all points in the point cloud, and take the minimum value: Where Q is a point in the point cloud, ∥PQ∥ is the Euclidean distance, and a distance tolerance ∈ is set. When d(P,Point Cloud)<∈, P is considered to be on the point cloud surface; if the point cloud surface can be defined as a closed body, the relationship between the projection direction of the normal vector through point P and the normal vector is further combined to determine whether it is inside the geometric body; secondly, use Calculate the cross-sectional contour volume; finally, based on the calculated cross-sectional volume, use the sampling method to calculate the over-excavation and under-excavation volume of the tunnel after blasting. The calculation formula is:

[0084] 4. Over-excavation and under-excavation early warning monitoring:

[0085] First, according to the Technical Specifications for Highway Tunnel Construction and tunnel construction conditions, the outer and inner lines of the design section are set to form a safety range, and the outer and inner lines are marked in the cross-section diagram to form an allowable range for over-excavation and under-excavation. Then, according to the degree of risk within the safety range, the warning level is divided. Normal range (green): between the inner and outer lines, no adjustment is required; minor warning (yellow): close to the outer or inner line, need attention; severe warning (red): exceeding the outer or inner line, immediate rectification is required. For over-excavation, yellow warning: exceeding the outer line by 0 to 10 cm, red warning: exceeding the outer line by more than 10 cm, yellow warning for under-excavation: entering the inner line by 0 to 5 cm, red warning: entering the inner line The threshold is tightened appropriately in special areas (such as weak surrounding rock and faults) for more than 3cm. Then the actual cross-section point cloud is compared with the theoretical cross-section to analyze the over-excavation and under-excavation. The over-excavation area is marked in red and the under-excavation area is marked in blue. At the same time, a monitoring system needs to be installed in the tunnel to automatically upload point cloud data and analyze the over-excavation and under-excavation. When a yellow warning is issued, the robot will prompt the construction personnel to pay attention to adjustments. When a red warning is issued, an audible and visual alarm will be triggered and the project management personnel will be notified. The point cloud data, deviation analysis results and alarm records of each monitoring will be saved to facilitate a comprehensive analysis of the over-excavation and under-excavation data of the entire tunnel and evaluate the construction quality and construction technology level.

[0086] It should be noted that the point cloud data scanned by the robot is directly processed inside it and does not need to be transmitted to a computer terminal. Finally, after the over-excavation and under-excavation are mentioned, the over-excavation and under-excavation areas are intuitively displayed through color mapping. According to real-time monitoring, when over-excavation and under-excavation occur, the robot will issue an early warning to remind the staff in time.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for detecting over-excavation and under-excavation of surrounding rock in a curved tunnel, characterized by: The method specifically comprises the following steps: S1: After tunnel blasting, a robot is used to enter the tunnel and extract the excavation point cloud data of the inner wall surface through 3D laser scanning; S2: Obtain the tunnel section based on the extracted point cloud data; S3: Determine the over-break and under-break volume of the tunnel according to the tunnel section; S4: Design monitoring thresholds based on the designed excavation section profile and conduct over-excavation and under-excavation monitoring of the tunnel.

2. The method for detecting over-break and under-break of surrounding rock of a curved tunnel according to claim 1, characterized in that: In step S1, it specifically includes: S11. First, the excavation point cloud data of the inner wall surface is extracted by using robot 3D scanning, including arranging target balls in the curved tunnel; S12. After the target sphere is arranged, the coordinates of the control points of each tunnel pile are measured by using a total station, so as to obtain the coordinates of the crosshair midpoint of the target sphere, and the segment point cloud is obtained by using a robot 3D laser scanner, from which the spherical target is extracted and its center is fitted by the principle of rear intersection; S13. Output point cloud data under unified coordinates.

3. The method for detecting over-excavation and under-excavation of surrounding rock of a curved tunnel according to claim 2, characterized in that: In step S12, it includes: 1) The principle of resection is to measure the angle or distance from the unknown point to several known points, and combine the coordinates of these known points to calculate the coordinates of the unknown point through geometric or mathematical methods; the known point is the target sphere, which requires at least two (usually three or more) control points with known coordinates. The positions of these points are called known points; 2) The robot performs a 360° scan in a direction perpendicular to the line direction or at a 45-degree angle to the line direction, extracts the spherical target and fits its center; 3) The three-dimensional coordinates of the center point of each target sphere measured by the total station using back intersection are called the reference point set P = {P1, P2, ..., P n }, the target sphere center coordinates extracted from the point cloud, these points are called the point set to be aligned Q = {q1,q2,...,q n }; According to the relative position of the target ball, establish the corresponding relationship between the center point of the ball in the point cloud and the center point of the ball measured by the total station to ensure that p i and q i One to one correspondence; 4) Rigid body transformation model, that is, transforming the point set Q of the point cloud coordinate system to the total station coordinate system P, the formula is, p i =R×q i +T; R represents the rotation matrix, T represents the translation vector; 5) Solve the rigid body change and use the least squares method to solve it. According to the definition of the objective function: Solve for the optimal R and T to minimize the error E, and use Horn to quickly solve the rigid body transformation parameters; 6) Apply rigid body transformation and use the calculated R and T to transform the target sphere center coordinate q in the point cloud data i Transform to total station coordinate system: Verify and optimize the calculated results and calculate the error If the error is large, the iterative closest point algorithm is used to further align the point cloud with the reference point.

4. The method for detecting over-break and under-break of surrounding rock of a curved tunnel according to claim 3, characterized in that: In step S2, it specifically includes: S21. Obtain at least 3 pairs of target sphere center coordinates and obtain the absolute point cloud through inter-station coordinate conversion: Calculate the transformation parameters from the local coordinate system of the point cloud to the global coordinate system through at least 3 pairs of target sphere center coordinates, including the rotation matrix R and the translation vector T: p absolute =R*P local +T, convert point cloud data from the local coordinate system to the absolute coordinate system; S22. Obtain the tunnel median line according to the processed point cloud data: extract the tunnel median line by projecting the point cloud on the XOY plane, extracting the outer rectangle of the point cloud on the XOY plane to determine the axial direction of the tunnel, calculating the centerline point in combination with the vault, and jointly designing the centerline to encrypt the tunnel centerline, and finally obtaining the tunnel median line; S23. Based on the median line extracted from the tunnel, the cross section of the vertical line cluster of the tunnel microelement is extracted.

5. The method for detecting over-break and under-break of surrounding rock of a curved tunnel according to claim 4, characterized in that: In step S22, the maximum and minimum coordinates of the tunnel point cloud in the XOY plane are X max , X min , Y max , Y min , it is possible to determine an enclosing rectangle, and determine the tunnel direction and approximate centerline position according to the direction of the long axis of the rectangle; the arch vertex, that is, the highest point in the tunnel, is used as the arch vertex in the cross-section point cloud; the arch vertex and the center point of the bottom of the cross-section are connected to calculate the geometric center of the cross-section as the centerline point; The tunnel centerline is encrypted by combining the designed centerline with the point cloud centerline points, comparing the deviations between the designed centerline points and the point cloud centerline points, evaluating the degree to which the actual tunnel deviates from the design, and using the weighted fitting method to combine the point cloud centerline points with the designed centerline points to generate an encrypted and optimized tunnel centerline.

6. The method for detecting over-break and under-break of surrounding rock of a curved tunnel according to claim 5, characterized in that: In step S23, it includes: 1) Using three-dimensional coordinates P i (x i ,y i ,z i ) represents the tunnel centerline, the construction of the centerline microelement and the vertical line cluster includes: microelement tangent vector calculation, and the calculation of the tangent vector of the centerline data points in sequence: T i Indicates that the midline is at P i The tangent vector at P i-1 ,P i+1 Indicates the points before and after the midline; 2) At each midline point P i , construct the normal plane, the normal plane is perpendicular to the tangent vector T i , with an optional reference direction U, the normal vector is calculated by cross product: N i =T i *U, forming three directions of the plane: tangent vector T i , normal vector N i , and its orthogonal complement vector B i =T i ×N i ; Vertical line cluster generation: At each point P i On the normal plane, P i Generate a vertical line cluster as the center, use radius R as the threshold, and limit the range of point cloud points on the normal plane; 3) Filter the point cloud and generate cross-section points, including: calculation of the distance from the point cloud point to the normal plane, that is, the calculation formula is: d = N i *(Q j -P i ), and at the same time, it is necessary to screen the point cloud points Q that satisfy |d|<∈ j ; Project the point cloud onto the plane and filter out the point cloud points Q j Project it onto the normal plane and calculate its two-dimensional coordinates. The calculation formula is: Q' j ={(B i *(Q j -P i ),N i *(Q i -P i ))}, then we get the two-dimensional point set of the tunnel section; 4) Fitting the cross-sectional curve, using the least squares method to fit the point cloud data, that is, Thus, the cross-section geometric parameters are obtained; the results are outputted, and all micro-element cross-sections are integrated into a cross-section point cloud of a complete micro-element vertical line cluster; based on the extracted cross-section point cloud, the point cloud is subjected to non-uniform downsampling optimization; 5) Calculate the local density of the point cloud of the section using the following formula: Ni is the number of neighborhood points of point Pi within radius r, and Vi is the neighborhood search volume. The point cloud data is divided into sparse and dense areas according to the calculated threshold, and the basis for the division is: ρ i >ρ thresh For dense areas, the opposite is sparse areas; for the point cloud in dense areas, the voxel grid downsampling method is used to optimize, and the voxel grid side length d is set dense , usually 5-10 mm, only one point is retained in this area; a smaller voxel side length d is used for sparse areas sparse , the steps are the same as those in the dense area; 6) In the key geometric feature areas including the vault and the wall base, retain the high curvature point cloud, according to the curvature C of the point i , the screening point meets C i >C thresh point, and Where λ1, λ2, λ3 are the eigenvalues ​​of the covariance matrix; the processed point cloud data are recombined to form an optimized cross-sectional point cloud D'={P1', P2', ..., P n '}; Project the optimized cross-section point cloud and perform noise reduction; 7) According to the tunnel design file or the extracted centerline, take the tunnel centerline point as the origin, project the point cloud vertically to the reference plane, and calculate the projection point P'(x', y', z') of the point P(x, y, z) on its reference plane; assuming the reference plane is a plane, the projection formula is: N is the normal vector of the reference plane, d is the offset of the reference plane, and the error is checked and corrected after projection. Let the distance from point P to the reference plane be d. If d>∈, ∈ is the error threshold, it is marked as an abnormal point, and the abnormal point is smoothly adjusted to ensure that it is projected within the range of the reference plane; 8) Apply SOR filtering technology to the projected point cloud data to remove noise points, calculate the number of points k in the neighborhood of each point, identify dense areas and sparse areas, and if the density difference is too large, it is necessary to fill in the sparse area or downsample the dense area. The number of neighborhood points k needs to be adjusted according to the density of the cross-section point cloud; for each point, calculate the average distance from its neighboring points to the point, and remove outliers whose distance mean exceeds the standard deviation multiple threshold. The standard deviation multiple σ is usually set to 1-2; use a smoothing algorithm to remove local irregular fluctuations; compare the projected point cloud after denoising with the original point cloud to confirm that the feature points are not over-smoothed, and finally save the projected and denoised point cloud data in a standard format for subsequent analysis or modeling.

7. The method for detecting over-break and under-break of surrounding rock of a curved tunnel according to claim 6, characterized in that: In step S3, the simplified point cloud is obtained, and the over-excavation and under-excavation volumes are calculated using the quasi-Monte Carlo method, which specifically includes: S31, constructing a minimum construction minimum circumscribed cube of the point cloud to provide a sampling space range; S32. The random body generated in the bounding volume generates coordinates through uniformly distributed random numbers, and its coordinate range must satisfy: x min ≤x≤x max ,y min ≤y≤y max ,z min ≤z≤z max ; Based on the distance detection from point to point cloud surface, determine whether it falls within the cross section described by the point cloud; S33, calculate the cross-sectional contour volume, the calculation formula is: V pcd is the volume of the model to be tested, V aabb is the volume of the axisymmetric bounding box, Np all Np is the total number of generated sampling points. in is the number of sampling points in the model; S34. Based on the cross-sectional contour volume obtained by the quasi-Monte Carlo method, the over-break and under-break volume after tunnel blasting is calculated using the sampling method. The calculation formula is: Where: V over is the tunnel over-excavation volume, V under is the undercut volume of the tunnel, V aabb is the volume of the bounding box, P over is the number of sampling points in the over-excavation area, P under is the number of sampling points in the under-excavation area, P all is the number of all sampling points.

8. The method for detecting over-break and under-break of surrounding rock of a curved tunnel according to claim 7, characterized in that: In step S4, it specifically includes: S41. According to the Technical Specifications for Highway Tunnel Construction and tunnel construction conditions, the outer and inner envelopes of the design section are set to form a safety range; S42, the outer and inner envelopes of the cross section are the designed cross section plus the allowable over-break range and the designed cross section minus the allowable under-break range, respectively; S43. According to the risk level within the safety range, set a percentage and divide the warning level.

9. A monitoring and early warning system for over-excavation and under-excavation of surrounding rock in a curved tunnel, characterized by: The system adopts the method as claimed in any one of claims 1 to 8.

Images