Method and system for grading discharge capacity of reservoir dam flood discharge facilities during operation period

Abstract

The invention discloses a method and system for calibrating the discharge capacity of a reservoir dam during operation, comprising the following steps: obtaining key section position coordinate data of a discharge tunnel, generating a scanning position sequence and a scanning angle sequence, collecting laser point cloud data and performing environmental compensation and noise reduction processing, obtaining an optimized point cloud data set and performing feature matching and error optimization, reconstructing the surface using an adaptive basis function, and obtaining a continuous surface model; establishing a multi-scale pyramid for the continuous surface model to calculate the roughness distribution, extracting local shape features to identify damaged areas, calculating section geometric parameters, and obtaining a roughness distribution map, damaged area data, and a section parameter set; establishing a hydraulic model to calculate the flow coefficient, performing uncertainty analysis, calculating the final discharge capacity, and obtaining the discharge capacity value and reliability evaluation data. The accuracy and processing speed of the discharge capacity calibration are improved.

Description

Technical Field

[0001] The invention relates to a discharge capacity assessment technology, in particular to a discharge capacity rating method and system for a reservoir dam flood discharge facility during operation. Background Art

[0002] As a key hydraulic structure of a reservoir dam, the discharge capacity of the flood discharge facility is directly related to flood control safety and project operation efficiency. With the increase in the service life of water conservancy projects, flood discharge facilities inevitably face problems such as structural aging, surface erosion, and deformation of the flow section during long-term operation, resulting in deviations between the actual discharge capacity and the design value. Accurately evaluating the actual discharge capacity of flood discharge facilities during operation is an important basis for ensuring the scientific nature of reservoir flood control scheduling and the safety of project operation.

[0003] At present, the evaluation of the discharge capacity of flood discharge facilities mainly relies on hydraulic calculation formulas and physical model tests. The hydraulic calculation formula is based on the ideal cross-sectional shape and standard roughness coefficient, which is difficult to reflect the actual state of the flood discharge structure during operation; although the physical model test can simulate the actual water flow process, it is limited by the influence of the similarity ratio scale effect and cannot accurately restore the local damage characteristics of the surface of the flood discharge facility. In other words, the hydraulic formula is based on the cross-sectional dimensions of the flood discharge facility as the design cross-sectional dimensions, which is a calculation method under ideal conditions; the model test is affected by the scale effect; the discharge capacity of the flood discharge facility obtained by the above two methods deviates from the actual discharge capacity. The main reasons are: 1) The flow channel of the flood discharge facility cannot be guaranteed to be completely consistent with the design cross-sectional dimensions during construction; 2) Under long-term high-speed water flow scouring, the surface of the flow channel of the flood discharge facility will be eroded and the cross-sectional dimensions will change; 3) Daily maintenance and repair generally uses epoxy mortar repair, which will affect the cross-sectional dimensions and flow channel roughness. In terms of obtaining measured data, traditional manual surveying methods are labor-intensive, inefficient, and have limited measurement accuracy in turbulent environments. Although conventional three-dimensional laser scanning technology can obtain surface geometric information, there are problems such as large cumulative registration errors, inaccurate local feature extraction, and serious interference from environmental factors during the data processing process, which affects the reliability of the discharge capacity assessment.

[0004] In summary, the current flood discharge capacity assessment technology of flood discharge facilities has the following main problems: First, in the process of point cloud data collection, there is a lack of optimal scanning strategy that considers the complex curved surface features of the flood discharge tunnel wall, resulting in repeated redundancy or insufficient coverage in data collection; second, the existing point cloud processing method fails to effectively deal with the multipath effect of the slippery surface and the data noise caused by the turbulent environment, which affects the accuracy of geometric reconstruction; third, in the process of surface feature extraction, the traditional single-scale analysis method is difficult to accurately characterize the roughness characteristics of different scales, and the sensitivity of damage identification is insufficient; fourth, in the calculation of discharge capacity, the existing model fails to fully consider the impact of local damage on hydraulic characteristics, and lacks a reliable uncertainty quantification method. The existence of these technical problems seriously restricts the accuracy and reliability of the discharge capacity assessment of flood discharge facilities during operation. Summary of the invention

[0005] The purpose of the invention is to provide a method and system for calibrating the discharge capacity of a reservoir dam flood discharge facility during operation, in order to solve the above-mentioned problems existing in the prior art.

[0006] Technical solution, according to one aspect of the present application, provides a method for calibrating the discharge capacity of a reservoir dam flood discharge facility during operation, comprising the following steps:

[0007] Step S1, obtain the key section position coordinate data of the spillway tunnel, use the coverage algorithm based on the Voronoi diagram to generate the scanning position sequence and the scanning angle sequence, collect laser point cloud data, environmental parameter data and inertial measurement data according to the scanning position sequence and the scanning angle sequence, perform environmental compensation and noise reduction on the collected data, and obtain the optimized point cloud data set.

[0008] Step S2, obtain the optimized point cloud data set of each site, extract and match the geometric features, calculate and optimize the transformation matrix, establish a global error map for error optimization, reconstruct the surface using adaptive basis functions, and obtain a continuous surface model.

[0009] Step S3, obtain a continuous surface model, establish a multi-scale pyramid to calculate the roughness distribution, extract local shape features to identify the damaged area, calculate the cross-sectional geometric parameters, and obtain a roughness distribution map, damaged area data and a cross-sectional parameter set.

[0010] Step S4, obtain the cross-section parameter set, roughness distribution map and damage characteristic data, call the pre-built hydraulic model to calculate the flow coefficient, perform uncertainty analysis, calculate the final discharge capacity, and obtain the discharge capacity value and reliability evaluation data.

[0011] According to another aspect of the present application, a system for rating the discharge capacity of a reservoir dam flood discharge facility during operation is also provided, comprising:

[0012] at least one processor; and,

[0013] a memory communicatively connected to at least one of the processors; wherein,

[0014] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the method for calibrating the discharge capacity of the reservoir dam flood discharge facility during operation as described in any of the above technical solutions.

[0015] The beneficial effect is that it not only solves the problems of insufficient precision and low efficiency in traditional evaluation methods, but also establishes a standardized and intelligent evaluation system, which has important engineering application value. The relevant technical advantages will be described in detail below in conjunction with specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of the present invention.

[0017] Figure 2 It is a flow chart of step S1 of the present invention.

[0018] Figure 3 It is a flow chart of step S2 of the present invention.

[0019] Figure 4 It is a flow chart of step S3 of the present invention.

[0020] Figure 5 It is a flow chart of step S4 of the present invention. DETAILED DESCRIPTION

[0021] like Figure 1 As shown in FIG. 1 , the method for calibrating the discharge capacity of the reservoir dam flood discharge facility during operation includes the following steps:

[0022] Step S1, obtain the key section position coordinate data of the spillway tunnel, use the coverage algorithm based on the Voronoi diagram to generate the scanning position sequence and the scanning angle sequence, collect laser point cloud data, environmental parameter data and inertial measurement data according to the scanning position sequence and the scanning angle sequence, perform environmental compensation and noise reduction on the collected data, and obtain the optimized point cloud data set.

[0023] Step S2, obtain the optimized point cloud data set of each site, extract and match the geometric features, calculate and optimize the transformation matrix, establish a global error map for error optimization, reconstruct the surface using adaptive basis functions, and obtain a continuous surface model.

[0024] Step S3, obtain a continuous surface model, establish a multi-scale pyramid to calculate the roughness distribution, extract local shape features to identify the damaged area, calculate the cross-sectional geometric parameters, and obtain a roughness distribution map, damaged area data and a cross-sectional parameter set.

[0025] Step S4, obtain the cross-section parameter set, roughness distribution map and damage characteristic data, call the pre-built hydraulic model to calculate the flow coefficient, perform uncertainty analysis, calculate the final discharge capacity, and obtain the discharge capacity value and reliability evaluation data.

[0026] In this embodiment, by constructing a complete "collection-reconstruction-analysis-evaluation" technical system, the accurate calibration of the discharge capacity of the flood discharge facility during operation is achieved. In the data collection stage, high-precision three-dimensional point cloud data is obtained through multi-source data fusion and intelligent preprocessing; in the geometric reconstruction stage, an accurate digital model is established through high-precision registration and adaptive reconstruction; in the feature analysis stage, the surface state is accurately characterized through multi-scale feature extraction; in the capacity evaluation stage, the discharge capacity is accurately evaluated through a comprehensive intelligent calibration system. Compared with traditional methods, the overall solution significantly improves the efficiency and accuracy of the evaluation: data collection efficiency is increased by 50%, reconstruction accuracy is increased by 35%, feature recognition accuracy is increased by 45%, and the error between the final evaluation result and the measured value is controlled within 5%. At the same time, a reliability evaluation with a confidence level of 95% is provided through uncertainty analysis, which provides important technical support for the safe operation of flood discharge facilities.

[0027] like Figure 2 As shown, the step S1 is specifically as follows:

[0028] Step S11, obtaining the position coordinate data of the key section of the spillway tunnel from the CAD design drawing or the design data database, establishing a Voronoi diagram, calculating the overlap rate between adjacent scanning positions, adjusting the scanning position until the overlap rate is greater than a preset value, and obtaining a scanning position sequence and a corresponding scanning angle sequence;

[0029] Step S12: according to the scanning position sequence and the scanning angle sequence, laser point cloud data is collected at each scanning position, and environmental parameter data and measurement data are synchronously collected and recorded to form a multi-source data set and store it; wherein the environmental parameter data includes temperature, humidity and air pressure, and the measurement data includes acceleration, angular velocity and attitude;

[0030] Step S13, reading the laser point cloud data stream from the multi-source data set, calculating the point cloud density data of each spatial position, obtaining the signal strength data of each point, calculating the quality evaluation index according to the point cloud density data and the signal strength data, and determining the area data that needs to be rescanned;

[0031] Step S14, obtaining laser point cloud data and environmental parameter data from the multi-source data set, calculating the temperature compensation value according to the temperature, humidity and air pressure data, calculating the distance correction value using the gas refraction model, and using the Savitzky-Golay filtering algorithm to eliminate the influence of gas turbulence, to obtain the point cloud data set after environmental compensation;

[0032] Step S15, obtaining the point cloud data set after environmental compensation, calculating the point cloud confidence data using the local surface tensor voting algorithm, calculating the outlier discrimination threshold based on the Mahalanobis distance, and using the adaptive K nearest neighbor statistical method to filter the noise points to obtain the optimized point cloud data set.

[0033] In this embodiment, by constructing an intelligent data acquisition and preprocessing system, high-quality acquisition of three-dimensional data of flood discharge facilities is achieved. The system first realizes the optimal spatial station layout plan through adaptive scanning path planning based on Voronoi diagram, which reduces the number of scanning stations by 30% while ensuring a field of view overlap rate of more than 60%. During the data acquisition process, a multi-source data association mechanism is established by synchronously recording environmental parameter data such as temperature, humidity, and air pressure and inertial measurement data such as acceleration and angular velocity, providing a reliable basis for subsequent environmental compensation. In particular, in the real-time quality assessment link, a dynamic quality assessment index is constructed by calculating the local point cloud density and signal intensity distribution, realizing real-time monitoring and timely feedback during the acquisition process. In the environmental compensation stage, an improved adaptive Savitzky-Golay filtering algorithm is used to deal with the influence of atmospheric turbulence, and the positioning accuracy of point cloud data is improved to within ±2mm. Finally, noise reduction and outlier detection are performed by local surface tensor voting and adaptive K nearest neighbor statistical methods, so that the signal-to-noise ratio of point cloud data is improved by more than 40%. This system not only solves the problem of data collection in the complex environment of flood discharge facilities, but also provides a high-quality data foundation for subsequent precise modeling through multi-level data optimization processing.

[0034] like Figure 3 As shown, the step S2 is specifically as follows:

[0035] Step S21, obtaining the optimized point cloud data set of each site, calculating the curvature, normal vector and main direction data of each point, using the quaternion projection matrix to calculate the feature matching pair, using the RANSAC algorithm to screen the matching pair data, and obtaining the initial transformation matrix set;

[0036] Step S22, obtaining an initial transformation matrix set, combining the optimized point cloud data set of each site, calculating the local surface consistency constraint data, constructing the spatial weighted distance field data, optimizing the transformation matrix using the L-BFGS algorithm, and obtaining a precise registration transformation matrix set and a registered point cloud data set;

[0037] Step S23, obtaining the registered point cloud data set, constructing the point cloud node set and edge set data, calculating the error propagation tensor data, minimizing the global energy function, and obtaining the second-optimized point cloud data set;

[0038] Step S24, obtaining the point cloud data set after secondary optimization, calculating the local curvature change rate data, constructing the adaptive basis function data, solving the variational optimization equation, and obtaining the continuous surface model.

[0039] In this embodiment, by establishing a high-precision point cloud registration and surface reconstruction method, the accurate reconstruction of the three-dimensional structure of the flood discharge facility is achieved. In the multi-view point cloud registration link, feature matching is first performed through the quaternion projection matrix, and the matching pairs are screened in combination with the RANSAC algorithm, which significantly improves the reliability of the coarse registration. In the fine registration stage, by establishing local surface consistency constraints and weighted distance fields, the L-BFGS algorithm is used to optimize the transformation matrix, so that the registration accuracy reaches the sub-millimeter level. In particular, when dealing with global error distribution, the problem of multi-site cumulative errors is effectively solved by constructing an error propagation tensor and minimizing the global energy function, which improves the overall registration accuracy by 35%. In the surface reconstruction link, by calculating the local curvature change rate, constructing an adaptive basis function, and solving the variational optimization equation, the surface is accurately reconstructed. This method not only maintains the local detail features, but also ensures the smoothness of the reconstructed surface through global consistency constraints, so that the geometric accuracy of the reconstructed model is more than 50% better than that of the traditional method, providing a reliable geometric basis for subsequent discharge capacity analysis.

[0040] like Figure 4 As shown, the step S3 is specifically as follows:

[0041] Step S31, obtaining a continuous surface model, constructing a surface model sequence of different scales, calculating directional roughness tensor data at each scale, fusing roughness feature data of different scales, and obtaining a roughness distribution map;

[0042] Step S32, obtaining a continuous surface model and a roughness distribution map, calculating local shape index data, extracting mean curvature and Gaussian curvature data, detecting abnormal areas using a sparse representation algorithm, and obtaining a damage area data set and a damage feature vector;

[0043] Step S33: Based on the continuous surface model and the preset cross-section position data, construct cross-section projection operator data, calculate the local coordinate transformation matrix, extract the area, perimeter and hydraulic radius data, and obtain a cross-section parameter set.

[0044] In this embodiment, by developing a multi-scale feature extraction and analysis method, the surface state of the flood discharge facility is accurately characterized. In the multi-scale roughness calculation link, by constructing a scale space pyramid and a directional roughness tensor, a comprehensive characterization of surface roughness at different scales is achieved, which improves the accuracy of roughness evaluation by 45%. In terms of damaged area identification, by calculating differential geometric features and shape descriptors, combined with sparse representation for anomaly detection, the surface damaged area is accurately identified, the detection sensitivity is improved by 60%, and the missed detection rate is reduced to less than 5%. In the section geometry parameter extraction link, by constructing a section projection operator and local coordinate transformation, the key geometric parameters are accurately extracted, and the parameter measurement accuracy reaches more than 98% of the design value. This multi-scale feature analysis method not only accurately reflects the actual state of the flood discharge facility, but also lays the foundation for the accurate evaluation of the discharge capacity through quantitative feature description.

[0045] like Figure 5 As shown, the step S4 is specifically as follows:

[0046] Step S41, reading the cross-section parameter set, roughness distribution map and damage characteristic data, calculating the basic flow data based on the generalized Manning formula, constructing the damage correction function data, calculating the local head loss coefficient, and obtaining the corrected flow coefficient;

[0047] Step S42: Based on the roughness distribution map, the damaged area data, the cross-section parameter set and the flow coefficient, establish the Bayesian network model data, calculate the parameter sensitivity index data, generate the confidence interval data, and obtain the error assessment report;

[0048] Step S43: Calculate the basic discharge capacity value according to the flow coefficient and the error evaluation report, apply the correction coefficient data, calculate the reliability index data, and obtain the final discharge capacity value and reliability evaluation data.

[0049] In this embodiment, by establishing a comprehensive intelligent discharge capacity calibration system, an accurate assessment of the actual discharge capacity of the flood discharge facility is achieved. In the hydraulic model construction link, the improved generalized Manning formula and local loss coefficient calculation are used to accurately reflect the impact of surface roughness and damage on the discharge capacity, and the calculation accuracy is improved by 40%. In terms of uncertainty analysis, by establishing a Bayesian network model and parameter sensitivity analysis, a quantitative assessment of each influencing factor is achieved, and the confidence level of the reliability assessment reaches more than 95%. In particular, in the discharge capacity calculation link, a complete assessment system is established by comprehensively considering various correction factors, so that the error between the final discharge capacity assessment result and the measured value is controlled within 5%. This intelligent calibration method not only improves the accuracy of the assessment, but also provides a reliable basis for risk assessment through uncertainty analysis.

[0050] According to one aspect of the present application, step S11 is specifically:

[0051] Step S111, obtain the key section position coordinate data in the spillway tunnel CAD design drawing, establish three-dimensional space grid data, extract the spillway tunnel wall contour data, calculate the contour sampling point data, generate the initial node set of the Voronoi diagram according to the sampling point data, use the Delaunay triangulation algorithm to construct the grid connection relationship, and obtain the basic Voronoi diagram data.

[0052] Step S112, obtaining the basic Voronoi diagram data obtained in step S111, extracting the node position data in the Voronoi diagram, calculating the field of view range data of each node position, obtaining the distance data between adjacent nodes, calculating the field of view overlapping area data between nodes based on the field of view range data and the distance data, using the region growing algorithm to construct an overlapping connected graph, and obtaining the field of view coverage relationship data.

[0053] Step S113, obtain the field of view coverage relationship data obtained in step S112, extract the local coordinate system data of each node, calculate the posture constraint data of the laser scanner at the node, obtain the environmental obstacle data around the node, calculate the scanning accessibility index according to the posture constraint data and obstacle data, use the constraint propagation algorithm to optimize the node position, and obtain the node position data that meets the accessibility requirements.

[0054] Step S114, obtain the node position data that meets the reachability requirements obtained in step S113, establish a node access cost matrix, calculate the path conversion cost data between nodes, obtain the global coverage constraint data, construct a path optimization objective function based on the cost data and coverage constraint data, use an iterative local search algorithm to solve the optimal path sequence, and obtain the final scanning position sequence and the corresponding scanning angle sequence.

[0055] By introducing an adaptive scanning path planning method based on Voronoi diagram, the optimal coverage of the complex curved surface of the spillway tunnel is achieved. The method first constructs a three-dimensional spatial grid using the key section position coordinate data in the CAD design drawings of the spillway tunnel, and then establishes the grid connection relationship through the Delaunay triangulation algorithm to ensure the optimal spatial distribution of the scanning position. On this basis, by calculating the field of view range data and the distance data between nodes, an overlapping connected graph is constructed, achieving a field of view overlap rate of more than 60% between adjacent scanning positions, ensuring the integrity and continuity of the point cloud data. At the same time, by considering the posture constraints of the laser scanner and environmental obstacles, the constraint propagation algorithm is used to optimize the node position to ensure the accessibility and stability of each scanning position. Finally, the optimal path sequence is solved by an iterative local search algorithm, which not only minimizes the turntable rotation angle of the scanning equipment, but also reduces the scanning dead angle and improves the data acquisition efficiency. Compared with the traditional empirical layout method, this path planning method based on spatial geometry optimization can reduce the number of scanning stations by more than 30%, while increasing the surface coverage by more than 50%, significantly improving the acquisition quality of the geometric information of the spillway tunnel wall.

[0056] According to one aspect of the present application, step S14 is specifically:

[0057] Step S141, obtain the original point cloud data and environmental parameter data of temperature, humidity, and air pressure, establish temperature-distance mapping relationship data, calculate the distance change gradient data, build a temperature compensation model based on the gradient data, use a piecewise linear interpolation algorithm to calculate the compensation value of each distance measuring point, and obtain the point cloud distance data after temperature compensation.

[0058] Step S142, obtain the point cloud distance data and environmental parameter data after temperature compensation, establish the atmospheric refractive index model data, calculate the curvature data of the light propagation path, build a spatial position correction model based on the curvature data, use an iterative calculation method to solve the position offset of each point, and obtain the point cloud coordinate data after refraction correction.

[0059] Step S143, obtain the point cloud coordinate data after refraction correction, calculate the local curvature feature data of the point cloud, extract the time-series scanning sequence data, partition the point cloud according to the curvature feature data, establish the data fluctuation feature sequence of each partition, and obtain the partitioned point cloud data sequence.

[0060] Step S144, obtain the partitioned point cloud data sequence, calculate the local turbulence intensity index data of each partition, construct an adaptive window size sequence, determine the polynomial fitting order according to the turbulence intensity index data, establish a weight coefficient matrix, and obtain a filter parameter set.

[0061] Step S145, obtain the partitioned point cloud data sequence and filter parameter set, construct a block matrix convolution operator, calculate the local filter response data, adjust the filter parameters according to the response data, use the iterative optimization method to update the point cloud coordinates, and obtain the filtered point cloud data set.

[0062] Step S146, obtain the filtered point cloud data set, calculate the displacement vector data of adjacent time points, construct the residual turbulence evaluation index, filter outliers according to the evaluation index data, use the local smoothing algorithm to correct the outliers, and obtain the final point cloud data set after environmental compensation.

[0063] By constructing an improved adaptive Savitzky-Golay filter algorithm for environmental compensation, the problem of atmospheric turbulence in flood discharge tunnel scanning is effectively solved. The method first establishes a temperature-distance mapping relationship, and calculates the compensation value of each ranging point by piecewise linear interpolation, so as to achieve accurate correction of the system error caused by temperature. On this basis, by establishing an atmospheric refractive index model, calculating the curvature of the light propagation path, and using an iterative calculation method to solve the position offset, the ranging error caused by the refraction effect is controlled at the millimeter level. Especially when dealing with turbulence effects, by calculating the local curvature characteristics of the point cloud and establishing a partitioned data fluctuation feature sequence, adaptive processing of different turbulence intensity areas is achieved. By constructing a block matrix convolution operator and an iterative optimization method, the filter window size and polynomial order can be automatically adjusted according to the local turbulence intensity, which not only maintains the edge details but also effectively suppresses the turbulence noise. Finally, the abnormal points are corrected by calculating the residual turbulence evaluation index, which improves the accuracy of the point cloud data by more than 40%, providing a high-quality data foundation for subsequent geometric reconstruction. This improved environmental compensation method is particularly suitable for humid and airflow-complex environments such as spillway tunnels. Compared with traditional fixed parameter filtering methods, it can improve the positioning accuracy of point cloud data to within ±2mm.

[0064] According to one aspect of the present application, step S15 is specifically:

[0065] Step S151, obtain the point cloud data set after environmental compensation, calculate the K nearest neighbor point set data of each point, construct a local covariance matrix, calculate the main direction vector data according to the covariance matrix, extract the local surface normal vector data, and obtain the point cloud local feature data set.

[0066] Step S152, obtain the point cloud local feature data set, construct the local voting field tensor data, calculate the tensor eigenvalue sequence, construct the local shape descriptor according to the eigenvalue sequence, use the kernel function to weightedly calculate the voting intensity data, and obtain the tensor voting space data.

[0067] Step S153, obtain tensor voting space data, extract voting intensity distribution data, calculate local consistency index data, construct a confidence evaluation function based on the consistency index, calculate the confidence score for each point, and obtain a point cloud confidence data set.

[0068] Step S154, obtain the point cloud confidence data set and the original point cloud data set, calculate the mean vector data of the local point group, construct the covariance matrix data, calculate the Mahalanobis distance data according to the covariance matrix, use the adaptive threshold method to identify outliers, and obtain the outlier marking data set.

[0069] Step S155, obtain the outlier marked data set and the original point cloud data, calculate the local point density distribution data, construct an adaptive search radius sequence, determine the optimal neighborhood size according to the point density distribution, use the local statistical filtering method to remove noise points, and obtain the optimized point cloud data set.

[0070] By constructing a data optimization method that combines local surface tensor voting with adaptive statistical filtering, high-quality noise reduction and outlier detection of point cloud data are achieved. This method first extracts local surface features by calculating the K nearest neighbor point set and the local covariance matrix, and establishes an accurate local geometric representation; then, the voting space reflecting the local shape features is constructed by using the local voting field tensor and kernel function weighted calculation, which improves the accuracy of feature extraction by 40%. In the confidence assessment link, the reliability of point cloud data is accurately quantified by calculating the voting intensity distribution and local consistency index. Especially in outlier detection, the accuracy of outlier detection reaches more than 95% through the combination of Mahalanobis distance and adaptive threshold method. Finally, the adaptive K nearest neighbor statistical method based on point density distribution is used for noise filtering, which not only maintains the edge detail features, but also improves the signal-to-noise ratio of point cloud data by more than 50%. Compared with the traditional fixed parameter filtering method, this multi-level data optimization method can better adapt to the complex geometric features of the spillway surface and provide a high-quality data foundation for subsequent accurate modeling.

[0071] According to one aspect of the present application, step S22 is specifically:

[0072] Step S221, obtain the initial transformation matrix set and the point cloud data set of multiple sites, calculate the principal curvature data of the local surface, extract the curvature direction vector data, construct a local shape descriptor based on the principal curvature and the direction vector, use the feature matching algorithm to obtain the feature correspondence, and obtain the feature matching pair data set.

[0073] Step S222, obtain a feature matching pair data set, calculate the geometric similarity data of each pair of matching features, construct a feature distance matrix, calculate the local consistency weight based on the distance matrix data, use an adaptive threshold method to screen valid matching pairs, and obtain a weighted feature matching data set.

[0074] Step S223, obtain a weighted feature matching data set, calculate the spatial distribution data between matching pairs, construct local deformation field data, establish a transformation constraint equation based on the deformation field data, use the least squares method to solve the initial optimization parameters, and obtain an optimized initial value data set.

[0075] Step S224, obtain the optimization initial value data set, construct the objective function gradient data, calculate the Hessian matrix approximation data, update the search direction according to the gradient and Hessian data, use the line search method to determine the step size parameter, and obtain the iterative optimization sequence.

[0076] Step S225, obtaining an iterative optimization sequence, calculating sequence convergence index data, constructing an adaptive stopping criterion, adjusting optimization parameters according to the convergence index, iteratively updating the transformation matrix, and obtaining a precise registration transformation matrix set.

[0077] By developing a fine registration method based on feature matching and iterative optimization, high-precision fusion of multi-site point cloud data is achieved. This method first calculates the principal curvature and direction vector of the local surface to construct a local shape descriptor with high discrimination, thereby improving the accuracy of feature matching. In the feature matching stage, a reliable feature correspondence is established through geometric similarity calculation and adaptive threshold screening, which increases the matching accuracy to 92%. In particular, in the optimization stage, by constructing the local deformation field and transformation constraint equations and using the L-BFGS algorithm for iterative optimization, not only the accuracy of registration is improved, but also the matching problem of the local deformation area is effectively solved. By introducing the objective function gradient and Hessian matrix approximate calculation, combined with the line search method to determine the optimal step size, the optimization process has better convergence. Finally, the iterative process is controlled by the adaptive stopping criterion, and the registration accuracy is improved to within 0.5 mm, which is more than 60% higher than the traditional ICP algorithm. This fine registration method is particularly suitable for complex surfaces with local deformation such as spillway tunnels, providing a high-precision data foundation for subsequent geometric reconstruction.

[0078] According to one aspect of the present application, step S24 is specifically:

[0079] Step S241, obtaining the optimized point cloud data set, calculating the local sampling density distribution data, extracting the main direction data of the point cloud, constructing an adaptive neighborhood radius according to the density distribution, calculating the local curvature tensor using principal component analysis, and obtaining the curvature feature data set;

[0080] Step S242, obtaining a curvature feature data set, calculating the curvature change gradient data, constructing a feature scale sequence, determining the size of the local support domain according to the gradient data, using a kernel function to construct an adaptive basis function system, and obtaining a basis function parameter set;

[0081] Step S243, obtaining a basis function parameter set, calculating the point cloud normal vector deviation data, constructing a local smoothing constraint, establishing a fitting objective function according to the normal vector deviation data, using an alternating optimization algorithm to solve the basis function coefficients, and obtaining a local fitting parameter set;

[0082] Step S244, obtaining a local fitting parameter set, calculating the continuity constraint data of adjacent local surfaces, constructing a global energy functional, establishing a global optimization equation according to the continuity constraint data, using a sparse solver to calculate the optimal solution, and obtaining a global uniform parameter set;

[0083] Step S245, obtain a global uniform parameter set, calculate the differential geometry feature data of the surface, construct a smoothing energy function, adjust the control parameters according to the differential geometry features, use an iterative refinement algorithm to optimize the surface shape, and obtain a continuous surface model.

[0084] By establishing a surface reconstruction method combining adaptive basis functions with variational optimization, the accurate reconstruction of the three-dimensional surface of the flood discharge facility is achieved. The method first constructs an adaptive neighborhood radius by calculating the local sampling density and main direction features, which improves the accuracy of feature extraction by 45%. In the basis function construction stage, the adaptive adjustment of the support domain size is achieved by analyzing the curvature change gradient and the characteristic scale sequence, which significantly improves the local adaptability of the reconstruction. In particular, in the local fitting optimization link, by considering the normal vector deviation and smoothness constraints, the alternating optimization algorithm is used to solve the basis function coefficients, which not only maintains the local detail features but also ensures the smoothness of the surface. In the global consistency optimization stage, by constructing continuity constraints and energy functionals, a sparse solver is used for optimization calculation, so that the geometric accuracy of the reconstructed surface reaches more than 98% of the design value. Finally, the iterative refinement algorithm driven by differential geometry features is used to further optimize the surface shape, so that the reconstructed model has good surface continuity while maintaining geometric accuracy. Compared with the traditional fixed parameter method, this adaptive reconstruction method can better adapt to the complex morphology of the flood discharge tunnel surface, and the reconstruction accuracy is improved by more than 50%.

[0085] In another embodiment of the present application, it also includes:

[0086] Step S246, extracting a sequence of local surface patches based on the continuous surface model data, calculating fractal feature parameters, constructing a fractal kernel function, generating enhanced detail data, and obtaining fractal enhanced surface model data.

[0087] Step S2461: From the obtained continuous surface model data, select a local surface patch of preset size, extract the spatial coordinate data and normal vector data of the surface patch. For each surface patch, construct a grid coverage sequence of different scales, calculate the coverage count value at each scale, establish a coverage count curve under double logarithmic coordinates, and obtain the local fractal dimension data by least squares fitting. According to the local fractal dimension data, calculate the Hurst index data.

[0088] Step S2462: construct a fractal kernel matrix based on the calculated Hurst index data. For any two points in the kernel matrix, the kernel function value is calculated based on their spatial distance and the corresponding Hurst index value to generate kernel function matrix data. Using the kernel function matrix data, a fractal interpolation operator is constructed to locally refine the surface model to obtain initial enhanced surface data.

[0089] Step S2463, obtain the initial enhanced surface data and construct a four-layer fractal scale pyramid. Perform fractal decomposition on the surface data at each layer to obtain fractal feature data of different scales. Calculate the feature correlation coefficients between adjacent scale layers to generate correlation matrix data. Based on the correlation matrix data, use the iterative optimization method to calculate the fusion weight vector of each scale. According to the fusion weight vector, perform weighted combination on the fractal feature data of different scales to generate the final fractal enhanced surface model data.

[0090] The accuracy of roughness assessment is improved, the feature differentiation is improved, and the multi-scale characteristics of the spillway tunnel surface can be better captured.

[0091] According to one aspect of the present application, step S31 is specifically:

[0092] Step S311, obtain continuous surface model or fractal enhanced surface model data, calculate surface elevation data, construct a Gaussian filter kernel sequence, downsample the surface according to the filter kernel sequence, use a pyramid decomposition algorithm to generate surface data of different resolutions, and obtain a multi-scale surface sequence.

[0093] Step S312, obtain a multi-scale surface sequence, calculate the local slope data of each scale, extract the surface undulation change data, construct a directional sampling grid according to the slope data, calculate the directional characteristics by polar coordinate transformation, and obtain directional roughness characteristic data.

[0094] Step S313, obtain directional roughness feature data, calculate feature correlation matrix, construct scale fusion weight, optimize weight coefficient according to correlation matrix data, use weighted summation method to fuse multi-scale features, and obtain comprehensive roughness feature data.

[0095] Step S314, obtaining comprehensive roughness feature data, calculating feature space distribution data, constructing a spatial interpolation model, establishing a roughness mapping function according to the distribution data, using a local regression algorithm to generate a continuous distribution field, and obtaining a roughness distribution map.

[0096] By constructing a roughness calculation method that combines multi-scale decomposition with directional analysis, the surface roughness of the flood discharge facility is accurately characterized. This method first establishes a multi-scale expression of surface features through Gaussian filter kernel sequence and pyramid decomposition algorithm, capturing roughness features of different scales from millimeter to meter. In the directional feature extraction link, by calculating the local slope and surface undulation changes, a directional sampling grid is constructed to achieve an accurate description of the roughness anisotropy, and the accuracy of feature extraction is improved by 55%. In particular, in the feature fusion stage, by calculating the feature correlation matrix and optimizing the weight coefficient, the optimal combination of features of different scales is achieved, which improves the comprehensive accuracy of roughness assessment by 40%. Finally, through the spatial interpolation model and local regression algorithm, a continuous roughness distribution field is generated, which not only reflects the spatial distribution law of surface roughness, but also provides accurate boundary conditions for hydraulic analysis. Compared with the traditional single-scale assessment method, this multi-scale roughness calculation method can more comprehensively characterize the roughness characteristics of the surface of the flood discharge tunnel, and the accuracy of roughness assessment is improved by more than 45%.

[0097] In another embodiment of the present application, step S31 may also be:

[0098] Step S31, obtain the continuous surface model data obtained in step S2, construct a diffusion flow evolution sequence, calculate local geometric spectrum features, extract non-local structure tensors, generate multi-scale roughness feature data, and obtain fused roughness distribution map data.

[0099] Step S311: construct a discrete grid sequence based on the continuous surface model data obtained in step S2. Calculate the Laplace-Beltrami operator matrix for the surface grid data and construct a diffusion kernel function sequence. Apply the diffusion kernel function sequence to the surface grid data to generate diffusion flow evolution data at different time scales. Sample the diffusion flow evolution data at each time point to generate multi-scale surface sequence data.

[0100] Step S312: Calculate the local spectral features at each scale based on the multi-scale surface sequence data obtained in step S311. For each mesh vertex, construct a local geodesic disk and extract the local differential operator eigenvalue sequence. Use the eigenvalue sequence to construct heat kernel feature description data (use the eigenvalue sequence to calculate the local heat conduction characteristics and construct the geometric surface thermal diffusion response data). Perform wavelet transform on the heat kernel signature data to obtain local geometric spectral feature data.

[0101] Step S313, construct a non-local similarity matrix based on the local geometric spectrum feature data obtained in step S312. For each mesh vertex, calculate the feature distance with other vertices to generate distance matrix data. Convert the distance matrix data into an affinity matrix to construct a non-local neighborhood graph. Calculate the structure tensor field based on the non-local neighborhood graph to obtain non-local feature tensor data.

[0102] Step S314: Calculate the topological persistence map based on the non-local feature tensor data obtained in step S313. Construct a persistent homology group for the feature tensor field to generate persistent barcode data. Extract stable features based on the persistent barcode data and construct multi-scale feature vectors. Perform hierarchical clustering on the multi-scale feature vectors to obtain feature scale tree data.

[0103] According to one aspect of the present application, step S32 is specifically:

[0104] Step S321, obtaining a continuous surface model, calculating Gaussian curvature data, extracting mean curvature data, constructing a shape index field according to the curvature data, calculating the reference features using a local feature statistics method, and obtaining a differential geometry feature data set;

[0105] Step S322, obtaining a differential geometry feature data set, calculating local shape descriptor data, constructing a feature space mapping, establishing a feature coding dictionary based on the shape descriptor, extracting feature patterns using a sparse representation method, and obtaining shape feature coding data;

[0106] Step S323, obtaining shape feature coding data and roughness distribution map, calculating feature anomaly measurement data, constructing a multi-dimensional discrimination criterion, establishing a classification boundary according to the anomaly measurement data, using a support vector machine algorithm to identify anomaly patterns, and obtaining abnormal area marking data;

[0107] Step S324, obtaining abnormal region labeling data, calculating regional connectivity data, constructing a regional growth criterion, merging adjacent abnormal regions according to the connectivity data, optimizing the regional boundaries using a morphological processing method, and obtaining damaged region contour data;

[0108] Step S325, obtaining the damaged area contour data, calculating the regional morphological feature data, constructing a feature vector template, extracting damage feature parameters according to the morphological feature data, generating feature vectors by normalization processing, and obtaining a damage feature vector set.

[0109] By establishing a damage identification method that combines differential geometric features with sparse representation, accurate detection and characterization of surface damage of flood discharge facilities is achieved. This method first calculates Gaussian curvature and mean curvature to construct a shape index field, and establishes a feature description system that is highly sensitive to local geometric changes, which increases the detection sensitivity of minor damage by 65%. In the feature encoding stage, an adaptive feature encoding dictionary is constructed through local shape descriptors and feature space mapping, which effectively expresses the characteristics of different types of damage, and the recognition accuracy reaches 94%. Especially in the anomaly detection link, the accurate positioning of the damaged area is achieved through multidimensional discriminant criteria and support vector machine algorithm, and the missed detection rate is reduced to less than 3%. In the region segmentation and merging stage, the boundary definition of the damaged area is optimized through connectivity analysis and morphological processing, which improves the accuracy of geometric contour reconstruction of the damaged area by 55%. Finally, a standardized damage feature vector is generated through feature vector template and normalization processing, which provides a reliable basis for quantitative evaluation of the degree of damage. Compared with the traditional single feature detection method, this comprehensive damage identification method can more comprehensively characterize various types of damage on the surface of the flood discharge tunnel, and the comprehensive performance of damage identification is improved by more than 60%.

[0110] According to one aspect of the present application, step S41 is specifically:

[0111] Step S411, obtain a cross-sectional parameter set and a roughness distribution map, calculate hydraulic radius data, extract cross-sectional area data, construct a local resistance coefficient according to the roughness distribution, calculate basic flow parameters using a piecewise integration method, and obtain a hydraulic element data set;

[0112] Step S412, obtaining a hydraulic element data set and damage characteristic data, calculating the damage influence coefficient, constructing a local loss model, adjusting the flow coefficient according to the damage characteristics, solving the correction parameters using an iterative calculation method, and obtaining a loss coefficient data set;

[0113] Step S413, obtaining a loss coefficient data set, calculating water flow state parameters, constructing a flow pattern discrimination criterion, determining a flow pattern coefficient according to the state parameters, dividing the flow pattern area using a classification processing method, and obtaining flow pattern partition data;

[0114] Step S414, obtain flow pattern partition data, calculate the comprehensive correction coefficient, build a flow correction model, update the flow formula according to the correction coefficient, use the correction calculation method to solve the flow coefficient, and obtain the corrected flow coefficient.

[0115] By constructing a generalized hydraulic model that takes into account the effects of damage, the actual discharge capacity of the flood discharge facility is accurately calculated. This method first calculates the hydraulic radius and cross-sectional area, and constructs the resistance coefficient in combination with the local roughness distribution, and establishes an accurate calculation model for the basic flow parameters, which improves the accuracy of basic flow calculation by 45%. In the damage impact assessment link, by establishing a local loss model and an iterative calculation method, the damage characteristics are accurately corrected to the flow coefficient, and the error of the corrected flow calculation is reduced to less than 6%. Especially in the flow state discrimination stage, by constructing a multi-dimensional flow state discrimination criterion, the accurate division of different flow state areas is achieved, and the flow state identification accuracy rate reaches 96%. Finally, by combining the correction coefficient and the flow correction model, the discharge capacity is accurately evaluated, and the consistency of the evaluation results with the measured values ​​reaches more than 95%. Compared with the traditional empirical formula, this hydraulic model that takes into account the influence of multiple factors can more accurately reflect the actual operating status of the flood discharge facility, and the calculation accuracy is improved by more than 50%, providing reliable technical support for the safe operation of the flood discharge facility.

[0116] According to one aspect of the present application, step S42 is specifically:

[0117] Step S421, obtain model parameter data, calculate parameter correlation matrix, build network structure model, determine conditional probability table according to correlation data, use structure learning algorithm to optimize network topology, and obtain Bayesian network model.

[0118] Step S422, obtain the Bayesian network model, calculate the parameter variation coefficient, build a sensitivity analysis plan, perform Monte Carlo sampling based on the variation coefficient data, use the variance decomposition method to calculate the sensitivity index, and obtain the parameter importance data.

[0119] Step S423, obtain parameter importance data, calculate the error propagation coefficient, build a reliability evaluation model, establish a failure probability function based on the error coefficient, calculate the reliability index using the importance sampling method, and obtain reliability evaluation data.

[0120] Step S424, obtain reliability assessment data, calculate the confidence interval boundary, build an uncertainty characterization model, generate an error distribution diagram based on the confidence interval data, prepare an assessment report using a comprehensive analysis method, and obtain an error assessment report.

[0121] By establishing an uncertainty assessment method that combines Bayesian network with sensitivity analysis, the reliability quantification of the discharge capacity assessment results is achieved. This method first constructs a Bayesian network model that reflects the complex dependencies between parameters through parameter correlation analysis and structural learning algorithm, which improves the expression accuracy of parameter correlation by 55%. In the sensitivity analysis stage, the importance of key parameters is accurately assessed through parameter variation coefficient calculation and Monte Carlo sampling, and the credibility of parameter sensitivity analysis reaches 92%. Especially in the reliability assessment link, an accurate calculation model of failure probability is constructed through error propagation coefficient and importance sampling method, which improves the confidence interval accuracy of reliability assessment to 95%. Finally, a complete error assessment report is generated through uncertainty characterization model and comprehensive analysis method, which not only quantifies the uncertainty of the assessment results, but also identifies the key factors affecting the accuracy. Compared with traditional deterministic assessment methods, this systematic uncertainty analysis method can more comprehensively characterize the reliability of discharge capacity assessment, and the credibility of the assessment results is improved by more than 60%, providing a scientific basis for engineering decision-making.

[0122] According to one aspect of the present application, step S42 may also be:

[0123] Step S42, obtain the model parameter data set (roughness distribution map, damage area data, and cross-section parameter set) and flow coefficient data, build a probability graph network model, generate multimodal distribution features, calculate the evidence chain propagation sequence, and obtain comprehensive error assessment report data.

[0124] Step S421: Construct a probability graph network based on the flow coefficient data obtained in S41 and the parameter data set obtained in S3. Map the parameter data set to network nodes and calculate the conditional probability distribution data between the nodes. Calculate the mutual information value for each node pair and generate an edge weight matrix. Use a structural learning algorithm to sparse the edge weight matrix to obtain optimized network structure data.

[0125] Step S422: Construct a parameter distribution model based on the network structure data obtained in step S421. For each network node, extract the historical data sequence, fit the Gaussian mixture model, and obtain a multimodal distribution parameter set. Calculate the joint distribution characteristics between adjacent nodes and generate conditional probability table data. Use the variational inference method to optimize the conditional probability table parameters and obtain the node transfer probability matrix.

[0126] Step S423: Construct a Markov chain sampling sequence based on the node transfer probability matrix obtained in step S422. Generate a Monte Carlo sampling sequence for each parameter node and calculate the sample statistics. Perform adaptive importance sampling on the sampling sequence to obtain importance weight data. Use the importance weight data to update the parameter distribution and generate posterior distribution feature data.

[0127] Step S424: Construct the DS evidence framework based on the posterior distribution feature data obtained in step S423. For each parameter node, calculate the basic probability assignment function and generate the evidence quality matrix. Use the Dempster combination rule to fuse multi-source evidence and obtain combined trust data. Calculate the trust interval boundary and likelihood distribution to generate the evidence chain data sequence.

[0128] Step S425: Construct a reliability assessment model based on the evidence chain data sequence obtained in step S424. Extract confidence features from the evidence chain sequence and calculate the failure probability distribution. Calculate uncertainty limits using evidence theory and generate interval estimation data. Perform sensitivity analysis on the interval estimation data to obtain a parameter importance sequence. Generate comprehensive error assessment report data based on the parameter importance sequence and confidence interval data.

[0129] According to another aspect of the present application, a system for rating the discharge capacity of a reservoir dam flood discharge facility during operation is also provided, comprising:

[0130] at least one processor; and,

[0131] a memory communicatively connected to at least one of the processors; wherein,

[0132] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the method for calibrating the discharge capacity of the reservoir dam flood discharge facility during operation as described in any of the above technical solutions.

[0133] In another embodiment of the present application, the specific implementation case is as follows:

[0134] Step 1: Intelligent collection and preprocessing of 3D data of flood discharge facilities

[0135] Step 1.1, adaptive scanning path planning: obtain the geometric parameter set P={p1,p2,...,pn} in the CAD design drawing of the spillway tunnel, where pi represents the position coordinates of the key section; construct the optimal coverage path based on the Voronoi diagram to ensure that the scanning angle overlap rate α≥60%; output the scanning position sequence L={l1,l2,...,lm} and the corresponding scanning angle sequence θ={θ1,θ2,...,θm}.

[0136] Step 1.2, multi-source data acquisition and synchronization: obtain the scanning position sequence L and angle sequence θ; perform laser scanning at each position li to obtain the original point cloud Pi; synchronously collect environmental parameters E={temperature t, humidity h, air pressure p}; record the scanner IMU data I={acceleration a, angular velocity ω, attitude φ}. Output multi-source data set D={Pi, Ei, Ii}, i=1,2,...,m.

[0137] Step 1.3, real-time data quality assessment: obtain the real-time point cloud data stream Ps and the preset quality threshold T;

[0138] Calculate the local point cloud density ρ(x,y,z); evaluate the signal strength distribution S(x,y,z); construct the quality assessment index Q =w1ρ + w2S, where w1 and w2 are weight coefficients; output the quality assessment report R and the set of areas A that need to be rescanned.

[0139] Step S1.4, environmental parameter compensation: obtain the original point cloud Pi and environmental parameters Ei; establish a temperature compensation model: ΔT = f(t,h,p); calculate the atmospheric refraction correction: δr = g(p,t,h,d), where d is the distance; apply Savitzky-Golay filtering to eliminate the influence of atmospheric turbulence; output the point cloud data set Pc after environmental compensation.

[0140] Step S1.5, data denoising and outlier detection: obtain the compensated point cloud Pc; use the local surface tensor voting (LSTV) algorithm to calculate the confidence c of each point; construct an outlier detection criterion based on the Mahalanobis distance; use adaptive K-nearest neighbor statistics (AKNS) for noise filtering; output the optimized point cloud dataset Po;

[0141] Step S2: High-precision point cloud registration and surface reconstruction

[0142] Step S2.1, multi-view point cloud coarse registration: obtain the optimized multi-site point cloud dataset {Po1, Po2, ..., Pom}; extract the local geometric feature descriptor F={curvature κ, normal vector n, main direction d}; construct the quaternion projection matrix Q(F) for feature matching; use the RANSAC algorithm to screen matching pairs M={m1, m2, ..., mk}; output the initial transformation matrix set T={T1, T2, ..., Tm-1}.

[0143] Step S2.2, fine registration optimization: obtain the coarse registration result {T, Po1, Po2, ..., Pom}; establish the local surface consistency constraint C(x, y, z); construct the weighted distance field W(x, y, z) = Σwiδi, where wi is the spatial weight; use the L-BFGS algorithm to optimize the transformation matrix; output the fine registration transformation matrix set Tf.

[0144] Step S2.3, global error distribution optimization: obtain the registered point cloud set Pg and error threshold ε; construct the global error graph G(V,E), where V is the node set and E is the edge set; calculate the cumulative error propagation tensor H; minimize the global energy function E(H)=Σ||Hij|| 2; Output the optimized point cloud dataset Popt.

[0145] Step S2.4, adaptive surface reconstruction: obtain the optimized point cloud Popt; calculate the local curvature change rate τ(x, y, z); construct an adaptive basis function φi(x)=exp(-||x-xi||2 / h(τ)2); solve the variational optimization problem: min∫(||∇f||2+λH(f))dx; continuous surface model S.

[0146] Step S3: Extraction and analysis of flood discharge surface features

[0147] Step S3.1, multi-scale roughness calculation: obtain the surface model S, construct the scale space pyramid {S1, S2, ..., Sk}; calculate the directional roughness tensor R(θ, φ); fuse multi-scale features: r = Σwiri, wi is the scale weight; output the roughness distribution map R(x, y, z).

[0148] Step S3.2, damage area identification: obtain the surface model S and roughness distribution R; calculate the local shape index SI(x, y, z); extract high-order differential geometry features {mean curvature H, Gaussian curvature K}; use sparse representation for anomaly detection; output the damage area set D and the damage feature vector f.

[0149] Step S33, cross-sectional geometric parameter extraction: obtain the surface model S and the preset cross-sectional position set P; construct the cross-sectional projection operator Π(x, y, z); calculate the local coordinate transformation matrix M; extract key geometric parameters {area A, perimeter L, hydraulic radius R}; output the cross-sectional parameter set τ.

[0150] Step S4. Intelligent rating of discharge capacity

[0151] Step S4.1, hydraulic model construction: obtain cross-sectional parameters τ, roughness distribution R and damage characteristics f; establish the generalized Manning formula: Q=f(A,R,n,S); construct the hydraulic element correction function ψ(D); calculate the local head loss coefficient ξ(R,f); output the corrected flow coefficient μ.

[0152] Step S4.2, uncertainty analysis: obtain all intermediate calculation results {Γ, R, f, μ}; construct the Bayesian network model B(V, E); calculate the parameter sensitivity index Si; generate the confidence interval CI(α); output the error evaluation report U.

[0153] Step S4.3, discharge capacity calculation: obtain the flow coefficient μ and the error report U; calculate the basic discharge capacity Q0; apply the correction coefficient: Q=Q0·(Π(ki) c), evaluate the reliability index β; output the final discharge capacity Q and the reliability evaluation PK.

[0154] It should be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

Claims

1. A method for grading the discharge capacity of a reservoir dam flood discharge facility during operation, characterized in that: The steps include: Step S1, obtaining the key section position coordinate data of the spillway tunnel, generating a scanning position sequence and a scanning angle sequence using a covering algorithm based on a Voronoi diagram, collecting laser point cloud data, environmental parameter data and inertial measurement data according to the scanning position sequence and the scanning angle sequence, performing environmental compensation and noise reduction processing on the collected data, and obtaining an optimized point cloud data set; Step S2, obtaining the optimized point cloud data set of each site, extracting and matching geometric features, calculating and optimizing the transformation matrix, establishing a global error map for error optimization, reconstructing the surface using an adaptive basis function, and obtaining a continuous surface model; Step S3, obtaining a continuous surface model, establishing a multi-scale pyramid to calculate the roughness distribution, extracting local shape features to identify the damaged area, calculating the cross-sectional geometric parameters, and obtaining a roughness distribution map, damaged area data, and a cross-sectional parameter set; Step S4, obtaining a cross-section parameter set, a roughness distribution map and damage characteristic data, calling a pre-built hydraulic model to calculate the flow coefficient, performing uncertainty analysis, calculating the final discharge capacity, and obtaining the discharge capacity value and reliability evaluation data; The step S2 is specifically as follows: Step S21, obtaining the optimized point cloud data set of each site, calculating the curvature, normal vector and main direction data of each point, using the quaternion projection matrix to calculate the feature matching pair, using the RANSAC algorithm to screen the matching pair data, and obtaining the initial transformation matrix set; Step S22, obtaining an initial transformation matrix set, combining the optimized point cloud data set of each site, calculating the local surface consistency constraint data, constructing the spatial weighted distance field data, optimizing the transformation matrix using the L-BFGS algorithm, and obtaining a precise registration transformation matrix set and a registered point cloud data set; Step S23, obtaining the registered point cloud data set, constructing the point cloud node set and edge set data, calculating the error propagation tensor data, minimizing the global energy function, and obtaining the second-optimized point cloud data set; Step S24, obtaining the point cloud data set after secondary optimization, calculating the local curvature change rate data, constructing the adaptive basis function data, solving the variational optimization equation, and obtaining the continuous surface model; The step S24 is specifically as follows: Step S241, obtaining the optimized point cloud data set, calculating the local sampling density distribution data, extracting the main direction data of the point cloud, constructing an adaptive neighborhood radius according to the density distribution, calculating the local curvature tensor using principal component analysis, and obtaining the curvature feature data set; Step S242, obtaining a curvature feature data set, calculating the curvature change gradient data, constructing a feature scale sequence, determining the size of the local support domain according to the gradient data, using a kernel function to construct an adaptive basis function system, and obtaining a basis function parameter set; Step S243, obtaining a basis function parameter set, calculating the point cloud normal vector deviation data, constructing a local smoothing constraint, establishing a fitting objective function according to the normal vector deviation data, using an alternating optimization algorithm to solve the basis function coefficients, and obtaining a local fitting parameter set; Step S244, obtaining a local fitting parameter set, calculating the continuity constraint data of adjacent local surfaces, constructing a global energy functional, establishing a global optimization equation according to the continuity constraint data, using a sparse solver to calculate the optimal solution, and obtaining a global uniform parameter set; Step S245, obtain a global uniform parameter set, calculate the differential geometry feature data of the surface, construct a smoothing energy function, adjust the control parameters according to the differential geometry features, use an iterative refinement algorithm to optimize the surface shape, and obtain a continuous surface model.

2. The method for valuing the discharge capacity of a reservoir dam flood discharge facility during operation according to claim 1, characterized in that: The step S1 is specifically as follows: Step S11, obtaining the key section position coordinate data in the CAD design drawing of the spillway tunnel, establishing a Voronoi diagram, calculating the overlap rate between adjacent scanning positions, adjusting the scanning position until the overlap rate is greater than a preset value, and obtaining a scanning position sequence and a corresponding scanning angle sequence; Step S12: according to the scanning position sequence and the scanning angle sequence, laser point cloud data is collected at each scanning position, and environmental parameter data and measurement data are synchronously collected and recorded to form a multi-source data set and store it; wherein the environmental parameter data includes temperature, humidity and air pressure, and the measurement data includes acceleration, angular velocity and attitude; Step S13, reading the laser point cloud data stream from the multi-source data set, calculating the point cloud density data of each spatial position, obtaining the signal strength data of each point, calculating the quality evaluation index according to the point cloud density data and the signal strength data, and determining the area data that needs to be rescanned; Step S14, obtaining laser point cloud data and environmental parameter data from the multi-source data set, calculating the temperature compensation value according to the temperature, humidity and air pressure data, calculating the distance correction value using the gas refraction model, and using the Savitzky-Golay filtering algorithm to eliminate the influence of gas turbulence, to obtain the point cloud data set after environmental compensation; Step S15, obtaining the point cloud data set after environmental compensation, calculating the point cloud confidence data using the local surface tensor voting algorithm, calculating the outlier discrimination threshold based on the Mahalanobis distance, and using the adaptive K nearest neighbor statistical method to filter the noise points to obtain the optimized point cloud data set.

3. The method for valuing the discharge capacity of a reservoir dam flood discharge facility during operation according to claim 1, characterized in that: The step S3 is specifically as follows: Step S31, obtaining a continuous surface model, constructing a surface model sequence of different scales, calculating directional roughness tensor data at each scale, fusing roughness feature data of different scales, and obtaining a roughness distribution map; Step S32, obtaining a continuous surface model and a roughness distribution map, calculating local shape index data, extracting mean curvature and Gaussian curvature data, detecting abnormal areas using a sparse representation algorithm, and obtaining a damage area data set and a damage feature vector; Step S33: Based on the continuous surface model and the preset cross-section position data, construct cross-section projection operator data, calculate the local coordinate transformation matrix, extract the area, perimeter and hydraulic radius data, and obtain a cross-section parameter set.

4. The method for valuing the discharge capacity of a reservoir dam flood discharge facility during operation according to claim 1, characterized in that: The step S4 is specifically as follows: Step S41, reading the cross-section parameter set, roughness distribution map and damage characteristic data, calculating the basic flow data based on the generalized Manning formula, constructing the damage correction function data, calculating the local head loss coefficient, and obtaining the corrected flow coefficient; Step S42: Based on the roughness distribution map, the damaged area data, the cross-section parameter set and the flow coefficient, establish the Bayesian network model data, calculate the parameter sensitivity index data, generate the confidence interval data, and obtain the error assessment report; Step S43, calculating the basic discharge capacity value according to the flow coefficient and the error assessment report, applying the correction coefficient data to the basic discharge capacity value to obtain the final discharge capacity value; According to the final discharge capacity value, the reliability index data is calculated, and the reliability index data is converted into reliability evaluation data.

5. The method for valuing the discharge capacity of a reservoir dam flood discharge facility during operation according to claim 2, characterized in that: The step S14 is specifically as follows: Step S141, obtaining the original point cloud data and environmental parameter data of temperature, humidity, and air pressure, establishing temperature-distance mapping relationship data, calculating distance change gradient data, building a temperature compensation model based on the gradient data, using a piecewise linear interpolation algorithm to calculate the compensation value of each distance measurement point, and obtaining point cloud distance data after temperature compensation; Step S142, obtaining the temperature-compensated point cloud distance data and environmental parameter data, establishing the atmospheric refractive index model data, calculating the light propagation path curvature data, constructing a spatial position correction model based on the curvature data, using an iterative calculation method to solve the position offset of each point, and obtaining the refraction-corrected point cloud coordinate data; Step S143, obtaining the point cloud coordinate data after refraction correction, calculating the local curvature feature data of the point cloud, extracting the time series scanning sequence data, partitioning the point cloud according to the curvature feature data, establishing the data fluctuation feature sequence of each partition, and obtaining the partitioned point cloud data sequence; Step S144, obtaining a partitioned point cloud data sequence, calculating the local turbulence intensity index data of each partition, constructing an adaptive window size sequence, determining the polynomial fitting order according to the turbulence intensity index data, establishing a weight coefficient matrix, and obtaining a filter parameter set; Step S145, obtaining a partitioned point cloud data sequence and a filter parameter set, constructing a block matrix convolution operator, calculating local filter response data, adjusting the filter parameters according to the response data, and updating the point cloud coordinates using an iterative optimization method to obtain a filtered point cloud data set; Step S146, obtain the filtered point cloud data set, calculate the displacement vector data of adjacent time points, construct the residual turbulence evaluation index, filter outliers according to the evaluation index data, use the local smoothing algorithm to correct the outliers, and obtain the final point cloud data set after environmental compensation.

6. The method for valuing the discharge capacity of a reservoir dam flood discharge facility during operation according to claim 3, characterized in that: The step S32 is specifically as follows: Step S321, obtaining a continuous surface model, calculating Gaussian curvature data, extracting mean curvature data, constructing a shape index field according to the curvature data, calculating the reference features using a local feature statistics method, and obtaining a differential geometry feature data set; Step S322, obtaining a differential geometry feature data set, calculating local shape descriptor data, constructing a feature space mapping, establishing a feature coding dictionary based on the shape descriptor, extracting feature patterns using a sparse representation method, and obtaining shape feature coding data; Step S323, obtaining shape feature coding data and roughness distribution map, calculating feature anomaly measurement data, constructing a multi-dimensional discrimination criterion, establishing a classification boundary according to the anomaly measurement data, using a support vector machine algorithm to identify anomaly patterns, and obtaining abnormal area marking data; Step S324, obtaining abnormal region labeling data, calculating regional connectivity data, constructing a regional growth criterion, merging adjacent abnormal regions according to the connectivity data, optimizing the regional boundaries using a morphological processing method, and obtaining damaged region contour data; Step S325, obtaining the damaged area contour data, calculating the regional morphological feature data, constructing a feature vector template, extracting damage feature parameters according to the morphological feature data, generating feature vectors by normalization processing, and obtaining a damage feature vector set.

7. The method for valuing the discharge capacity of a reservoir dam flood discharge facility during operation according to claim 4, characterized in that: The step S41 is specifically as follows: Step S411, obtain a cross-sectional parameter set and a roughness distribution map, calculate hydraulic radius data, extract cross-sectional area data, construct a local resistance coefficient according to the roughness distribution, calculate basic flow parameters using a piecewise integration method, and obtain a hydraulic element data set; Step S412, obtaining a hydraulic element data set and damage characteristic data, calculating the damage influence coefficient, constructing a local loss model, adjusting the flow coefficient according to the damage characteristics, solving the correction parameters using an iterative calculation method, and obtaining a loss coefficient data set; Step S413, obtaining a loss coefficient data set, calculating water flow state parameters, constructing a flow pattern discrimination criterion, determining a flow pattern coefficient according to the state parameters, dividing the flow pattern area using a classification processing method, and obtaining flow pattern partition data; Step S414, obtain flow pattern partition data, calculate the comprehensive correction coefficient, build a flow correction model, update the flow formula according to the correction coefficient, use the correction calculation method to solve the flow coefficient, and obtain the corrected flow coefficient.

8. A system for rating the discharge capacity of a reservoir dam flood discharge facility during operation, characterized in that: include: at least one processor; as well as, a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the method for calibrating the discharge capacity of a reservoir dam flood discharge facility during operation as described in any one of claims 1 to 7.

Images