Intelligent splitting method based on data segmentation algorithm

Abstract

The invention discloses an intelligent splitting method based on a data segmentation algorithm, and particularly relates to the technical field of intelligent splitting, and the method comprises the steps: carrying out the finite element multi-scale stress analysis of a fusion model, mapping the stress data to a point cloud, marking a high-stress region, a regular geometric region and a transition region based on an adaptive DBSCAN clustering algorithm, and obtaining a high-stress region, a regular geometric region and a transition region; the marking and partitioning efficiency and the transition region identification precision are effectively improved; according to the method, graph segmentation is carried out on the feature partitions to obtain a steel structure preliminary module division scheme, and geometric segmentation is carried out on the steel structure preliminary module division scheme to obtain a steel structure refined module division scheme and machining process parameters; based on an NSGA-II algorithm and a PPO algorithm, dynamic multi-objective optimization is performed on a steel structure refined module splitting scheme, and a final Pareto leading-edge solution set and an optimization scheme comparison matrix are output, so that the segmentation accuracy is improved, the module division dispersion is reduced, the adaptation rate of process parameters and the process is improved, and multi-objective collaborative optimization is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent splitting technology, and more specifically, to an intelligent splitting method based on a data segmentation algorithm. Background Art

[0002] With the rapid development of the construction industry, steel structures are increasingly used in various building projects. Traditional methods for disassembling steel structure modules often rely on manual experience and lack precise data analysis and optimization strategies. This leads to low disassembly efficiency, severe resource waste, and difficulty meeting complex project requirements. Therefore, there is an urgent need for an intelligent, efficient, and precise method for optimizing the disassembly of steel structure modules to improve both production efficiency and disassembly quality.

[0003] The existing intelligent splitting method of steel structure obtains the point cloud and BIM model by manually selecting feature points to achieve preliminary alignment of the point cloud and the BIM model, and then uses the classic iterative closest point algorithm for precise alignment. After alignment, single-scale finite element analysis is performed and high-stress areas are marked. When splitting the steel structure, the steel structure is cut according to the standard component size. Finally, with the goal of minimizing cost, a genetic algorithm is used to generate a local optimal solution, and cutting parameters are selected from a fixed process library, which to a certain extent ensures the quality of steel structure splitting and production efficiency.

[0004] However, the existing methods still have some problems: the error bandwidth of the single-scale model used in stress analysis during feature recognition is high, the efficiency of manual marking and partitioning is low, and the accuracy of transition zone recognition is insufficient. The feature recognition process should be optimized; the existing methods use a single segmentation dimension for steel structure segmentation, which has insufficient segmentation accuracy, the discreteness of module division exceeds the standard, and the static matching of processing parameters is low in adaptability to the process, which can only ensure cost minimization. The segmentation process should be optimized. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides an intelligent splitting method based on a data segmentation algorithm to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: an intelligent splitting method based on a data segmentation algorithm, comprising the following steps:

[0007] S1. Original data acquisition: Scan the steel structure to be split to obtain the surface point cloud data of the steel structure and extract the BIM model data of the steel structure to be split;

[0008] S2. Steel structure data fusion: De-noise the steel structure surface point cloud data, and then align it with the steel structure BIM model data to obtain a fusion model;

[0009] S3, Feature Marking: Perform finite element multi-scale stress analysis on the fusion model, map the stress data to the point cloud, mark the high stress area, regular geometry area and transition area based on the adaptive DBSCAN clustering algorithm, and output the feature marking results;

[0010] S4. Hierarchical intelligent segmentation: Perform graph segmentation on the feature partitions to obtain a preliminary module division scheme for the steel structure. Perform geometric segmentation on the preliminary module division scheme for the steel structure to obtain a refined module division scheme for the steel structure, a cutting edge weight table, and processing parameters.

[0011] S5. Dynamic optimization of splitting scheme: Based on the NSGA-II algorithm and PPO algorithm, dynamic multi-objective optimization is performed on the refined module splitting scheme of the steel structure and the final Pareto front solution set and optimization scheme comparison matrix are output;

[0012] S6. Closed-loop verification: Perform finite element simulation verification and digital twin feedback on the optimization solution, and output a finite element analysis report, a digital twin verification report, and a tolerance compensation parameter table;

[0013] S7. Dynamic adjustment of schemes: Dynamically adjust the arrangement order of the optimization schemes based on the closed-loop verification results, mark the first-ranked scheme as the target splitting scheme, mark the second-ranked and third-ranked schemes as the first alternative scheme and the second alternative scheme respectively, and upload the marked schemes to the steel structure splitting scheme management terminal.

[0014] The technical effects and advantages of the present invention are as follows:

[0015] The present invention performs finite element multi-scale stress analysis on the fusion model, maps the stress data to a point cloud, marks high stress areas, regular geometric areas and transition areas based on an adaptive DBSCAN clustering algorithm, and outputs feature marking results, which effectively reduces the error bandwidth, improves the marking partitioning efficiency, and improves the transition area recognition accuracy. The present invention performs graph segmentation on the feature partitioning to obtain a preliminary module division scheme for the steel structure, performs geometric segmentation on the preliminary module division scheme for the steel structure to obtain a refined module division scheme for the steel structure, a cutting edge weight table, and processing process parameters. Based on the NSGA-II algorithm and the PPO algorithm, the refined module splitting scheme for the steel structure is dynamically multi-objective optimized and the final Pareto frontier solution set and the optimization scheme comparison matrix are output, which improves the segmentation accuracy and reduces the module division discreteness. The dynamic process parameter matching improves the adaptation rate to the process, realizes multi-objective collaborative optimization, and improves the splitting quality and production efficiency of the final segmentation scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A diagram showing the steps of the method of the present invention.

[0017] Figure 2 This is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] like Figure 1 This embodiment provides an intelligent splitting method based on a data segmentation algorithm, comprising the following steps:

[0020] S1. Original data acquisition: Scan the steel structure to be split to obtain the surface point cloud data of the steel structure and extract the BIM model data of the steel structure to be split;

[0021] Furthermore, the extracted BIM model data of the steel structure to be split includes geometric parameters, connection relationships, material properties, non-geometric information and metadata. The geometric parameters include component geometry, component size and component spatial position coordinates; the connection relationship includes the node type and connection method between components; the material properties include steel grade, material strength and surface treatment requirements; the non-geometric information includes component number, component type and design load; the metadata includes BIM model version information and model coordinate system benchmark.

[0022] It should be specifically noted in this embodiment that a three-dimensional laser scanner can be used to obtain point cloud data of the steel structure surface, and BIM data such as geometric parameters and connection relationships of the steel structure model can be exported from the Revit platform.

[0023] S2. Steel structure data fusion: De-noise the steel structure surface point cloud data, and then align it with the steel structure BIM model data to obtain a fusion model;

[0024] Furthermore, the specific steps of steel structure data fusion are as follows:

[0025] A1. Point cloud denoising based on curvature-based non-local mean filtering;

[0026] A2. Calculate the centroid of the point cloud and the BIM model, then translate the point cloud so that its centroid coincides with the centroid of the BIM model.

[0027] A3. Calculate the covariance matrix for the point cloud and BIM model respectively, then perform eigendecomposition to obtain the eigenvectors of the point cloud and BIM model. Construct a rotation matrix and apply it to align the principal axis of the point cloud with the BIM model.

[0028] A4. After determining the initialization parameters, find the nearest point of each point in the point cloud in the BIM model. Solve the optimal rotation matrix and optimal translation vector for each iteration by minimizing the squared sum of the distances between corresponding points. Then update the point cloud coordinates and continue iterative processing until the iteration termination criteria are met and the steel structure data fusion is completed.

[0029] It should be specifically noted that in this embodiment, the specific steps of point cloud denoising are as follows:

[0030] A11. Search neighborhood: Set the initial neighborhood radius r and search for the point cloud with the i-th point cloud coordinate p ai Points whose distance is less than or equal to the initial neighborhood radius constitute the neighborhood D ai ;

[0031] A12. Calculate curvature: For the i-th point cloud coordinate p ai , calculate the eigenvalue λ of the covariance matrix within its initial neighborhood radius a1 ≧λ a2 ≧λ a3 , the i-th point cloud coordinate p ai The curvature value C(p ai ) is the ratio of the minimum eigenvalue to other eigenvalues. The specific formula is:

[0032] A13. Calculate point p aj Point p ai The weight w(p ai ,p aj ): p aj Neighborhood D ai The three-dimensional coordinates of the j-th original point, σ d is the spatial distance weight parameter, the default value is equal to the initial neighborhood radius, σ c is the curvature difference weight parameter, the default value is 0.1, C(p aj ) is the neighborhood D ai The curvature value of the jth original point in the inner space is calculated in the same way as the i-th point cloud coordinate p ai The curvature value C(p ai ) is calculated in the same way;

[0033] A14. For the i-th point p in the point cloud ai , weighted average of all points p in its neighborhood aj The denoised three-dimensional coordinates p are obtained after the coordinates are obtained. ai * , the specific formula is:

[0034] Specifically, in this embodiment, the steps for aligning the point cloud with the centroid of the BIM model are as follows:

[0035] A21. Calculate the centroid μ of the denoised point cloud a , the specific formula is: N a is the total number of points in the point cloud, and the centroid μ of the BIM model is calculated d , the specific formula is: N d is the number of vertices in the BIM model, q dj is the coordinate of the jth vertex of the BIM model;

[0036] A22. Translate the point cloud so that its centroid coincides with the centroid of the BIM model. Then the coordinates of the ith denoised point cloud after centroid alignment are p ai 'Satisfied: p ai '=p ai * -μ a +μ d .

[0037] Specifically, in this embodiment, the steps for aligning the point cloud with the main axis direction of the BIM model are as follows:

[0038] A31. Calculate the point cloud covariance matrix M a , the specific formula is: (p ai '-μ d ) T For (p ai '-μ d ) to calculate the covariance matrix M of the BIM model d , the specific formula is: (q dj '-μ d ) T for (q dj '-μ d )’s transposed matrix;

[0039] A32. Perform eigendecomposition on the point cloud covariance matrix to obtain the eigenvector V a , V a =[v a1 ,v a2 ,v a3 ], perform eigendecomposition on the covariance matrix of the BIM model and obtain the eigenvector V d , V d =[v d1 ,v d2 ,v d3 ];

[0040] A33. Construct the rotation matrix R PCA , the specific formula is: R PCA =Vd ·V a -1 , V a -1 V a The inverse matrix of

[0041] A34. Apply the rotation matrix to align the point cloud with the main axis direction of the BIM model. The i-th point cloud coordinate p after the main axis direction is aligned ai "Satisfied: p ai ”=R PCA ·p ai '.

[0042] It should be specifically noted that in this embodiment, the specific steps of A4 are as follows:

[0043] A41. Determine the initialization parameters: the i-th point cloud coordinate p when no iteration is performed ai (0) =p ai ", the initial rotation matrix R0 = I, I is the unit matrix, which means the main diagonal elements are 1 and the rest of the elements are 0. The initial translation vector t0 = 0;

[0044] A42, for each point p in the point cloud ai Find its closest point q in the BIM model dj (i);

[0045] A43. The optimal rotation matrix and optimal translation vector for the kth iteration are solved by minimizing the square sum of the distances between corresponding points for the kth iteration. The mathematical expression for minimizing the square sum of the distances between corresponding points for the kth iteration is: R k =V k-1 ·U k-1 T , V k-1 represents the main direction distribution of the decentralized coordinates of the BIM model at the k-1th iteration, U k-1 represents the main direction distribution of the decentralized coordinates of the k-1th iteration point cloud, t k =μ q,k-1 -R k μ p,k-1 , μ q,k-1 represents the centroid of the nearest point set in the BIM model for each point in the k-1th iteration point cloud, μ p,k-1 represents the centroid of the k-1th iteration point cloud;

[0046] A44. Update the point cloud coordinates p after the kth iteration ai (k) , the specific formula is: p ai (k) =R k ·pai (k-1) +t k ;

[0047] A45. Iteration termination judgment: When Or k = T max The iteration terminates when the maximum number of iterations T is usually taken max =100, convergence threshold∈=10 -5 .

[0048] S3, Feature Marking: Perform finite element multi-scale stress analysis on the fusion model, map the stress data to the point cloud, mark the high stress area, regular geometry area and transition area based on the adaptive DBSCAN clustering algorithm, and output the feature marking results;

[0049] Furthermore, the specific steps of performing finite element multi-scale stress analysis on the fusion model are as follows:

[0050] B11. Macroscale Modeling: Select shell or beam element types based on the steel structure material properties, load conditions, and expected analysis objectives, and preliminarily set the solver's nonlinear options and convergence tolerance. Generate a sparse mesh for the entire structure, calculate the global stress distribution, and identify preliminary high-stress areas.

[0051] B12. Microscale modeling: Establish a refined sub-model of solid elements in the high-stress area marked by the macroscale. Use the macroscale results as the boundary conditions of the sub-model through nodal displacement interpolation or stress field mapping. Refine the sub-model mesh size until the stress gradient resolution requirement is met.

[0052] B13. Solver Setup and Execution: Select a compatible solver based on the element type, configure the load step, substep, and output request. If the microanalysis generates a new high stress area, return to B11 to expand the macro-marked area and reanalyze.

[0053] B14. Result output: Output the overall Von Mises stress field σ of the macro model vm (x, y, z), the microscopic sub-model includes the refined stress field of the principal stress components, the displacement vector U of each node w [U wx ,U wy ,U wz ], maximum displacement position, maximum stress value and its spatial coordinates (x max ,y max ,z max ), a safety factor based on a specified yield strength and the geometric curvature distribution of stress concentration areas.

[0054] Furthermore, the specific steps for mapping stress data to point cloud are as follows:

[0055] B21. Stress interpolation: For point P in the point cloud bi (x,y,z), search for the number of finite element nodes N within its initial neighborhood radius b Then calculate point P based on the interpolation formula bi The stress value σ pbi , the specific formula is: σ j Point P bi The stress value of the jth finite element node within the initial neighborhood radius, d i,j Point P bi The distance to the jth finite element node;

[0056] B22. Global normalization: Normalization is performed based on the stress range of all finite element nodes. The specific formula is: σ max , σ min They are the maximum stress value and the minimum stress value of the finite element node respectively;

[0057] B23, outlier truncation: when point P bi When the stress value is greater than the preset value, it is marked as a singular point and removed;

[0058] B24, Data output: Output point cloud set with stress labels {p bi (x,y,z,σ pbi * )}.

[0059] Furthermore, the specific steps for marking high stress areas, regular geometry areas, and transition areas based on the adaptive DBSCAN clustering algorithm are as follows:

[0060] B31. Calculate point p bi The k-nearest neighbor average stress weighted density ρ pbi , the specific formula is: N(p bi ,k) represents point p bi The k-nearest neighbor set, m pbj is the jth neighbor point p bj The number of neighborhood points under the initial neighborhood radius;

[0061] B32. Calculate the neighborhood radius dynamic adjustment value r c , the specific formula is: c =r0×exp(-α s ×ρ pbi ), r0 is the initial neighborhood radius, α s is the attenuation coefficient;

[0062] B33, according to the local stress gradient Dynamically set minPts, the specific formula is: Among them, minPts0 is the minimum number of basic points, β P is the adaptive adjustment coefficient, β P ∈[0.1,1.0];

[0063] B34. Calculate the principal curvatures C1 and C2 through PCA, and standardize the principal curvatures C1 and C2 to obtain C1 * 、C2 * , the specific formula is: C e , σ a are the global curvature mean and standard deviation,

[0064] B35. Construct the eigenvector F after balancing the dimensions through Mahalanobis distance pbi , F pbi =[σ pbi * ,C1 * ,C2 * ];

[0065] B36. Setting the upper stress threshold σ based on the quantile of the global stress-curvature distribution y1 , curvature change upper threshold C y1 , stress lower limit threshold σ y2 And the curvature change lower threshold C y2 ;

[0066] B37, will σ pbi * >σ y1 And ||C1-C2||>C y1 The point is marked as the high stress area, and σ pbi * <σ y2 And ||C1-C2|| <C y2 The points are marked as regular geometric areas, and the remaining points are marked as transition areas.

[0067] Specifically, it should be noted that in this embodiment, the feature labeling results include partition category labels, multi-level label distribution maps; average stress, maximum stress, stress standard deviation, regional area, point cloud ratio, high stress node coordinates, high stress node stress value and average curvature change rate of the high stress area, average stress, regional area, point cloud ratio, principal curvature mean and curvature entropy of the regular geometric area, average stress, stress gradient mean, regional area, point cloud ratio and curvature change rate standard deviation of the transition zone; geometric boundary point coordinates and stress boundary point coordinates of the boundary point set, boundary type labels and closed boundary loops.

[0068] It should be specifically noted in this embodiment that the upper stress threshold, the upper curvature change threshold, the lower stress threshold, and the lower curvature change threshold are all selected based on actual needs and are not limited to specific values here. In terms of numerical relationship, the upper stress threshold is greater than the lower stress threshold, and the upper curvature change threshold is greater than the lower curvature change threshold.

[0069] S4. Hierarchical intelligent segmentation: Perform graph segmentation on the feature partitions to obtain a preliminary module division scheme for the steel structure. Perform geometric segmentation on the preliminary module division scheme for the steel structure to obtain a refined module division scheme for the steel structure, a cutting edge weight table, and processing parameters.

[0070] Furthermore, the specific steps of hierarchical intelligent segmentation are as follows:

[0071] C1. Graph model abstraction and edge weight calculation: Define the center point of the high stress area, the centroid of the geometric regular area, and the key point of the transition area as graph nodes, and calculate the weight w of the edge connecting the i and j nodes. bi,j , the specific formula is: Where α y , σ i , σ i , β h 、f i,j , γ l ,θ i,j They are stress weight coefficient, stress value of node i, stress value of node j, geometric complexity weight coefficient, geometric connection area ratio of edge (i, j), connection strength weight coefficient and connection strength of edge (i, j);

[0072] C2. Multi-constraint graph segmentation optimization: Determine the objective function, the specific mathematical expression is: Minimize∑w bi,j ×x i,j +λ a ×Im balance(S k ), x i,j The value is 1 when the edge (i, j) is cut, and the value is 0 when the edge (i, j) is not cut. a is the weight adjustment coefficient, Imbalance(S k ) represents the submodule scale balance penalty term, Im balance(S k )=max(|S k |-Avgsize) 2 , S k is the kth submodule obtained after graph segmentation, Avgsize is the ratio of the total number of nodes to the number of submodules, |S k ∣ is the submodule S k The number of nodes, process constraints and mechanical constraints are set;

[0073] C3, Geometric Decomposition and Manufacturability Verification: Extract the module's outer contour based on the Alpha Shape algorithm to avoid generating non-convex shapes, insert process segmentation surfaces for complex connection areas, and interact with the processing equipment database to verify component size matching;

[0074] C4. Quantification of cutting edge priority: The data expression of the priority score is the product of the stress gradient and safety factor on both sides of the edge and the product of the groove form matching degree;

[0075] C5. Data output: Output the geometric boundary description of the segmentation scheme, the refined sub-module CAD model, the cutting edge weight table and the processing parameters. The geometric boundary description includes the boundary dominant type, boundary point coordinates, boundary average stress gradient, boundary curvature and recommended connection method. The refined sub-module CAD model includes process characteristics, assembly datum and metadata. The cutting edge weight table includes the cutting edge ID, starting point coordinates, end point coordinates, weight, priority score, recommended cutting angle, groove type and thermal deformation compensation. The processing parameters include temporary process parameters and final process parameters. The final process parameters include sub-module ID, cutting sequence, welding sequence, groove angle, welding current and post-processing requirements.

[0076] S5. Dynamic optimization of splitting scheme: Based on the NSGA-II algorithm and PPO algorithm, dynamic multi-objective optimization is performed on the refined module splitting scheme of the steel structure and the final Pareto frontier solution set and optimization scheme comparison matrix are output;

[0077] Furthermore, the specific steps of dynamic optimization of the segmentation scheme are as follows:

[0078] D1. Setting optimization objectives and dynamic constraints: Minimizing total cost, maximizing structural strength, and minimizing the number of modules are set as optimization objectives, and transportation constraints and processing constraints are set as dynamic constraints;

[0079] Specifically, in this embodiment, the total cost is the sum of the cutting cost, the welding cost, and the transportation cost; and the structural strength is the minimum of the ratio of the maximum equivalent stress to the material yield strength and the ratio of the maximum displacement to the allowable displacement.

[0080] D2. Application of NSGA-II algorithm: Apply NSGA-II algorithm to obtain the initial Pareto frontier solution set and optimization scheme comparison matrix;

[0081] Furthermore, the specific steps of applying the NSGA-II algorithm are as follows:

[0082] D21, after inputting the search space of the cutting edge selection sequence and the population size N, N initial splitting schemes are randomly generated, and each scheme is encoded as a binary string;

[0083] D22. After calculating the three objective values for each solution, the population is stratified using the Pareto dominance relationship. Rank 1 is the optimal non-dominated solution set, Rank 2 is the second best, and so on.

[0084] D23. For individuals of the same rank, sort them by each objective function value and calculate the crowding distance of their adjacent solutions: fm is the mth objective function value;

[0085] D24, randomly select k individuals from the population, select individuals with higher rank or larger crowding distance to enter the mating pool, and perform crossover on the selected parent individuals according to the crossover probability p. c Exchange some genes with a mutation probability p m Flip gene position;

[0086] D25. Merge the original population and offspring to obtain 2N individuals. Select the top N best individuals according to non-dominated sorting and crowding distance to form a new generation population.

[0087] D26: If convergence has not been achieved before the maximum number of generations, the algorithm will be terminated directly. If the growth rate of the Pareto frontier hypervolume of two adjacent generations is lower than the preset value or the algorithm has not converged after reaching the maximum number of generations, the algorithm will be terminated forcibly.

[0088] D3, PPO collaborative optimization: obtains the hypervolume growth rate, solution set diversity, and constraint violation rate of the current generation of NSGA-II in real time, adjusts the crossover probability or target weight according to the current state, and updates the NSGA-II parameters;

[0089] D4. Termination judgment: The termination is judged when the Pareto frontier super volume growth rate of two adjacent generations is lower than the preset value and there is no improvement for 10 consecutive generations. The termination is judged when the maximum number of generations is reached or the processing feasibility verification pass rate is higher than the preset value;

[0090] D5. Output the final Pareto frontier solution set, optimization scheme comparison matrix, and dynamic parameter trajectory.

[0091] Furthermore, the final Pareto front solution set includes the unique scheme identifier, sub-module list, total cost, structural strength score, number of sub-modules, processing feasibility score, cutting edge details, connection method, cutting priority, stress gradient, maximum stress and maximum displacement of each solution set; the optimization scheme comparison matrix includes scheme ID, cost weight, strength weight, module number weight, hypervolume contribution diversity score, cost-strength slope and processing feasibility score; the dynamic parameter trajectory includes crossover probability, mutation probability, PPO dynamically adjusted weight and dynamic change data set of average hypervolume.

[0092] S6. Closed-loop verification: Perform finite element simulation verification and digital twin feedback on the optimization solution, and output a finite element analysis report, a digital twin verification report, and a tolerance compensation parameter table;

[0093] Specifically, it should be noted that the specific steps of performing finite element simulation verification on the optimization scheme are as follows:

[0094] E11. Import the CAD model containing process features and material properties into the finite element analysis software and convert the cutting parameters into contact pairs or binding constraints in the finite element analysis;

[0095] E12. Load and Constraint Design: Load conditions include static loads, dynamic loads, and construction loads. Constraint conditions include hinged supports (releasing rotation), sliding supports (releasing translation), and foundation spring stiffness (set according to the geological report).

[0096] E13, Intelligent Meshing: Identify high stress areas and apply second-order hexahedral meshes to them, control mesh size gradients in transition areas, and apply first-order tetrahedral meshes to regular areas;

[0097] E14. Safety Verification and Quantification: Set numerical safety standards for stress, displacement, and buckling, calculate the maximum VonMises stress, displacement contours, and buckling factor, and compare the calculated results with the safety standards to determine whether the optimized solution meets the safety standards.

[0098] E15, Anomaly Detection: Automatically mark out-of-limit areas and associate them with corresponding components in the CAD model;

[0099] E16. Data output: Output finite element analysis report and abnormal area mark. The finite element analysis report includes stress / strain distribution diagram, safety factor and critical load value; the abnormal area mark includes the abnormal area location and abnormal area abnormality type, abnormality cause and abnormality risk level.

[0100] Specifically, in this embodiment, the specific steps of performing digital twin feedback on the optimization solution are as follows:

[0101] E21, data acquisition: collect module installation position deviation (Δx, Δy, Δz) and cutting size error;

[0102] E22. Map measured data to the digital twin model and update geometric and mechanical properties;

[0103] E23, Deviation Analysis: Calculate the installation error. The specific formula is: Calculate the cutting accuracy, the specific formula is:

[0104] E24. Generate dynamic compensation instructions based on deviation analysis results;

[0105] E25. Output the digital twin verification report and tolerance compensation parameter table. The digital twin verification report includes the module ID, design stress, measured stress, installation error, compensation instructions, processing feasibility score, and total cost. The tolerance compensation parameter table includes the compensation item, theoretical value, measured error, and compensation value.

[0106] S7. Dynamic adjustment of schemes: Dynamically adjust the arrangement order of the optimization schemes based on the closed-loop verification results, mark the first-ranked scheme as the target splitting scheme, mark the second-ranked and third-ranked schemes as the first alternative scheme and the second alternative scheme respectively, and upload the marked schemes to the steel structure splitting scheme management terminal.

[0107] In this embodiment, it should be specifically explained that the specific steps of dynamically adjusting the solution are as follows:

[0108] F1. Retrieve the safety factor S for closed-loop verification of each solution qi , total cost W zi , the number of abnormal areas N yi , Processing feasibility score M xi And the mean value of tolerance compensation T ei ;

[0109] F2, safety factor S qi , total cost W zi , the number of abnormal areas N yi , Processing feasibility score M xi And the mean value of tolerance compensation T ei Perform normalization processing, the specific formula is as follows: S qmax 、S qmin 、W zmax 、W zmin 、N ymax 、M xmax 、M xmin 、T emax 、T emin They are the maximum safety factor, minimum safety factor, maximum total cost, minimum total cost, maximum number of abnormal areas, maximum processing feasibility score, minimum processing feasibility score, maximum mean tolerance compensation value, and minimum mean tolerance compensation value of closed-loop verification of the optimization scheme, respectively.

[0110] F3, respectively, with β Sb , β Wb ,λ Nbi , β Mb , β TbRepresents the safety weight, cost weight, abnormal penalty coefficient, process weight, and tolerance weight after dynamic adjustment. The data value of each weight before closed-loop verification is obtained, and β is used to represent the safety weight, cost weight, abnormal penalty coefficient, process weight, and tolerance weight after dynamic adjustment. Sa , β Wa ,λ Na , β Ma , β Ta Indicates that λ Na =0;

[0111] F4, N yi When β is greater than 0, Sb =β Sa +0.1, T ei When β is greater than the designed compensation threshold, Wb =β Wa -0.05, M xi When the score is less than the set minimum processing feasibility score, β Mb =β Ma +0.05, β Tb =β Ta ,λ Nbi =N gi +0.2, N gi is the number of high-risk anomalies in the i-th solution;

[0112] F5. Calculate the comprehensive score Z of the i-th solution i , the specific formula is:

[0113] F6. Sort the optimization schemes from high to low according to the comprehensive score values, mark the top-ranked scheme as the target split scheme, mark the second-ranked and third-ranked schemes as the first alternative scheme and the second alternative scheme respectively, and upload the marked schemes to the steel structure split scheme management terminal.

[0114] What needs to be specifically explained in this embodiment is that the preset values, expected values, set values, design values, thresholds, weight coefficients, and weight adjustment coefficients used are all selected based on actual needs and are not limited to specific values here.

[0115] like Figure 2 The embodiment shown provides an intelligent splitting system based on a data segmentation algorithm, including an original data acquisition module, a steel structure data fusion module, a feature processing module, an intelligent segmentation module, a segmentation scheme dynamic optimization module, a closed-loop verification module, a scheme dynamic adjustment module and a database. The original data acquisition module, the steel structure data fusion module, the feature processing module, the intelligent segmentation module, the segmentation scheme dynamic optimization module, the closed-loop verification module and the scheme dynamic adjustment module are connected in sequence, the original data acquisition module is connected to the feature processing module, the intelligent segmentation module and the closed-loop verification module respectively, and all modules in the system are connected to the database.

[0116] The raw data acquisition module scans the steel structure to be split to obtain surface point cloud data of the steel structure and extracts BIM model data of the steel structure to be split;

[0117] The steel structure data fusion module denoises the steel structure surface point cloud data and then registers it with the steel structure BIM model data to obtain a fusion model;

[0118] The feature processing module performs finite element multi-scale stress analysis on the fusion model, maps stress data to a point cloud, marks high stress areas, regular geometry areas, and transition areas based on an adaptive DBSCAN clustering algorithm, and outputs feature marking results;

[0119] The intelligent segmentation module performs graph segmentation on the feature partitions to obtain a preliminary module division scheme for the steel structure, and performs geometric segmentation on the preliminary module division scheme for the steel structure to obtain a refined module division scheme for the steel structure, a cutting edge weight table, and processing parameters;

[0120] The splitting scheme dynamic optimization module performs dynamic multi-objective optimization on the steel structure refined module splitting scheme based on the NSGA-II algorithm and the PPO algorithm and outputs the Pareto frontier solution set and the optimization scheme comparison matrix;

[0121] The closed-loop verification module performs finite element simulation verification and digital twin feedback on the optimization scheme, and outputs a finite element analysis report, a digital twin verification report, and a tolerance compensation parameter table;

[0122] The scheme dynamic adjustment module dynamically adjusts the arrangement order of the optimization schemes based on the closed-loop verification results, marks the first-ranked scheme as the target splitting scheme, marks the second-ranked and third-ranked schemes as the first alternative scheme and the second alternative scheme, respectively, and uploads the marked schemes to the steel structure splitting scheme management terminal;

[0123] The database is used to store data information of all modules in the system.

[0124] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent splitting method based on a data segmentation algorithm, characterized by: The following steps are involved: S1. Original data acquisition: Scan the steel structure to be split to obtain the surface point cloud data of the steel structure and extract the BIM model data of the steel structure to be split; S2. Steel structure data fusion: De-noise the steel structure surface point cloud data, and then align it with the steel structure BIM model data to obtain a fusion model; S3. Feature tags: Perform finite element multi-scale stress analysis on the fusion model, map the stress data to a point cloud, mark high stress areas, regular geometry areas, and transition areas based on the adaptive DBSCAN clustering algorithm, and output feature marking results; S4. Hierarchical intelligent segmentation: Perform graph segmentation on the feature partitions to obtain a preliminary module division scheme for the steel structure. Perform geometric segmentation on the preliminary module division scheme for the steel structure to obtain a refined module division scheme for the steel structure, a cutting edge weight table, and processing parameters. S5. Dynamic optimization of splitting scheme: Based on the NSGA-II algorithm and PPO algorithm, dynamic multi-objective optimization is performed on the refined module splitting scheme of the steel structure and the final Pareto front solution set and optimization scheme comparison matrix are output; S6. Closed-loop verification: Perform finite element simulation verification and digital twin feedback on the optimization solution, and output a finite element analysis report, a digital twin verification report, and a tolerance compensation parameter table; S7. Dynamic adjustment of schemes: Dynamically adjust the arrangement order of the optimization schemes based on the closed-loop verification results, mark the first-ranked scheme as the target splitting scheme, mark the second-ranked and third-ranked schemes as the first alternative scheme and the second alternative scheme respectively, and upload the marked schemes to the steel structure splitting scheme management terminal.

2. The intelligent splitting method based on the data segmentation algorithm according to claim 1, characterized in that: The BIM model data of the steel structure to be split extracted in step S1 includes geometric parameters, connection relationships, material properties, non-geometric information and metadata, wherein the geometric parameters include component geometric shape, component size and component spatial position coordinates; The connection relationship includes the node type and connection method between components; material properties include steel grade, material strength and surface treatment requirements; non-geometric information includes component number, component type and design load; metadata includes BIM model version information and model coordinate system benchmark.

3. The intelligent splitting method based on the data segmentation algorithm according to claim 1 is characterized in that: The specific steps of the steel structure data fusion in step S2 are as follows: A1. Point cloud denoising based on curvature-based non-local mean filtering; A2. Calculate the centroid of the point cloud and the BIM model, then translate the point cloud so that its centroid coincides with the centroid of the BIM model. A3. Calculate the covariance matrix for the point cloud and BIM model respectively, then perform eigendecomposition to obtain the eigenvectors of the point cloud and BIM model. Construct a rotation matrix and apply it to align the principal axis of the point cloud with the BIM model. A4. After determining the initialization parameters, find the nearest point of each point in the point cloud in the BIM model. Solve the optimal rotation matrix and optimal translation vector for each iteration by minimizing the squared sum of the distances between corresponding points. Then update the point cloud coordinates and continue iterative processing until the iteration termination criteria are met and the steel structure data fusion is completed.

4. The intelligent splitting method based on the data segmentation algorithm according to claim 1 is characterized in that: The specific steps of step S3 for performing finite element multi-scale stress analysis on the fusion model are as follows: B11. Macroscale Modeling: Select shell or beam element types based on the steel structure material properties, load conditions, and expected analysis objectives, and preliminarily set the solver's nonlinear options and convergence tolerance. Generate a sparse mesh for the entire structure, calculate the global stress distribution, and identify preliminary high-stress areas. B12. Microscale modeling: Establish a refined sub-model of solid elements in the high-stress area marked by the macroscale. Use the macroscale results as the boundary conditions of the sub-model through nodal displacement interpolation or stress field mapping. Refine the sub-model mesh size until the stress gradient resolution requirement is met. B13. Solver Setup and Execution: Select a compatible solver based on the element type, configure the load step, substep, and output request. If the microanalysis generates a new high stress area, return to B11 to expand the macro-marked area and reanalyze. B14. Result output: Output the overall Von Mises stress field σ of the macro model vm (x, y, z), the microscopic sub-model includes the refined stress field of the principal stress components, the displacement vector U of each node w [U wx ,U wy ,U wz ], maximum displacement position, maximum stress value and its spatial coordinates (x max ,y max ,z max ), a safety factor based on a specified yield strength and the geometric curvature distribution of stress concentration areas.

5. The intelligent splitting method based on the data segmentation algorithm according to claim 1 is characterized in that: The specific steps of step S3 for mapping stress data to point cloud are as follows: B21. Stress interpolation: For point P in the point cloud bi (x,y,z), search for the number of finite element nodes N within its initial neighborhood radius b Then calculate point P based on the interpolation formula bi The stress value σ pbi , the specific formula is: σ j Point P bi The stress value of the jth finite element node within the initial neighborhood radius, d i,j Point P bi The distance to the jth finite element node; B22. Global normalization: Normalization is performed based on the stress range of all finite element nodes. The specific formula is: σ max , σ min They are the maximum stress value and the minimum stress value of the finite element node respectively; B23, outlier truncation: when point P bi When the stress value is greater than the preset value, it is marked as a singular point and removed; B24, Data output: Output point cloud set with stress labels {p bi (x,y,z,σ pbi * )}.

6. The intelligent splitting method based on the data segmentation algorithm according to claim 1, characterized in that: The specific steps of step S3 for marking high stress areas, regular geometric areas and transition areas based on the adaptive DBSCAN clustering algorithm are as follows: B31. Calculate point p bi The k-nearest neighbor average stress weighted density ρ pbi , the specific formula is: N(p bi ,k) represents point p bi The k-nearest neighbor set, m pbj is the jth neighbor point p bj The number of neighborhood points under the initial neighborhood radius; B32. Calculate the neighborhood radius dynamic adjustment value r c , the specific formula is: c =r0×exp(-α s ×ρ pbi ), r0 is the initial neighborhood radius, α s is the attenuation coefficient; B33, according to the local stress gradient Dynamically set minPts, the specific formula is: Among them, minPts0 is the minimum number of basic points, β P is the adaptive adjustment coefficient, β P ∈[0.1,1.0]; B34. Calculate the principal curvatures C1 and C2 through PCA, and standardize the principal curvatures C1 and C2 to obtain C1 * 、C2 * , the specific formula is: C e , σ a are the global curvature mean and standard deviation, B35. Construct the eigenvector F after balancing the dimensions through Mahalanobis distance pbi , F pbi =[σ pbi * ,C1 * ,C2 * ]; B36. Setting the upper stress threshold σ based on the quantile of the global stress-curvature distribution y1 , curvature change upper threshold C y1 , stress lower limit threshold σ y2 And the curvature change lower threshold C y2 ; B37, will σ pbi * >σ y1 And ||C1-C2||>C y1 The point is marked as the high stress area, and σ pbi * <σ y2 And ||C1-C2|| <C y2 The points are marked as regular geometric areas, and the remaining points are marked as transition areas.

7. The intelligent splitting method based on the data segmentation algorithm according to claim 1 is characterized in that: The specific steps of step S4 of hierarchical intelligent segmentation are as follows: C1. Graph model abstraction and edge weight calculation: Define the center point of the high stress area, the centroid of the geometric regular area, and the key point of the transition area as graph nodes, and calculate the weight w of the edge connecting the i and j nodes. bi,j , the specific formula is: Where α y , σ i , σ i , β h 、f i,j , γ l ,θ i,j They are stress weight coefficient, stress value of node i, stress value of node j, geometric complexity weight coefficient, geometric connection area ratio of edge (i, j), connection strength weight coefficient and connection strength of edge (i, j); C2. Multi-constraint graph segmentation optimization: Determine the objective function, the specific mathematical expression is: Minimize∑w bi,j ×x i,j +λ a ×Imbalance(S k ), x i,j The value is 1 when the edge (i, j) is cut, and the value is 0 when the edge (i, j) is not cut. a is the weight adjustment coefficient, Imbalance(S k ) represents the submodule scale balance penalty term, Imbalance(S k )=max(|S k |-Avgsize) 2 , S k is the kth submodule obtained after graph segmentation, Avgsize is the ratio of the total number of nodes to the number of submodules, |S k ∣ is the submodule S k The number of nodes, process constraints and mechanical constraints are set; C3, Geometric Decomposition and Manufacturability Verification: Extract the module's outer contour based on the Alpha Shape algorithm to avoid generating non-convex shapes, insert process segmentation surfaces for complex connection areas, and interact with the processing equipment database to verify component size matching; C4. Quantification of cutting edge priority: The data expression of the priority score is the product of the stress gradient and safety factor on both sides of the edge and the product of the groove form matching degree; C5. Data output: Output the geometric boundary description of the segmentation scheme, the refined sub-module CAD model, the cutting edge weight table and the processing parameters.

8. The intelligent splitting method based on the data segmentation algorithm according to claim 1 is characterized in that: The final Pareto front solution set output in step S5 includes the unique identifier of each solution set, submodule list, total cost, structural strength score, number of submodules, processing feasibility score, cutting edge details, connection method, cutting priority, stress gradient, maximum stress and maximum displacement; the optimization solution comparison matrix includes solution ID, cost weight, strength weight, module number weight, super volume contribution diversity score, cost-strength slope and processing feasibility score.

Images