Investigation and recovery method for choke plug failure water permeation accident of underground powerhouse of hydropower station

Abstract

The invention discloses an investigation and restoration method for a hydropower station underground powerhouse bulkhead failure water permeation accident, and belongs to the field of accident detection.According to the method, three-dimensional information of important parts such as an accident site bulkhead is efficiently collected under the complex environment condition, complete splicing is conducted through a feature fitting algorithm, key parameters of the bulkhead are quantitatively analyzed, and the bulkhead failure process is accurately restored; the problems of incomplete three-dimensional point cloud data acquisition, low point cloud data feature extraction and registration precision, lack of dynamic recovery capability and the like in a complex environment are solved, and a set of choke plug failure accident quantitative and precise investigation system from data acquisition, feature analysis to dynamic simulation is formed. The method breaks through the limitation of a traditional method in the aspects of integrity, precision and mechanism restoration, provides technical support for safe operation of the hydropower station, and provides method reference for investigation of similar accident reasons.

Description

Technical Field

[0001] The present invention relates to the field of accident detection, and particularly to an investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station. Background Art

[0002] As a key sealing structure of the penstock system, the failure of the bulkhead may lead to catastrophic water leakage accidents. Traditional accident investigation methods rely on manual exploration, two-dimensional measurement, and empirical inference, making it difficult to accurately restore the failure mechanism of the bulkhead. With the application of 3D scanning technology, some studies have attempted to analyze the accident causes through reverse modeling of point cloud data. However, due to complex working condition constraints and technical bottlenecks, existing methods still have significant deficiencies in data integrity, model accuracy, and dynamic restoration capabilities, which are specifically manifested in the following aspects: First, the integrity of data acquisition is insufficient. Most existing 3D scanning technologies use fixed-resolution single-site scanning, which is easily affected by occlusion and reflection interference in the complex environment of the underground powerhouse of a hydropower station, resulting in the lack of point cloud in key areas (such as the fracture surface of the bulkhead). In addition, after the separation of the bulkhead shedding body and the residual body, existing methods often ignore the independent scanning of the shedding body and only rely on the data of the on-site residual body for reconstruction, resulting in the distortion of the geometric features of the model and being unable to reflect the true shape before failure. Second, the accuracy of point cloud processing and feature extraction is low. Traditional denoising algorithms are difficult to distinguish noise from tiny structural features, especially in key areas such as bulkhead welds and bolt holes, which are prone to incorrect deletion. Edge detection mostly relies on the normal vector variance method, but it is not sensitive enough to areas with sudden curvature changes, resulting in a large deviation in the fracture path of the fracture line extraction. The center fitting of the curved surface often uses the least squares method, without considering the distribution deviation caused by asymmetric wear. In addition, the registration of point cloud data often uses mainstream registration technologies (such as the standard ICP algorithm), but this technology only relies on the minimization of geometric distance and does not incorporate the actual stress conditions of the bulkhead (such as water pressure distribution and bolt pre-tightening force), resulting in insufficient mechanical rationality of the assembled model. Third, the lack of dynamic restoration ability. Existing studies mostly generate 3D models based on static point clouds, and can only infer the failure process through cross-section analysis, lacking dynamic simulation of crack propagation paths and bulkhead shedding trajectories, and the credibility of extracting key parameters (such as detonation points and stress peaks) is low. Summary of the Invention

[0003] In view of the above deficiencies in the prior art, the present invention provides an investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station.

[0004] To achieve the above invention object, the technical solution adopted by the present invention is as follows: An investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station, comprising the following steps: S1. Collect the three-dimensional point cloud data of the residual body of the plug head at the accident site of the underground powerhouse of the hydropower station through non-contact multi-station three-dimensional scanning, and collect the three-dimensional point cloud data of the detached body of the plug head through non-contact multi-station three-dimensional scanning in a stable indoor place; S2. Perform splicing and denoising preprocessing on the three-dimensional point cloud data of the complex environment of the residual body of the plug head and the three-dimensional point cloud data of the detached body of the plug head obtained in S1; S3. Calculate the minimum variance unbiased estimate value of the surface center of the three-dimensional point cloud data of the detached body of the plug head and the three-dimensional point cloud data of the residual body of the plug head after splicing, and use the vector method to extract the edge feature lines of the residual body of the plug head and the detached body of the plug head to generate an edge point cloud data set; S4. Force the minimum variance unbiased estimate values of the surface centers of the three-dimensional point cloud data of the detached body of the plug head and the three-dimensional point cloud data of the residual body of the plug head calculated in S3 to coincide, and calculate the optimal solution of the surface fitting rotation matrix based on the obtained edge point cloud data set; S5. Perform point cloud coordinate rotation transformation again, complete the complete point cloud fitting of the plug head body to obtain the complete point cloud characteristics of the plug head body, and perform three-dimensional quantitative analysis based on the obtained complete point cloud characteristics of the plug head body to restore the accident occurrence process.

[0005] Further, the S1 specifically includes the following steps: S11. Set the scanning route and arrange the scanning stations according to the visibility, stability, overlap degree conditions of the scanning site and the position and shape of the scanning object; S12. Level the three-dimensional scanning instrument at the designed station, evenly arrange the common targets for splicing, and select the resolution and precision scanning parameters to collect the 360-degree three-dimensional information of the station; S13. Reset the scanning parameters, perform local encrypted scanning on the residual body of the plug head, and collect the data of the next station after completion; S14. Collect the three-dimensional point cloud data of the detached body of the plug head through non-contact multi-station three-dimensional scanning; Further, the S2 specifically includes the following steps: S21. Denoise the cloud data of each station according to the distance range, reflectivity, isolated points, and curvature factors; S22. Uniformly number the common targets in the cloud data of each station, and use the 3D point cloud target detection algorithm to automatically identify and extract the center points of the common targets; S23. Perform coordinate registration and transformation calculation on the adjacent station data, and perform weighted fusion on the overlapping areas to respectively obtain the point cloud data of the residual body of the plug head and the detached body of the plug head after splicing.

[0006] Further, the S3 specifically includes the following steps: S31. Respectively extract the maximum numerical profile curves of the point cloud data of the residual body of the plug head and the detached body of the plug head after splicing; S32. Make the maximum numerical profile curves of the plug remaining body and the plug detachment body point cloud data lie in the same plane, and use this plane as the center of a circle, fit with a certain radius, and calculate the optimization objective function of the sum of squared errors; S33. Let the optimization objective function of the sum of squared errors be the minimum value, and obtain the minimum variance unbiased estimate of the surface center of the plug remaining body and the plug detachment body point cloud data by solving the minimum values of the center and radius of the plane in S32 through a quadratic differential equation; S34. Construct a spherical domain centered at an arbitrary sampling point with a radius of twice the point cloud resolution, construct vectors from a point to each point within the domain and normalize the vectors, and calculate the sum of the vectors within the domain; S35. Determine the position of this point according to the magnitude of the sum of the vectors within the domain calculated. The larger the value, the more marginal this point is; S36. Uniformly sample in the plug remaining body and the plug detachment body point cloud data, repeat steps S34 - S35, and screen according to the maximum value of the sum of the vectors within the domain to obtain the edge point cloud data set.

[0007] Furthermore, the optimization objective function of the sum of squared errors in S32 is expressed as:

[0008] In the formula, is the optimization objective function of the sum of squared errors, is the feature point on the surface of the plug remaining body or the plug detachment body point cloud data, is the feature point index, is the total number of feature points.

[0009] Furthermore, the specific steps of S4 are as follows: S41. Let the surface center coordinates of the plug remaining body be , and the surface center coordinates of the plug detachment body be , calculate the translation vector as , and apply a translation transformation to all the point cloud coordinates of the plug detachment body to force the surface center of the plug remaining body to coincide with the surface center of the plug detachment body, eliminating the global position deviation; S42. Search for the point with the closest Euclidean distance in the plug remaining body for each point in the edge point cloud data of the translated plug detachment body to form a set of point pairs, calculate the centroid of the set of point pairs, then construct a covariance matrix and perform singular value decomposition on the covariance matrix to obtain the initial solution of the surface registration rotation matrix; S43. Repeat steps S41 - S42 until the registration error converges or reaches the maximum number of iterations; S44. Starting from the initial solution of the surface stitching rotation matrix obtained, perform non-linear optimization on the stitching objective function PT function. When the change amount of the stitching objective function PT function is less than the threshold or the maximum number of iterations is reached, the optimal solution of the stitching rotation matrix is obtained.

[0010] Further, the stitching objective function PT in S44 is expressed as:

[0011] In the formula, is the initial solution of the surface stitching rotation matrix, is the traditional ICP registration error, is the surface curvature consistency constraint term, is the mechanical balance constraint term, , , is the weight coefficient, which is calibrated according to specific experiments.

[0012] Further, the S5 specifically includes the following steps: S51. Apply the rotation matrix to the translated plug detachment body point cloud, and further merge the obtained rotated data with the plug residue body point cloud to obtain the three-dimensional point cloud data of the complete plug body; S52. Mark the potential fracture area through the curvature mutation detection algorithm, convert the three-dimensional point cloud data of the complete plug body into a NURBS surface or voxel model, and reconstruct the three-dimensional geometry of the plug; S53. Set the accident condition boundary, perform transient dynamic dynamics simulation, and simulate the plug failure process.

[0013] The present invention has the following beneficial effects: ① In terms of data acquisition and modeling, by adopting the multi-station dynamic parameter adjustment and local dense scanning technology, combined with the reflectivity and curvature multi-modal denoising algorithm, the problem of occlusion interference in the complex environment of the underground powerhouse is overcome.

[0014] ② Through the vector method edge detection and the minimum variance unbiased estimation surface fitting algorithm, the deviation of the fracture path extraction is reduced, and the center location accuracy is improved. In the model stitching link, innovatively integrate the mechanical balance constraint and curvature consistency into the point cloud registration, construct the multi-objective optimization function PT, and break through the limitation of the traditional ICP algorithm that only relies on geometric alignment.

[0015] ③ Aiming at the problem of restoring the accident dynamic process, the present invention extracts parameters such as fracture lines, weld lines, and detonation points, and combines virtual reality technology to simulate the crack propagation path, plug detachment trajectory, and water penetration process, realizing the leap from "two-dimensional empirical judgment" to "three-dimensional data-driven analysis".

[0016] The present invention is significantly innovative in three aspects: method architecture (3D scanning + quantitative restoration), algorithm improvement (edge extraction, stitching optimization), and application scenario (failure investigation of the plug in hydropower stations), significantly improving the accuracy and efficiency of the failure investigation of the plug in hydropower stations. Description of the Drawings

[0017] Figure 1 It is a schematic flow chart of the investigation and restoration method for the water leakage accident caused by the failure of the plug in the underground powerhouse of a hydropower station according to the present invention. Detailed Embodiments

[0018] The following describes the detailed embodiments of the present invention to facilitate those skilled in the art of the present technology to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the detailed embodiments. For those of ordinary skill in the art of the present technology, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions made using the concept of the present invention are within the scope of protection.

[0019] An investigation and restoration method for the water leakage accident caused by the failure of the plug in the underground powerhouse of a hydropower station, as Figure 1 shown, includes the following steps: S1. Collect the three-dimensional point cloud data of the remaining body (R) of the plug at the accident site in the underground powerhouse of the hydropower station through non-contact multi-site three-dimensional scanning, and collect the three-dimensional point cloud data of the detached body (L) of the plug in a stable indoor location through non-contact multi-site three-dimensional scanning; In this embodiment, the three-dimensional point cloud data of the remaining body (R) of the plug and the complex environment at the accident site in the underground powerhouse of the hydropower station are collected through non-contact multi-site three-dimensional scanning.

[0020] S11. Set the scanning route and arrange the scanning stations according to the visibility, stability, overlap degree conditions of the scanning site and the position and shape of the scanning object; S12. Level the three-dimensional scanning instrument at the designed station, evenly arrange the common targets for stitching, and select the scanning parameters of resolution and accuracy to collect the 360-degree three-dimensional information of the station.

[0021] S13. Reset the scanning parameters, perform local encrypted scanning on the remaining body of the plug, and collect the data of the next station after completion; S14. Collect the three-dimensional point cloud data of the detached body (L) of the plug through non-contact multi-site three-dimensional scanning; S2. Stitch and preprocess the three-dimensional point cloud data of the complex environment of the remaining body of the plug and the three-dimensional point cloud data of the detached body of the plug obtained in S1; In this embodiment, it specifically includes the following steps: S21. Denoise the cloud data of each station according to the distance range, reflectivity, isolated points, and curvature factors; S22. Uniformly number the common targets in the cloud data of each site, and automatically identify and extract the center points of the common targets using a 3D point cloud target detection algorithm; S23. Perform coordinate registration and transformation calculations on the data of adjacent sites, and perform weighted fusion on the overlapping areas to respectively obtain the point cloud data of the spliced blind head residue (R) and the blind head shedding body (L).

[0022] S3. Calculate the minimum variance unbiased estimate value of the surface center of the three-dimensional point cloud data of the spliced blind head shedding body and the three-dimensional point cloud data of the blind head residue, and use the vector method to extract the edge feature lines of the blind head residue and the blind head shedding body to generate an edge point cloud data set; In this embodiment, the specific steps are as follows: S31. Respectively extract the maximum numerical profile curves (Q R 、Q L ) of the point cloud data of the spliced blind head residue and the blind head shedding body; S32. Make the maximum numerical profile curves of the point cloud data of the blind head residue and the blind head shedding body in the same plane, and use this plane as the center of the circle, as the radius for fitting, and calculate the optimization objective function of the sum of squared errors; Let Q R 、Q L be in the same plane, and respectively perform fitting of the center of the plane and the radius . The optimization objective function of the sum of squared errors is: , where ( , ) is the feature point on the curve of Q R or Q L .

[0023] S33. Let the optimization objective function of the sum of squared errors be the minimum value, and obtain the minimum variance unbiased estimate values 、 of the center of the surface of the point cloud data of the blind head residue and the blind head shedding body by solving the minimum values of the center of the plane and the radius in S32 through a quadratic differential equation; Intelligently extract the edge feature lines of R and L using the vector method to generate an edge point cloud data set.

[0024] S34. Construct a spherical domain centered on any sampling point with a radius of twice the point cloud resolution. The set of points within the domain , k is the number of points within the domain. Construct the vector between the point , perform vector normalization, and then calculate the sum of vectors within the domain ; S35. According to the magnitude of the value, make an intelligent judgment. The larger the value, the more marginal the point is; S36. Uniformly sample in the point cloud data of the plug remaining body and the plug shedding body, repeat steps S34 - S35, and screen according to the maximum value of the sum of vectors within the domain value to obtain the edge point cloud data set.

[0025] S4. Force the minimum variance unbiased estimate value of the surface center calculated in S3 to coincide, and calculate the optimal solution of the surface fitting rotation matrix based on the obtained edge point cloud data set; This embodiment specifically includes the following steps: S41. Let the surface center coordinates of the plug remaining body be , and the surface center coordinates of the plug shedding body be , calculate the translation vector as , and then apply a translation transformation to all point cloud coordinates of the plug shedding body (DL) , force the surface center of L to coincide with the surface center OR of R, and eliminate the global position deviation; S42. Search for the point with the closest Euclidean distance in the plug remaining body for each point in the edge point cloud data of the translated plug shedding body to form a point pair set, calculate the centroid of the point pair set, then construct a covariance matrix, and perform a singular value decomposition on the covariance matrix to obtain the initial solution of the surface fitting rotation matrix; Based on the ICP registration of the plug shedding body and the plug remaining body, ① For each point in the edge point cloud data set of the translated plug shedding body , search for the point with the closest Euclidean distance in DR , and form a point pair set .

[0026] ② Calculate the centroids of the two sets of points , , then construct a covariance matrix , and perform a singular value decomposition (SVD) on H: , to obtain the initial solution of the surface fitting rotation matrix .

[0027] Among them is the transpose of the matrix, is the left singular vector matrix, and its column vectors are the eigenvectors of, representing the main direction of the target point cloud (DR). is the right singular vector matrix, and its column vectors are The eigenvector, representing the point cloud to be registered The main direction of is a diagonal matrix containing The singular values of, characterizing the alignment strength of the main directions of the two groups of point clouds

[0028] S43. Repeat steps S41 - S42 until the registration error converges or the maximum number of iterations is reached; S44. Perform non - linear optimization of the PT function starting from the initial solution of the surface - fitting rotation matrix. When the change in the fitting objective function PT is less than the threshold or the maximum number of iterations is reached, the optimal solution of the fitting rotation matrix is obtained.

[0029] Since the ICP algorithm only relies on geometric distance matching and will ignore the continuity of the blind - head surface, mechanical constraints, etc., optimize the PT function

[0030] In the formula, is the initial solution of the surface - fitting rotation matrix is the traditional ICP registration error is the mechanical balance constraint term , , is the weight coefficient, calibrated according to specific experiments is the surface curvature consistency constraint term, calculating the curvature difference between R and L after fitting is the mechanical balance constraint term, based on the force model before the blind - head failure (such as water flow pressure, bolt pre - tightening force, etc.), verifying the rationality of the force after fitting , , is the weight coefficient, calibrated according to specific experiments

[0031] Starting from the obtained , use the Levenberg - Marquardt algorithm to perform non - linear optimization on PT(T). When The change is less than the threshold (such as 1e - 5) or the maximum number of iterations is reached, obtain the optimal solution of the surface - fitting rotation matrix ( ).

[0032] S5. Perform point - cloud coordinate rotation transformation again to complete the fitting of the complete point cloud of the blind - head body and obtain the complete point - cloud characteristics of the blind - head body. Based on the obtained complete point - cloud characteristics of the blind - head body, perform three - dimensional quantitative analysis to restore the accident process.

[0033] In this embodiment, it specifically includes the following steps: S51. Apply the rotation matrix to the point cloud of the translated plug-off body, and further merge the obtained rotated data with the point cloud of the plug-remaining body to obtain the three-dimensional point cloud data of the complete plug body; For the point cloud of the translated off-body Apply the rotation matrix

[0034] Merge with the remaining body point cloud DR to obtain the three-dimensional point cloud data of the complete plug body (W).

[0035] S52. Mark the potential fracture regions through the curvature mutation detection algorithm, convert the three-dimensional point cloud data of the complete plug body into a NURBS surface or a voxel model, and reconstruct the three-dimensional geometric structure of the plug; Mark the potential fracture regions through the curvature mutation detection algorithm, manually assist in correcting the fracture path, and extract parameters such as the fracture line length, trend angle, and fracture surface roughness. Identify the weld region based on the difference in point cloud reflectivity, measure the weld width and penetration depth. Near the endpoints of the fracture line, determine the energy release source through local stress concentration analysis, determine the initiation point, extract the discontinuous regions in the point cloud after splicing, verify the detachment position in combination with the bolt hole position coordinates, determine the detachment point, and convert the point cloud W into a NURBS surface or a voxel model to reconstruct the three-dimensional geometric structure of the plug.

[0036] S53. Set the accident condition boundary and perform transient dynamic simulation to simulate the plug failure process. Set the accident condition boundary (such as water flow pressure, temperature, etc.), perform transient dynamic simulation to simulate the plug failure process, and combine virtual reality technology to display the crack propagation, plug detachment trajectory, and water penetration process in time series.

[0037] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0038] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the function specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 specified in one block or a plurality of blocks.

[0039] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the function specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 specified in one block or a plurality of blocks.

[0040] Specific embodiments are used in the present invention to illustrate the principles and implementation manners of the present invention. The descriptions of the above embodiments are only for helping to understand the method and its core idea of the present invention; meanwhile, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

[0041] Those of ordinary skill in the art will realize that the embodiments described herein are for helping the reader understand the principles of the present invention, and it should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations that do not depart from the essence of the present invention according to the technical revelations disclosed in the present invention, and these deformations and combinations are still within the protection scope of the present invention.

Claims

1. An investigation and restoration method for the water penetration accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station, characterized in that, It includes the following steps: S1. Collect the three-dimensional point cloud data of the residual body of the bulkhead at the accident site of the underground powerhouse of the hydropower station through non-contact multi-station three-dimensional scanning, and collect the three-dimensional point cloud data of the bulkhead shedding body through non-contact multi-station three-dimensional scanning in a stable indoor place; S2. Perform splicing and denoising preprocessing on the three-dimensional point cloud data of the complex environment of the residual body of the bulkhead and the three-dimensional point cloud data of the bulkhead shedding body obtained in S1; S3. Calculate the minimum variance unbiased estimate value of the surface center of the three-dimensional point cloud data of the bulkhead shedding body and the three-dimensional point cloud data of the residual body of the bulkhead after splicing, and use the vector method to extract the edge feature lines of the residual body of the bulkhead and the bulkhead shedding body to generate an edge point cloud data set; S4. Force the minimum variance unbiased estimate values of the surface centers of the three-dimensional point cloud data of the bulkhead shedding body and the three-dimensional point cloud data of the residual body of the bulkhead calculated in S3 to coincide, and calculate the optimal solution of the surface fitting rotation matrix based on the obtained edge point cloud data set; S5. Perform point cloud coordinate rotation transformation again to complete the complete point cloud fitting of the bulkhead body to obtain the complete point cloud characteristics of the bulkhead body, and perform three-dimensional quantitative analysis based on the obtained complete point cloud characteristics of the bulkhead body to restore the accident occurrence process.

2. The investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station according to claim 1, characterized in that, The specific steps of S1 are as follows: S11. Set the scanning route and arrange the scanning stations according to the visibility, stability, overlap degree conditions of the scanning site and the position and shape of the scanning object; S12. Level the three-dimensional scanning instrument at the designed station, uniformly arrange the common targets for splicing, and select the scanning parameters of resolution and accuracy to collect the 360-degree three-dimensional information of the station; S13. Reset the scanning parameters, perform local encrypted scanning on the residual body of the bulkhead, and collect the data of the next station after completion; S14. Collect the three-dimensional point cloud data of the bulkhead shedding body through non-contact multi-station three-dimensional scanning.

3. The investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station according to claim 1, characterized in that, The specific steps of S2 are as follows: S21. Denoise the cloud data of each station according to the distance range, reflectivity, isolated points, and curvature factors; S22. Uniformly number the common targets in the cloud data of each station, and use the 3D point cloud target detection algorithm to automatically identify and extract the center points of the common targets; S23. Perform coordinate registration and transformation calculation on the adjacent station data, and perform weighted fusion on the overlapping areas to obtain the point cloud data of the residual body of the bulkhead and the bulkhead shedding body after splicing respectively.

4. The investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station according to claim 1, characterized in that, The specific steps of S3 are as follows: S31. Extract the maximum numerical profile curves of the point cloud data of the residual body of the bulkhead and the bulkhead shedding body after splicing respectively; S32. Make the maximum numerical profile curves of the plug residual body and the plug detachment body point cloud data lie in the same plane, and use this plane as the center of the circle, fit with the radius, and calculate the optimization objective function of the sum of squared errors; S33. Let the optimization objective function of the squared error be the minimum value, and obtain the minimum value of the center and radius of the plane in S32 through a quadratic differential equation to obtain the minimum variance unbiased estimate value of the surface center of the point cloud data of the residual body of the bulkhead and the bulkhead shedding body; S34. Construct a spherical region centered at an arbitrary sampling point with a radius of twice the point cloud resolution, construct vectors from a point to each point within the region and normalize the vectors, and calculate the sum of the vectors within the region; S35. Judge the position of this point according to the magnitude of the sum value of the vectors within the calculated domain. The larger the value, the more marginal this point is; S36. Uniformly sample in the point cloud data of the residual body of the bulkhead and the bulkhead shedding body, repeat steps S34-S35, and screen according to the maximum value of the sum value of the vectors within the domain to obtain an edge point cloud data set.

5. The investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station according to claim 4, characterized in that, The optimization objective function of the squared error in S32 is expressed as: In the formula, is the optimization objective function of the squared error, is the feature point of the surface of the blind head residue or the blind head detachment body point cloud data, is the feature point index, is the total number of feature points.

6. The investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station according to claim 1, characterized in that, The specific steps of S4 are as follows: S41. Set the surface center coordinates of the plug remaining body as , and the surface center coordinates of the plug detached body as . Calculate the translation vector as , and apply the translation transformation to all the point cloud coordinates of the plug detached body to force the surface center of the plug remaining body to coincide with the surface center of the plug detached body, eliminating the global position deviation; S42. Search for the point with the closest Euclidean distance to each point in the edge point cloud data of the translated plug body in the plug residue body to form a set of point pairs, calculate the centroid of the set of point pairs, then construct a covariance matrix and perform singular value decomposition on the covariance matrix to obtain the initial solution of the surface fitting rotation matrix; S43. Repeat steps S41 - S42 until the registration error converges or the maximum number of iterations is reached; S44. Perform non - linear optimization of the fitting objective function PT function starting from the initial solution of the surface fitting rotation matrix obtained. When the change amount of the fitting objective function PT function is less than the threshold or the maximum number of iterations is reached, obtain the optimal solution of the fitting rotation matrix.

7. The investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station according to claim 6, characterized in that, The fitting objective function PT in S44 is expressed as: In the formula, is the initial solution of the surface fitting rotation matrix, is the traditional ICP registration error, is the surface curvature consistency constraint term, is the mechanical equilibrium constraint term, , , are weight coefficients, which are calibrated according to specific experiments.

8. The investigation and restoration method for the water leakage accident caused by the failure of the bulkhead in the underground powerhouse of a hydropower station according to claim 1, characterized in that, The specific steps of S5 are as follows: S51. Apply the rotation matrix to the point cloud of the translated plug body, and further merge the obtained rotated data with the point cloud of the plug residue body to obtain the three - dimensional point cloud data of the complete plug body; S52. Mark the potential fracture regions through the curvature mutation detection algorithm, convert the three - dimensional point cloud data of the complete plug body into a NURBS surface or a voxel model, and reconstruct the three - dimensional geometric structure of the plug; S53. Set the accident condition boundary, perform transient dynamic kinetic simulation, and simulate the plug failure process.

Images