Intelligent modeling method for cable-stayed bridge based on automatic analysis of CAD drawings and improvement of graphic element semantics

Abstract

The application relates to the technical field of bridge engineering and discloses a cable-stayed bridge intelligent modeling method based on CAD drawing automatic analysis and figure element semantic improvement, which comprises the following steps: acquiring DXF format cable-stayed bridge CAD vector drawing, intelligently classifying layers of the DXF format cable-stayed bridge CAD vector drawing and improving semantics in multiple strategies to obtain component data with structural semantics; automatically reconstructing a space hash node topology of the component data with structural semantics to obtain node-unit topology relationship data; correlating text-geometric space and intelligently identifying materials of the node-unit topology relationship data to obtain finite element model basic data with material parameters; detecting tower positions and correcting symmetry of the finite element model basic data with material parameters to obtain corrected finite element model basic data; and performing finite element analysis on the corrected finite element model basic data and self-adaptively retreating a multistage solver to obtain complete cable-stayed bridge finite element analysis results.

Description

Technical Field

[0001] This invention relates to the field of bridge engineering technology, specifically to an intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives. Background Technology

[0002] The design of large bridges typically follows a multi-stage sequential workflow: conceptual design → CAD drawing → finite element modeling → numerical analysis → result evaluation. Among these stages, the conversion from CAD drawings to finite element models (i.e., the drawing-to-model step) is the most time-consuming and experience-dependent step. Engineers need to manually extract the geometric information of components (centerline positions of beams, bottom and top coordinates of towers, anchorage endpoints of cables, etc.) from 2D or 3D CAD drawings, and then construct the numerical model node by node and element by element in the finite element software. For a typical double-tower cable-stayed bridge, this manual modeling process usually takes several days and is highly susceptible to human error.

[0003] The main problems include: data gaps between CAD and CAE; lack of unified standards for layer naming and lack of structural semantics for graphic elements; failure of text annotation information to automatically drive parameter assignment; and lack of automatic detection and correction capabilities for errors between drawings and models. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes an intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives, in order to solve the aforementioned technical problems.

[0005] A smart modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives includes: Obtain DXF format cable-stayed bridge CAD vector drawings, and perform intelligent layer classification and multi-strategy semantic enhancement on the DXF format cable-stayed bridge CAD vector drawings to obtain component data with structural semantics; The component data with structural semantics is subjected to spatial hashing and automatic topology reconstruction to obtain node-unit topology relationship data; Text-geometric spatial association and intelligent material recognition are performed on the node-unit topological relationship data to obtain the basic data of the finite element model with material parameters; For symmetrically arranged cable-stayed bridges, tower location detection and symmetry correction are performed on the finite element model basic data with material parameters to obtain the corrected finite element model basic data. Finite element analysis and multi-level solver adaptive backoff are performed on the corrected finite element model base data to obtain complete finite element analysis results for the cable-stayed bridge.

[0006] Furthermore, the DXF format cable-stayed bridge CAD vector drawing is subjected to intelligent layer classification and multi-strategy semantic enhancement to obtain component data with structural semantics, including: Based on a pre-built keyword library, weighted keyword matching is performed on all layers of the DXF format cable-stayed bridge CAD vector drawing to obtain layer matching scores. Based on the layer matching score, the layers are automatically classified into structure layers, annotation layers, material text layers, and ignored layers; The structural layers are further subdivided into beam, tower / column, and cable types; Different semantic enhancement algorithms are used for different structural component types to identify beam components, tower column components, and cable components respectively, and to obtain component data with structural semantics.

[0007] Furthermore, differentiated semantic enhancement algorithms are employed for different structural component types to identify beam components, tower-column components, and cable-stayed components, including: The elements in the beam layer are processed using a parallel line detection algorithm to identify beam components and extract the beam's centerline and cross-sectional width; The primitives in the tower column layer are processed using a vertical line segment clustering algorithm to identify tower column components and extract the center coordinates of the bottom and top of the tower column; The primitives in the cable-type layer are processed using a diagonal line segment direction grouping algorithm to identify cable components and extract the coordinates of the cable's anchorage endpoints.

[0008] Furthermore, the component data with structural semantics is subjected to spatial hashing and automatic topology reconstruction to obtain node-unit topology relationship data, including: Extract the endpoints of all components from the component data with structural semantics to obtain the component endpoint set; Based on the set of component endpoints, a three-dimensional spatial mesh index is constructed; Map all endpoints in the component endpoint set to the grid cells corresponding to the three-dimensional spatial grid index; The endpoints within the same grid cell are merged into shared nodes and assigned global node IDs sequentially to obtain a set of shared nodes; Traverse all components in all component data with structural semantics, establish corresponding finite element elements based on the shared nodes connected to the components, and obtain node-element topology relationship data.

[0009] Furthermore, text-geometric spatial association and intelligent material recognition are performed on the node-unit topological relationship data to obtain basic data for the finite element model with material parameters, including: Extract all text entities from the DXF format CAD vector drawing of the cable-stayed bridge to obtain a text entity set; For each text entity in the text entity set, extract its insertion point coordinates and text content to obtain text attribute data; Based on the insertion point coordinates in the text attribute data, search all components within a preset radius of each text entity, establish a spatial association between each text entity and the nearest component, and obtain text-component association data; The text content in the text-component association data is parsed to obtain the component's dimensions, material grade, and load information; The dimensions, material grade, and load information are assigned to the corresponding components to obtain the basic data of the finite element model with material parameters.

[0010] Furthermore, the text content in the text-component association data is parsed to obtain the component's dimensions, material grade, and load information, including: Using a hierarchical regular expression rule library, dimensional parameters, material grades, and load information are extracted from the text content of the text-component association data. Based on the extracted material grade, query the pre-constructed material mechanical property database to obtain the corresponding material mechanical parameters; By integrating the dimensional parameters, material mechanical parameters, and load information, the properties of the component are obtained.

[0011] Furthermore, for symmetrically arranged cable-stayed bridges, tower location detection and symmetry correction are performed on the finite element model foundation data with material parameters to obtain corrected finite element model foundation data, including: Extract all node coordinates from the basic data of all finite element models with material parameters to obtain a set of node coordinates; Discretize the coordinates in the set of node coordinates into equal-width intervals to construct a node density histogram; Identify the sealing peak value in the node density histogram, and take the coordinates corresponding to the density peak value that meets the preset conditions as the tower column position; Calculate the lengths of the left and right spans of the cable-stayed bridge based on the locations of the towers. Based on the left span length and the right span length, the node coordinate set is symmetry corrected to obtain the corrected node coordinate set; The corrected set of node coordinates is updated to the finite element model base data with material parameters to obtain the corrected finite element model base data.

[0012] Furthermore, based on the left span length and right span length, the node coordinate set is symmetry corrected to obtain the corrected node coordinate set, including: For a symmetrically arranged cable-stayed bridge, the lengths of the left span and the right span are compared. If they are not equal, the coordinates of the right span nodes are corrected by proportional scaling to obtain a preliminary set of corrected node coordinates. Perform mirror symmetry correction on the right span node in the preliminary correction node coordinate set, so that the right span node is accurately mirrored on the corresponding left span node, to obtain the final correction node coordinate set.

[0013] Furthermore, finite element analysis and multi-level solver adaptive backoff are performed on the corrected finite element model base data to obtain complete finite element analysis results for the cable-stayed bridge, including: Construct a multi-level solver chain that includes a master solver and at least one standby solver; Within each load step of the nonlinear finite element analysis, the solution is performed using the solver at the current level; If the current solver fails to converge, it will automatically switch to the next level solver to continue solving the current load step; If the current solver converges successfully, it will return to the master solver in the next load step to solve the problem. After solving all load steps, the finite element analysis data is extracted to obtain the complete finite element analysis results of the cable-stayed bridge.

[0014] The invention employing the above technical solution has the following advantages: 1. This invention achieves full-chain automation from DXF drawing parsing to FEA result post-processing, compressing the traditional manual modeling time of several days to the hour level, which is conducive to improving the efficiency of design analysis iteration.

[0015] 2. This invention addresses three different structural component types—beams, towers, and cables—by employing three distinct semantic enhancement rules: parallel line pair detection, vertical line segment clustering, and oblique line segment grouping. This approach avoids attempting to handle all component types with a single, universal detector. This differentiated strategy is rooted in the inherent geometrical differences of bridge components: beams exhibit horizontal double-boundary characteristics, towers exhibit vertical multi-segment clustering characteristics, and cables exhibit oblique single-line characteristics. This makes the method highly adaptable and robust to different bridge layouts and component sizes.

[0016] 3. This invention automatically identifies tower locations by utilizing the peak node density characteristics generated by multiple segments of the tower column, eliminating the need for manual pre-marking of tower column positions on drawings. This method is physically intuitive, as the structural characteristics of the tower column (multiple vertical components at the same X position) naturally map to the spatial distribution characteristics of node density. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention, the accompanying drawings used in the specific embodiments will be briefly described below. In all the drawings, the elements or parts are not necessarily drawn to scale.

[0018] Figure 1 This is the flowchart of the intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives, as described in this invention. Figure 1 ; Figure 2 This is the flowchart of the intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives, as described in this invention. Figure 2 ; Figure 3 This is the DXF image of a cable-stayed bridge in the intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives in this invention; Figure 4 This is a schematic diagram of the finite element model (FEA) in the intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives in this invention; Figure 5 This invention presents a schematic diagram of the FEA results for a double-tower cable-stayed bridge in its intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives. Detailed Implementation

[0019] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0020] This invention uses a double-tower cable-stayed bridge with a total span of 210m (span arrangement of 55m+100m+55m, tower height of 80m, and a total of 8 stay cables) as an example to explain in detail the implementation process of this invention. The input file is a DXF drawing in AutoCAD2008 format, containing geometric elements of three types of structural components: beams, towers, and stay cables, as well as text labels for span, cross-sectional dimensions, and material grades.

[0021] like Figures 1-5 As shown, the intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives of the present invention includes: Obtain CAD vector drawings of cable-stayed bridges in DXF format, and perform intelligent layer classification and multi-strategy semantic enhancement on the DXF format CAD vector drawings of cable-stayed bridges to obtain component data with structural semantics.

[0022] In this embodiment, layer intelligent classification and multi-strategy semantic enhancement are performed on the DXF format cable-stayed bridge CAD vector drawing to obtain component data with structural semantics, including: Based on a pre-built keyword library, weighted keyword matching is performed on all layers of a DXF format cable-stayed bridge CAD vector drawing to obtain layer matching scores. Based on the layer matching score, the layers are automatically classified into structure layers, annotation layers, material text layers, and ignored layers; The structural layers are further subdivided into beam, tower / column, and cable categories; Different semantic enhancement algorithms are used for different structural component types to identify beam components, tower column components, and cable components respectively, and to obtain component data with structural semantics.

[0023] In this embodiment, a differentiated semantic enhancement algorithm is used for different structural component types to identify beam components, tower column components, and cable components, including: The elements in the beam layer are processed using a parallel line detection algorithm to identify beam components and extract the beam's centerline and cross-sectional width; The primitives in the tower column layer are processed using a vertical line segment clustering algorithm to identify tower column components and extract the center coordinates of the bottom and top of the tower column; The primitives in the cable-type layer are processed using a diagonal line segment direction grouping algorithm to identify cable components and extract the coordinates of the cable's anchorage endpoints.

[0024] Specifically, keyword-based layer semantic recognition: Build a keyword library containing Chinese, English and industry abbreviations, and automatically classify all layers of DXF drawings into four categories according to semantics: structural layers, dimension layers, material text layers and ignore layers.

[0025] The structural layer is further subdivided into beam, pier, and cable categories to provide prior constraints on component categories for subsequent multi-strategy semantic enhancement.

[0026] Weighted keyword matching algorithm: Layer classification uses a weighted keyword matching algorithm. Let the layer name string be... The first keyword in the keyword database The keyword set for the class layer is Define the matching score for: in, For indicator functions (when keyword) for If the substring is a '1', then '0' is used; otherwise, '1' is used. Predefined weights for keywords. Layers Classified as making The largest category.

[0027] In this embodiment, the BEAM_CENTER layer matches the keyword BEAM (weight 1.0) and is classified as a structural beam layer; the TOWER layer matches the keyword TOWER (weight 1.0) and is classified as a structural tower layer; the CABLE layer matches the keyword cable (weight 1.0) and is classified as a structural cable layer; the DIM and TEXT layers match the keywords dimension and material, respectively, and are classified as annotation layers and material text layers.

[0028] Multi-strategy semantic boosting rules: For different structural component types, differentiated semantic enhancement rules based on domain knowledge are adopted. This mainly utilizes the essential differences in the geometric morphology of beams, towers, and cables in bridge engineering to design targeted primitive and component recognition logic for each type. Beam Parallel Line Pair Detection: In the beam-hint layer, search for all LINE / LWPOLYLINE elements. For any two line segments... and Calculate the angle between its direction vectors. Normal distance between the midpoints of the two line segments .when and When (this threshold is a configurable parameter, and its value is determined based on the actual dimensions of the beam components in the input drawing), it is determined to be a pair of parallel edges.

[0029] The line connecting the midpoints of two parallel lines is taken as the center line (RawBeam) of the beam, and the distance between the parallel lines is taken as the width of the beam section.

[0030] Pier / Tower Vertical Segment Clustering: In the pier-hint layer, filter all vertical segments with an angle less than 15° to the Z-axis (meeting the following conditions). <sin(15°)≈0.259)。

[0031] in, Let be the modulus of the horizontal projection component. Let be the magnitude of the direction vector.

[0032] Project the endpoints of all vertical line segments onto the XY plane and perform one-dimensional clustering based on the X coordinate (clustering tolerance 0.5m). Each cluster corresponds to a tower, and the lowest point of all line segments within the cluster is taken as the base center of the tower, and the highest point is taken as the top center of the tower.

[0033] Grouping of diagonal line segments of cable members: In the cable-hint layer, filter all diagonal line segments with an angle greater than 20° to the horizontal plane (this threshold is a configurable parameter, and its value is determined based on the actual dimensions of the cable members in the input drawing).

[0034] Represent each diagonal line segment as a unit direction vector from the starting point to the ending point. Angle clustering (tolerance 5°) is performed on the direction vectors, and line segments within the same direction group are identified as the same cable bundle.

[0035] In this embodiment, the semantics of the beam component are enhanced: all LINE elements are extracted from the beamhint layer, and multiple pairs of parallel lines are identified.

[0036] For each pair of parallel lines and : 1. Calculate the direction vector and The included angle ; 2. Calculate the normal projection length of the line connecting the midpoints. ; 3. If and Take the line connecting the midpoints as the center line of the beam. As Liang Kuan.

[0037] Semantic enhancement of pier components: Filtering components that meet the requirements in the pier-hint layer. The vertical line segments are clustered in one dimension according to the X coordinate. Two towers are identified (Tower A is located at X≈55m, and Tower B is located at X≈155m), each tower consisting of 21 vertical sub-segments.

[0038] Semantic enhancement of cable-stayed components: In the cable-hint layer, diagonal line segments with an angle greater than 20° to the horizontal plane are filtered out, and clustered by direction vector angle to identify 8 inclined cables (4 on the side of tower A and 4 on the side of tower B).

[0039] Semantic enhancement of cable-stayed components: In the cablehint layer, diagonal line segments with an angle >20° to the horizontal plane are filtered out and clustered according to the direction vector angle, and 8 cable stays are identified (4 on the side of tower A and 4 on the side of tower B).

[0040] Automatic reconstruction of node topology using spatial hashing is performed on component data with structural semantics to obtain node-unit topology relationship data.

[0041] In this embodiment, spatial hashing of component data with structural semantics is used for automatic topology reconstruction to obtain node-unit topology relationship data, including: Extract the endpoints of all components from the component data with structural semantics to obtain the component endpoint set; Construct a three-dimensional spatial mesh index based on the component endpoint set; Map all endpoints in the component endpoint set to the mesh cells corresponding to the 3D spatial mesh index; The endpoints within the same grid cell are merged into shared nodes and assigned global node IDs sequentially to obtain a set of shared nodes; Traverse all components in all component data with structural semantics, establish corresponding finite element elements based on the shared nodes connected to the components, and obtain node-element topology relationship data.

[0042] Specifically, the spatial hash deduplication algorithm: Let the set of all component endpoints (including beam endpoints, tower column endpoints, and cable anchorage endpoints) be . ,in, .

[0043] Construct a 3D spatial mesh index, mesh size ( This threshold is a configurable parameter for node merging tolerance, and its value is determined based on the actual dimensions of the components in the input drawing.

[0044] For each endpoint Calculate the coordinates of its grid cells: Where g is the side length of a cell in the spatial hash grid. Let be the three-dimensional coordinates of the endpoint of the i-th component.

[0045] All endpoints falling into the same grid cell are considered candidates for the same node. For each candidate set, its geometric center is taken as the shared node coordinates, and global node IDs are assigned sequentially starting from 1.

[0046] The theoretical time complexity of this algorithm is... The complexity of the naive pairwise distance comparison algorithm is O(n log n). .

[0047] Automatic topology establishment: After deduplication, iterate through all component objects: if the centerline of a beam connects to the node and nodes Then at node and Establish beam elements between them; Specifically, tower column segments are constructed by establishing tower column units between corresponding nodes, and cables are constructed by establishing cable units between corresponding nodes.

[0048] Since the beam nodes and tower columns at the bridge deck elevation coincide spatially (both located at the beam-tower intersection), the aforementioned spatial hash algorithm automatically merges them into the same shared node, thereby implicitly establishing the beam-tower connection relationship.

[0049] In this embodiment, the main girder is divided into 42 segments along the bridge axis (10 segments in the left span + 22 segments in the middle span + 10 segments in the right span), totaling 43 main girder nodes (including beam end support nodes 1 and 12, and beam-tower intersection nodes 4 and 9). Each tower has 11 sub-segments corresponding to 22 nodes, for a total of 44 tower node identifiers (node ​​4 and node 9 are shared with the main girder).

[0050] Before merging, there were a total of 87 node identifiers (43 main beams + 44 tower columns). After spatial hashing to remove duplicates, the main beam identifiers and tower column identifiers of the two beam-tower intersection nodes (node ​​4 and node 9) were merged into a shared node, ultimately generating 85 unique nodes.

[0051] Automatic merging of shared nodes: The beam endpoint (X=55m, Z=0m) and the tower A endpoint at the bridge deck elevation (X=55m, Z=0m) fall into the same spatial grid and are merged into a single node. This merging automatically establishes the beam-tower connection relationship without the need for manual specification of constraints.

[0052] By performing text-geometric spatial association and intelligent material identification on the node-element topology relationship data, the basic data of the finite element model with material parameters is obtained.

[0053] In this embodiment, text-geometric spatial association and intelligent material recognition are performed on the node-element topological relationship data to obtain the basic data of the finite element model with material parameters, including: Extract all text entities from the DXF format CAD vector drawing of the cable-stayed bridge to obtain a text entity set; For each text entity in the text entity set, extract its insertion point coordinates and text content to obtain text attribute data; Based on the insertion point coordinates in the text attribute data, search all components within a preset radius of each text entity, establish a spatial association between each text entity and the nearest component, and obtain text-component association data; The text content in the text-component association data is parsed to obtain the component's dimensions, material grade, and load information. The dimensions, material grade, and load information are assigned to the corresponding components to obtain the basic data of the finite element model with material parameters.

[0054] In this embodiment, the text content in the text-component association data is parsed to obtain the component's dimensions, material grade, and load information, including: Using a hierarchical regular expression rule library, dimensional parameters, material grades, and load information are extracted from the text content of text-component association data. Based on the extracted material grade, the corresponding material mechanical parameters are obtained by querying the material mechanical property database of the pre-constructed components. By integrating dimensional parameters, material mechanical parameters, and load information, the properties of the component are obtained.

[0055] Specifically, text-component spatial association: Extract all text / text entities from the DXF drawing, with each text entity t having its insertion point coordinates. and text content.

[0056] For each text entity, search its radius. For all components within the range (beams, tower columns, cables), select the nearest component to establish a connection: Where t is a TEXT / MTEX entity in the DXF drawing, and E is the set of all structural components after semantic enhancement. For set A component in, For text insertion point To components The shortest spatial distance, where R is the preset search radius. The values ​​of the independent variables that minimize the objective function, For text The associated component (function return value).

[0057] Size and material parsing driven by regularity rule library: A hierarchical regular expression rule base can be built to identify the following information categories: Table 1 Automatic material database lookup: Create a materials mechanical properties database material_db.json, using the material grade as the key to store the elastic modulus. Yield strength Ultimate strength ,density Isomechanical parameters.

[0058] When the regular rule base extracts a material grade (such as "Q345qD") from the text, it automatically queries the database to obtain complete mechanical parameters: In this embodiment, all TEXT entities (a total of 24 text annotations) are extracted from the drawings. For each text, the nearest component within an 18m range is searched and an association is established.

[0059] For example, the text Q345qD is located at X=105m, Y=5m, and its closest distance to the center line of the main beam is 5m, and it is associated with it.

[0060] The regular expression rule base identified Q345qD as a steel grade, and the material database returned the result. .

[0061] Specifically, C55 is identified as the concrete grade, and the return value is as follows: (Design value of 55MPa axial compressive strength of concrete) Tower location detection and symmetry correction were performed on the basic data of the finite element model with material parameters to obtain the corrected basic data of the finite element model.

[0062] In this embodiment, for a symmetrically arranged cable-stayed bridge, tower location detection and symmetry correction are performed on the finite element model foundation data with material parameters to obtain the corrected finite element model foundation data, including: Extract all node coordinates from the basic data of all finite element models with material parameters to obtain a set of node coordinates; Discretize the coordinates in the set of node coordinates into equal-width intervals and construct a node density histogram; Identify the sealing peak value in the node density histogram and use the coordinates of the density peak value that meets the preset conditions as the tower column position; Calculate the lengths of the left and right spans of the cable-stayed bridge based on the location of the towers; Based on the left and right span lengths, the node coordinate set is symmetry corrected to obtain the corrected node coordinate set; The corrected set of node coordinates is updated to the finite element model base data with material parameters to obtain the corrected finite element model base data.

[0063] In this embodiment, based on the left span length and the right span length, the node coordinate set is symmetry corrected to obtain the corrected node coordinate set, including: For a symmetrically arranged cable-stayed bridge, compare the lengths of the left span and the right span. If they are not equal, perform a proportional scaling correction on the coordinates of the right span nodes to obtain a preliminary set of corrected node coordinates. Perform mirror symmetry correction on the right span node in the preliminary correction node coordinate set so that the right span node is precisely mirrored on the corresponding left span node, thus obtaining the final correction node coordinate set.

[0064] Specifically, the X-coordinate node density histogram is constructed as follows: Discretize the X-coordinates of all nodes into equal-width intervals (binwidth=0.1m), and count the frequency of nodes within each interval. .

[0065] Since the tower column contains multiple vertical segments at the same X coordinate, while the main beam has only 1-2 nodes at each X position, the node density at the tower column position is significantly higher than at other positions.

[0066] Automatic tower location identification: Scan the node density histogram to identify nodes that meet the requirements. and The X-coordinate (excluding density concentration in the beam end support area) is used as the tower column location. Let the identified tower location X-coordinate be... (Sorted in ascending order).

[0067] For a double-tower cable-stayed bridge ,set up .

[0068] Span ratio calculation and mirror symmetry correction: Calculate the left span length right span length .like (Tolerance 0.05m), then perform X-coordinate proportional scaling correction on all nodes and component endpoints located to the right of tower B: Where x' is the corrected X-coordinate of the node, and x is the X-coordinate of any node before correction. B The X coordinate is the rightmost column (the rightmost one).

[0069] Mirror symmetry correction: After proportional scaling correction, a precise mirror alignment is further performed. For each right-side node's X coordinate... Search for the closest mirror image target position in the set of X coordinates of the left node, i.e., search the left node. Make and Minimize the difference, then... Revised to .

[0070] In this embodiment, a node density histogram of the X coordinate (bin=0.1m) was constructed, revealing 22 nodes at X=55m and 22 nodes at X=155m (tower B), while the main beam region has only 1-2 nodes at each X position. Tower A is automatically identified as being located at X=55m, and tower B as being located at X=155m.

[0071] Calculate the left span right cross Since the spans are equal, no proportional scaling is required. Perform mirror symmetry correction: for the X coordinate of each node in the right span, search for the closest mirror target of the left node.

[0072] Comparison before and after correction: Node 10 changes from 173.333m to 175.000m, Node 11 changes from 191.667m to 195.000m, and all right-spanning child nodes 50-57 are precisely mirrored and aligned.

[0073] Finite element analysis and adaptive back-off of the multi-stage solver were performed on the corrected finite element model base data to obtain complete finite element analysis results for the cable-stayed bridge.

[0074] In this embodiment, finite element analysis and multi-level solver adaptive backoff are performed on the corrected finite element model base data to obtain complete finite element analysis results for the cable-stayed bridge, including: Construct a multi-level solver chain that includes a master solver and at least one standby solver; Within each load step of the nonlinear finite element analysis, the solution is performed using the solver at the current level; If the current solver fails to converge, it will automatically switch to the next level solver to continue solving the current load step; If the current solver converges successfully, it will return to the master solver in the next load step to solve the problem. After solving all load steps, the finite element analysis data is extracted to obtain the complete finite element analysis results of the cable-stayed bridge.

[0075] Specifically, based on the semantic enhancement results mentioned above, an OpenSeesPyPython script for finite element analysis is automatically generated, including: Material definition: The tower column uses Concrete01 (core concrete f) c =35.5MPa, concrete protective layer) and Steel02 (HRB400 steel reinforcement f y =400MPa), the main beam and cables are made of Elastic material.

[0076] Section definitions: The tower column adopts a fiber section (concrete patch + steel layer), the main beam adopts an elastic section (A, Iy, Iz, J are calculated based on the section dimensions), and the cables adopt an elastic section (stiffness E = 1.95 × 10⁻⁶). 11 Pa).

[0077] Element generation: 21 segments of tower column × 2 towers = 42 nonlinearBeamColumn elements, 42 elasticBeamColumn elements of main beam, 8 corotTruss elements of cable, for a total of 92 elements.

[0078] Constraints and Loads: The tower base is fully fixed (6-DOF constraint), and the beam ends are hinged (translational constraint only). The dead load is a uniformly distributed line load (vertically downward) on the beam elements. The cable prestress is automatically applied at the material constitutive level via InitStressMaterial. No additional load modes are required.

[0079] Single-stage static analysis, 200-step LoadControl integration.

[0080] Problem Analysis: When performing finite element analysis on cable-stayed bridges using the large-scale finite element analysis software OpenSees, the tower columns are constructed using nonlinear fiber beam-column elements. The cross-sections of these elements are composed of fibers from the Concrete01 concrete constitutive model and the Steel02 steel reinforcement constitutive model (Giuffré-Menegotto-Pinto model). The softening segment of the concrete constitutive model and the Bauschinger effect of the steel reinforcement introduce strong nonlinearity.

[0081] Three-level solver adaptive backoff chain: This invention constructs a three-level solver backoff chain, which automatically switches the solution strategy based on the convergence state: Level 1 (Master Solver): NewtonLineSearch + BandGeneral System Solver + Transformation Constraint Processor + RCM Node Numbering Optimization + NormDispIncr Convergence Criterion (Tolerance 1×10⁻⁶) -6 (Maximum 200 iterations).

[0082] Second level (backup 1): If the first level fails to converge in any load step (the tolerance is not met even after more than 200 iterations), it automatically switches to the Krylov Newton solver + BandGeneral system solver + the same convergence criterion. The Krylov Newton solver accelerates convergence by constructing Krylov subspaces and has better robustness to ill-conditioned stiffness matrices.

[0083] Level 3 (Backup 2): If Level 2 still fails to converge, switch to Newton solver + SparseGeneral sparse solver + stricter convergence criteria (tolerance 1×10). -5 ).

[0084] The three-level solver switches at the load step level. That is, a convergence failure in a certain load step only triggers the degradation of the solver in that step. Subsequent load steps will still start from the first-level solver to try again, thereby minimizing performance loss.

[0085] In this embodiment, the single-stage static analysis involves 200 LoadControl integration steps (step size 0.005), requiring only one load mode (dead load, vertically downward). The cable prestress is automatically applied at the material constitutive level through InitStressMaterial, without the need for additional external load modes.

[0086] Because the cable element retains the true elastic modulus (E=1.95×10⁻⁶), 11 Pa) participates in the overall stiffness matrix, which helps to eliminate the matrix ill-conditioned problem caused by near-zero cable stiffness in the external force method. The main solver NewtonLineSearch, together with the SparseGeneral system solver, successfully converged for all load steps without the need to activate the backup solver.

[0087] Calculation results: Lateral displacement at the top of the tower: TowerA: +8.5mm, TowerB: -8.5mm (both tilting towards the mid-span, strictly symmetrical); Vertical deflection at mid-span of the main beam: approximately -5.4mm (downward); Vertical compression of the tower column: approximately 1.5mm.

[0088] In other embodiments, post-processing of the results is also included: After the solution is completed, extract the data from the recorder output file and generate a visualization report (outputting the structural response of the combination of dead load and cable force): main beam bending moment My; main beam deflection; main beam axial force; tower column lateral displacement: generate charts (4 sub-charts: bending moment chart, axial force chart, deflection chart, tower column displacement chart) and text data report.

[0089] Specifically, this invention integrates six modules into a single unit: intelligent layer classification, multi-strategy semantic enhancement, spatial hash topology reconstruction, text-geometric association, tower location detection and symmetry correction, and multi-level solver adaptive backoff. Furthermore, data is transferred between modules through a unified component data structure (component data with structural semantics) and node-unit topological relationships, allowing the output of preceding modules to be directly used as input for subsequent modules without any manual format conversion or intermediate intervention.

[0090] It should be noted that in the tower location detection and symmetry correction steps of the present invention, mirror symmetry correction is only applicable to bridges designed to be symmetrically arranged; for asymmetrical bridges, the system can automatically skip the correction step or only output the tower location detection result, and the other steps are unaffected.

[0091] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for intelligent modeling of cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives, characterized in that: include: Obtain DXF format cable-stayed bridge CAD vector drawings, and perform intelligent layer classification and multi-strategy semantic enhancement on the DXF format cable-stayed bridge CAD vector drawings to obtain component data with structural semantics; The component data with structural semantics is subjected to spatial hashing and automatic topology reconstruction to obtain node-unit topology relationship data; Text-geometric spatial association and intelligent material recognition are performed on the node-unit topological relationship data to obtain the basic data of the finite element model with material parameters; For symmetrically arranged cable-stayed bridges, tower location detection and symmetry correction are performed on the finite element model basic data with material parameters to obtain the corrected finite element model basic data. Finite element analysis and multi-level solver adaptive backoff are performed on the corrected finite element model base data to obtain complete finite element analysis results for the cable-stayed bridge.

2. The intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives as described in claim 1, is characterized in that, The DXF format cable-stayed bridge CAD vector drawing is subjected to intelligent layer classification and multi-strategy semantic enhancement to obtain component data with structural semantics, including: Based on a pre-built keyword library, weighted keyword matching is performed on all layers of the DXF format cable-stayed bridge CAD vector drawing to obtain layer matching scores. Based on the layer matching score, the layers are automatically classified into structure layers, annotation layers, material text layers, and ignored layers; The structural layers are further subdivided into beam, tower / column, and cable types; Different semantic enhancement algorithms are used for different structural component types to identify beam components, tower column components, and cable components respectively, and to obtain component data with structural semantics.

3. The intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives as described in claim 2, is characterized in that, Different semantic enhancement algorithms are used for different structural component types to identify beam components, tower-column components, and cable-stayed components, including: The elements in the beam layer are processed using a parallel line detection algorithm to identify beam components and extract the beam's centerline and cross-sectional width; The primitives in the tower column layer are processed using a vertical line segment clustering algorithm to identify tower column components and extract the center coordinates of the bottom and top of the tower column; The primitives in the cable-type layer are processed using a diagonal line segment direction grouping algorithm to identify cable components and extract the coordinates of the cable's anchorage endpoints.

4. The intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives as described in claim 1, is characterized in that, The component data with structural semantics is subjected to spatial hashing for automatic node topology reconstruction to obtain node-unit topology relationship data, including: Extract the endpoints of all components from the component data with structural semantics to obtain the component endpoint set; Based on the set of component endpoints, a three-dimensional spatial mesh index is constructed; Map all endpoints in the component endpoint set to the grid cells corresponding to the three-dimensional spatial grid index; The endpoints within the same grid cell are merged into shared nodes and assigned global node IDs sequentially to obtain a set of shared nodes; Traverse all components in all component data with structural semantics, establish corresponding finite element elements based on the shared nodes connected to the components, and obtain node-element topology relationship data.

5. The intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives as described in claim 1, characterized in that, By performing text-geometric spatial association and intelligent material recognition on the node-element topology relationship data, the basic data of the finite element model with material parameters is obtained, including: Extract all text entities from the DXF format CAD vector drawing of the cable-stayed bridge to obtain a text entity set; For each text entity in the text entity set, extract its insertion point coordinates and text content to obtain text attribute data; Based on the insertion point coordinates in the text attribute data, search all components within a preset radius of each text entity, establish a spatial association between each text entity and the nearest component, and obtain text-component association data; The text content in the text-component association data is parsed to obtain the component's dimensions, material grade, and load information; The dimensions, material grade, and load information are assigned to the corresponding components to obtain the basic data of the finite element model with material parameters.

6. The intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives as described in claim 5, is characterized in that, The text content in the text-component association data is parsed to obtain the component's dimensions, material grade, and load information, including: Using a hierarchical regular expression rule library, dimensional parameters, material grades, and load information are extracted from the text content of the text-component association data. Based on the extracted material grade, query the pre-constructed material mechanical property database to obtain the corresponding material mechanical parameters; By integrating the dimensional parameters, material mechanical parameters, and load information, the properties of the component are obtained.

7. The intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives as described in claim 1, characterized in that, For symmetrically arranged cable-stayed bridges, tower location detection and symmetry correction are performed on the finite element model foundation data with material parameters to obtain corrected finite element model foundation data, including: Extract all node coordinates from the basic data of all finite element models with material parameters to obtain a set of node coordinates; Discretize the coordinates in the set of node coordinates into equal-width intervals to construct a node density histogram; Identify the sealing peak value in the node density histogram, and take the coordinates corresponding to the density peak value that meets the preset conditions as the tower column position; Calculate the lengths of the left and right spans of the cable-stayed bridge based on the locations of the towers. Based on the left span length and the right span length, the node coordinate set is symmetry corrected to obtain the corrected node coordinate set; The corrected set of node coordinates is updated to the finite element model base data with material parameters to obtain the corrected finite element model base data.

8. The intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives as described in claim 7, is characterized in that, Based on the left and right span lengths, the node coordinate set is symmetry corrected to obtain the corrected node coordinate set, including: For a symmetrically arranged cable-stayed bridge, the lengths of the left span and the right span are compared. If they are not equal, the coordinates of the right span nodes are corrected by proportional scaling to obtain a preliminary set of corrected node coordinates. Perform mirror symmetry correction on the right span node in the preliminary correction node coordinate set, so that the right span node is accurately mirrored on the corresponding left span node, to obtain the final correction node coordinate set.

9. The intelligent modeling method for cable-stayed bridges based on automatic parsing of CAD drawings and semantic enhancement of primitives as described in claim 1, characterized in that, Finite element analysis and adaptive backoff of a multi-stage solver are performed on the corrected finite element model base data to obtain complete finite element analysis results for the cable-stayed bridge, including: Construct a multi-level solver chain that includes a master solver and at least one standby solver; Within each load step of the nonlinear finite element analysis, the solution is performed using the solver at the current level; If the current solver fails to converge, it will automatically switch to the next level solver to continue solving the current load step; If the current solver converges successfully, it will return to the master solver in the next load step to solve the problem. After solving all load steps, the finite element analysis data is extracted to obtain the complete finite element analysis results of the cable-stayed bridge.

Images