A method and system for parsing multi-source CAD native formats and constructing a unified data model

By constructing a unified data model through a self-developed geometric kernel and plug-in parsing mechanism, the problem of autonomous controllability of multi-source CAD data in an intranet offline environment was solved, enabling cross-format identification and incremental updates of the same part, reducing computational overhead and geometric accuracy loss.

CN123068489APending Publication Date: 2026-11-10CHANGSHA JINGYUE GALAXY TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611239695.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-17
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies cannot achieve native format parsing and unified data model construction of multi-source CAD data in an offline intranet environment, resulting in insufficient autonomy and controllability. Multiple format conversions cause geometric accuracy loss and topology destruction, making it impossible to identify the same part across formats. After design changes, a full rework is required, resulting in high calculation costs.

Method used

It adopts a self-developed geometric kernel to parse CAD files, realizes multi-source format parsing through a plug-in parsing mechanism, constructs a unified data model including geometric layer, structural layer and semantic layer, uses globally unique identifiers and geometric fingerprints for indexing, supports incremental updates and surface-level geometric comparison, and generates multi-precision lightweight mesh representations.

Benefits of technology

It enables multi-source CAD data management in an offline intranet environment, avoiding geometric accuracy loss and topology destruction, supporting the identification of the same part across formats, processing only the changed parts, and reducing computational overhead and data management complexity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123068489A_ABST
    Figure CN123068489A_ABST
Patent Text Reader

Abstract

This invention relates to the field of CAD data processing technology, and in particular to a method and system for parsing multi-source CAD native formats and constructing a unified data model. The method includes: receiving CAD files; identifying the format of the CAD files; scheduling a parser for the corresponding format based on the identification results; the parser parses the files based on a self-developed geometric kernel, converting heterogeneous CAD data into a unified intermediate geometric representation; calculating the content hash value based on the part geometric data as a globally unique identifier, and building an associated data model containing geometric, structural, and semantic layers; calculating the geometric fingerprint of each part and constructing multiple types of indexes; when importing a new version of the CAD file, comparing the globally unique identifier only processes the changed parts, performing surface-level geometric comparison on the modified parts, and outputting the difference location results. This invention achieves native parsing and unified data modeling of multi-source CAD formats, supports automatic deduplication and geometric retrieval of cross-format parts, and allows incremental processing and precise location of geometric differences in version changes.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of CAD data processing technology, and in particular to a method and system for parsing multi-source CAD native formats and constructing a unified data model. Background Technology

[0002] In the product development and production processes of discrete manufacturing, 3D CAD models are core data assets. Due to the differences in design software between upstream and downstream industries, enterprises generally have multiple CAD data formats, including neutral exchange formats such as STEP, IGES, JT, and 3DXML, native formats of commercial geometry kernels such as x_t based on the Parasolid kernel, and proprietary formats of software such as SOLIDWORKS, CATIA, and Rhino. Unified management of multi-source heterogeneous data has become a common challenge in the industry.

[0003] Existing multi-format CAD data management solutions mainly fall into two categories. The first category is format conversion solutions based on commercial geometric kernels or cloud services, which perform bidirectional format conversion between different CAD formats or use a neutral format as an intermediary to achieve multi-format mutual conversion. This solution has three drawbacks: First, it relies on foreign commercial kernel authorization or cloud conversion services, requires connection to an external network, and cannot operate independently on a physically isolated enterprise intranet, making it difficult to meet the requirements of independent control and data not leaving the network in a domestic IT innovation environment; second, multiple conversions are prone to geometric accuracy loss and topology damage due to differences in the geometric expression methods and accuracy tolerance systems of different kernels, and a large amount of product manufacturing information and assembly constraint information is easily lost; third, bidirectional conversion between N formats requires the configuration of N×N converters, and the maintenance cost increases exponentially with the number of formats.

[0004] The second type is a product data management solution based on file archives. This solution stores CAD files as binary attachments and only indexes external metadata such as filename, version number, and approval status, without parsing the internal geometric and structural information. Its drawbacks include: inability to identify the same part object across formats and files, easily generating a large amount of duplicate data; lack of support for geometric-based retrieval, comparison, and deduplication; and the inability to locate changed parts after design changes, requiring a complete rework of downstream lightweighting conversion, simulation preprocessing, and other processes, resulting in high computational costs.

[0005] In summary, existing technologies lack a multi-source CAD data management method that can run in an offline intranet environment, supports native parsing of multiple formats, has a unified data model and geometric retrieval capability, and enables incremental updates. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for parsing multi-source CAD native formats and constructing a unified data model.

[0007] To achieve the above objectives, in a first aspect, the present invention provides a method for parsing multi-source CAD native formats and constructing a unified data model. The method includes the following steps: receiving a CAD file, performing format recognition on the CAD file to obtain a recognition result, and scheduling a parser for the corresponding format according to the recognition result; the parser parses the CAD file based on a self-developed geometric kernel to obtain CAD data, and converts the CAD data from different sources into a unified intermediate geometric representation; calculating a content hash value based on the part geometric data in the intermediate geometric representation, using the content hash value as a globally unique identifier, constructing a data model including a geometric layer, a structural layer, and a semantic layer based on the intermediate geometric representation, and associating each layer with the globally unique identifier; calculating the geometric fingerprint of each part, and constructing multiple indexes based on the data model and the geometric fingerprint; when importing a new version of the CAD file of the same product, comparing the globally unique identifiers of the new and old versions, only performing data entry, fingerprint calculation, and index update for the parts that have been added or modified, and performing a surface-level geometric comparison for the modified parts, and outputting the geometric difference location result. This invention enables native parsing and unified data modeling of multi-source CAD formats without relying on multiple format conversions. It can identify the same part across formats and automatically remove duplicates, supports multi-dimensional geometric retrieval, and only processes changed parts and accurately locates surface-level differences when versions change, effectively reducing the computational overhead of data management and version iteration.

[0008] Optionally, the step of receiving the CAD file, performing format recognition on the CAD file to obtain a recognition result, and scheduling a parser for the corresponding format based on the recognition result includes: after receiving the CAD file to be imported, reading the binary signature and internal structural features of the CAD file header to perform format recognition, obtaining a recognition result, and scheduling the parser for the corresponding format based on the recognition result. The parser adopts a plug-in registration mechanism. This invention identifies CAD formats through file header features, unaffected by extension tampering; the plug-in parsing mechanism facilitates the addition of new format extensions, effectively reducing maintenance costs.

[0009] Optionally, the parser parses the CAD file based on a self-developed geometric kernel to obtain CAD data, and converts the CAD data from different sources into a unified intermediate geometric representation. This includes: after completing data reading, the parser converts the CAD data output by different CAD systems into a unified intermediate geometric representation according to a unified data mapping rule. During the conversion process, the multi-dimensional geometric data of the different CAD systems are normalized, and a mapping relationship between source tolerance and global tolerance is established. This invention converts heterogeneous CAD data into a unified intermediate geometric representation, eliminates kernel differences through tolerance normalization and mapping, and avoids geometric accuracy loss and topology misjudgment.

[0010] Optionally, the intermediate geometric representation includes topological information, geometric information, assembly structure information, attribute information, and product manufacturing information. The intermediate geometric representation of this invention fully covers topological, geometric, assembly, attribute, and product manufacturing information, providing comprehensive data support for subsequent retrieval, comparison, and engineering applications.

[0011] Optionally, the step of calculating the content hash value based on the part geometric data in the intermediate geometric representation and using the content hash value as a globally unique identifier includes: performing normalized serialization processing on the part geometric data, removing fields irrelevant to the geometric substance, calculating the content hash value based on the normalized serialized part geometric data, and using the content hash value as the globally unique identifier. This invention generates content hash identifiers based on normalized geometric data, enabling cross-format and cross-file identification of the same part, naturally achieving geometric data deduplication and unique identification.

[0012] Optionally, the step of constructing a data model based on the intermediate geometric representation, including a geometric layer, a structural layer, and a semantic layer, and associating each layer with the globally unique identifier, includes: the geometric layer storing multiple geometric representation data of the part, with the multiple geometric representation data corresponding to the same part being associated through the same globally unique identifier; the structural layer expressing the assembly hierarchy relationship in a graph structure, adopting a storage method that separates part definitions from part instances, with the multiple geometric representation data of the same part stored only once in the part definition, and the part instance storing space transformation location information; and the semantic layer storing part attributes, engineering semantic information, and material mapping relationships. This invention adopts a three-layer data model architecture, significantly compressing the storage space for repetitive assembly through the storage method that separates part definitions from instances, resulting in a clear and easily manageable data hierarchy.

[0013] Optionally, the calculation of the geometric fingerprint for each part involves constructing multiple indexes based on the data model and the geometric fingerprint. This includes calculating the geometric fingerprint for each part, where the geometric fingerprint is a multi-dimensional feature vector possessing rigid body transformation invariance and scale invariance, including part scale features, bounding box features, moment of inertia features, and surface distribution features. Based on the data model and the geometric fingerprint, a geometric similarity index, a structural relationship index, and an attribute semantic index are constructed. This invention constructs a geometric fingerprint with rigid body and scale invariance, combining three types of indexes to achieve multi-dimensional and efficient retrieval, supporting the search for similar parts and the tracing of structural relationships.

[0014] Optionally, the step of performing face-level geometric comparison on the modified parts and outputting geometric difference location results includes: converting the parts to be compared into attribute adjacency graphs, with individual faces as graph nodes and face adjacency relationships as graph edges, and encoding the geometric and topological attributes of the corresponding faces; calculating the face-level similarity matrix of the parts to be compared, and solving the global maximum weight face matching relationship using the Hungarian algorithm; combining the face matching ratio and the topological adjacency consistency rate to determine the degree of part change, and outputting the geometric difference location results, which include a set of difference faces composed of newly added faces, deleted faces, and faces with modified geometric parameters. This invention achieves face-level geometric comparison based on attribute adjacency graphs and a global matching algorithm, which can accurately determine the degree of part change and locate difference faces, supporting design change review.

[0015] Optionally, the method for parsing multi-source CAD native formats and constructing a unified data model further includes: discretizing the geometric data of the imported parts into a multi-precision mesh to generate a multi-level detail lightweight mesh representation; establishing a bidirectional mapping relationship between the lightweight mesh units and the source geometric surfaces; the multi-level detail lightweight mesh representation and the geometric data share the same globally unique identifier, and are incrementally generated as the parts are incrementally updated. This invention generates multi-precision lightweight meshes and establishes a bidirectional mapping, balancing browsing efficiency and data accuracy, supporting incremental generation, and effectively reducing the computational overhead of version iteration.

[0016] Secondly, this invention provides a system for parsing multi-source CAD native formats and constructing a unified data model. The system executes the method for parsing multi-source CAD native formats and constructing a unified data model provided by this invention. The system includes an input device, an output device, a processor, and a memory, which are interconnected. The memory stores a computer program, which includes program instructions, and the processor is configured to call the program instructions. This invention provides an integrated CAD data management system capable of independently completing multi-source format parsing and model construction, meeting the deployment requirements of enterprise intranet autonomy and controllability, and ensuring data remains within the network. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the method for parsing multi-source CAD native formats and constructing a unified data model according to an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the three-layer structure of the unified data model in an embodiment of the present invention.

[0019] Figure 3 This is a system framework diagram for multi-source CAD native format parsing and unified data model construction according to an embodiment of the present invention. Detailed Implementation

[0020] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.

[0021] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale.

[0022] Please see Figure 1 One embodiment of the present invention provides a method for parsing multi-source CAD native formats and constructing a unified data model. All steps in the method are completed offline on an enterprise intranet server without relying on any external network services. Specifically, the method includes the following steps: S1. Receive a CAD file, perform format recognition on the CAD file, obtain the recognition result, and schedule a parser for the corresponding format according to the recognition result.

[0023] In this embodiment, after receiving the CAD file to be imported, the system does not determine the format based on the file extension, but instead reads the binary signature in the file header and the internal structural features for identification.

[0024] Specifically, for STEP files, the "ISO-10303-21" identifier string in the file header is detected, and the FILE_SCHEMA field in the HEADER segment is further parsed to distinguish between AP203, AP214, and AP242 application protocols; for JT files, the version identifier byte sequence in the file header is detected; for Parasolid x_t files, the Parasolid identifier in the text header is detected; for IGES files, the fixed 80-column record format and S-segment start marker are detected; for proprietary formats such as SOLIDWORKS, CATIA, and Rhino, the magic number in the compound document or proprietary file header is detected.

[0025] After identification, the corresponding native parser is scheduled according to the registry under "Format Type → Parser Plugin". The parser uses a unified plugin interface, which is defined as follows: input file byte stream, output intermediate geometric representation defined in the following steps. When adding a new format, only one parser plugin needs to be added, without modifying the rest of the system, thus reducing the N×N scale conversion problem between N formats to an N×1 scale parsing problem.

[0026] S2. The parser parses the CAD file based on its self-developed geometric kernel to obtain CAD data, and converts the CAD data from different sources into a unified intermediate geometric representation.

[0027] In this embodiment, the scheduler selects a parser for the corresponding file format, and the parser reads the model data based on a self-developed geometric kernel.

[0028] Specifically, the self-developed geometric kernel first defines a unified CAD parser interface. All types of file parsers implement this interface and complete registration during the system initialization phase, establishing a mapping relationship between file feature identifiers and parser instances. When importing a model, the scheduler first reads the binary signature of the file header and internal structural features to identify the model type, then retrieves the corresponding parser from the parser registry, and calls the unified parsing interface to execute model parsing, realizing automatic scheduling and switching between different CAD formats.

[0029] For open standard formats such as STEP and IGES, the parser parses the file content according to standards such as ISO 10303 and GB / T 14213-2008. The first stage performs lexical and syntactic parsing, structuring the plaintext into a "mapping table from instance ID to entity parameters" and reading the original byte sequence. The second stage completes semantic assembly and topology reconstruction. Based on the Brep, Shell, Face, Loop, and Edge reference chains, it constructs a complete hierarchy from Body to Vertex. It solves the reuse problem of multiple faces sharing the same boundary through a global "ID to entity" mapping table and generates forward or reverse edges based on the direction marker (IGES requires topology inference based on group entities or geometric overlap). The third stage performs geometric evaluation. Analytical surfaces are directly instantiated using a positioning coordinate system and constant parameters. For NURBS freeform surfaces, the order, node vectors, and weighted control point mesh are read to construct a precise surface object that can be evaluated and discretized.

[0030] For proprietary CAD formats such as SOLIDWORKS, CATIA, and Creo, the parser performs native structured parsing of CAD model data, tailored to the data organization structures of different CAD systems. Specifically, it extracts the feature tree of the CAD file. For SOLIDWORKS, it parses the parent-child dependencies between features to reconstruct the modeling process; for CATIA, it parses its unique unordered organization of "geometric feature sets"; and for Creo, it extracts its regenerated feature sequences. At the assembly level, it restores the assembly hierarchy by parsing the instance matrix mapping table within each format. During parsing, it reads topological information such as parts, assemblies, entities, faces, edges, and vertices, while simultaneously acquiring surface parameters, parameter domains, colors, names, feature identifiers, assembly levels, and attribute data, and restoring the adjacency and reference relationships between each topological entity.

[0031] To address the systemic differences in data representation across different CAD kernels, this solution employs hierarchical normalization to construct a unified data foundation that ensures geometric and topological consistency and complete assembly relationships. The necessity of normalization lies in the fact that different CAD platforms employ varying underlying conventions, such as length units, modeling tolerances, coordinate system definitions, and surface parameter domains. Direct cross-format data fusion would introduce systemic deviations in geometric determinations (e.g., vertex coincidence, edge coincidence, face adjacency) and assembly positioning, leading to topology reconstruction errors or loss of assembly positional relationships. Specifically, this includes unit conversion, absolute / relative tolerance mapping, and parameter domain standardization. The normalization of the precision tolerance system involves: reading the modeling tolerances declared in the source file, unifying vertex coincidence and boundary closure determinations to the system-defined global tolerance benchmark, and recording the mapping relationship between the source tolerance and the global tolerance to avoid topology misjudgments caused by inconsistent tolerances during cross-format assembly. Through the above normalization process, the differences in data representation between different CAD platforms are eliminated, and a unified data structure of Body, Face, Edge, and Vertex and their topological relationships are constructed, providing a consistent data foundation for subsequent geometric feature recognition, attribute adjacency graph construction, and face-level difference matching.

[0032] The intermediate geometric representation specifically includes: (1) Topological information: The B-rep topology and the adjacency relationships between each level are completely preserved according to the "volume-shell-face-loop-edge-vertex" hierarchy; (2) Geometric information: The analytical surfaces (plane, cylindrical, conical, spherical, torus, etc.) in each format are kept in analytical form, and the freeform surfaces are unified into NURBS expression; (3) Assembly structure information: Parse the assembly tree, including sub-assemblies, part definitions, part instances and the spatial transformation matrix of each instance; (4) Attribute information: Parse key-value pairs such as part name, number, material, and custom attributes; (5) Product Manufacturing Information (PMI): For formats that carry product manufacturing information (such as STEP AP242, JT), analyze the dimensions, tolerances, datums and other annotation objects, and retain the relationship between each annotation and the geometric elements (faces, edges) to which it is attached.

[0033] During the parsing process, the original identifier (such as STEP entity number) of each geometric element in the source file is recorded, forming a bidirectional traceability table of "source file element - intermediate representation element", which is used for subsequent problem backtracking and incremental updates.

[0034] S3. Calculate the content hash value based on the part geometric data in the intermediate geometric representation, use the content hash value as the globally unique identifier (GUID) of the part, build a data model including a geometric layer, a structural layer and a semantic layer based on the intermediate geometric representation, and associate each layer with the globally unique identifier.

[0035] In this embodiment, the intermediate geometric representation is organized into a three-layer unified data model, such as... Figure 2 As shown: (1) Geometry layer: Stores B-rep data, NURBS geometry data, Level of Detail (LOD) mesh representation, and geometric fingerprint for each part, including but not limited to each part. Multiple representations of the same part (precise B-rep and meshes at each level) are associated with the same identifier, achieving "one geometry, multiple representations"; (2) Structure Layer: Storage includes, but is not limited to, the assembly hierarchy expressed in a directed acyclic graph, part definitions, part instances, transformation matrices, and reference reuse relationships. Part definitions and part instances are stored separately: part definition nodes store geometric data references, and instance nodes store transformation matrices and references to the corresponding definition nodes. When the same part is reused M times in the assembly, only one copy of the geometric data is stored, and M instances store only M sets of transformation matrices and definition references, which can significantly reduce the storage volume in large-scale repetitive assembly scenarios such as standard parts and arrays; (3) Semantic layer: Stores, but is not limited to, attributes, product manufacturing information, and material information, and establishes a mapping relationship with the enterprise's bill of materials code.

[0036] The three layers are linked by a globally unique identifier. This globally unique identifier is generated using a content hashing rule: Normalized serialization processing is performed on the part's geometric data, including serializing the B-rep data according to deterministic rules, removing fields irrelevant to the geometric substance such as timestamps and author information, calculating the content hash value based on the normalized serialized part's geometric data, and using this content hash value as the globally unique identifier for the part definition. The SHA-256 hashing algorithm adopts the NISTFIPS PUB 180-4 standard. Therefore, parts with identical geometric content, regardless of their format or file, will generate the same globally unique identifier, naturally achieving cross-format and cross-file identification and deduplication of the same part.

[0037] The specific B-rep data serialization rules are shown in Table 1, and the serialization format is shown in Table 2.

[0038] Table 1 B-rep Geometric Canonical Sequence Rules

[0039] Table 2 B-rep geometry canonical serialization byte format

[0040] S4. Calculate the geometric fingerprint of each part, and construct various indexes based on the data model and the geometric fingerprint.

[0041] In this embodiment, a geometric fingerprint is constructed for each part to describe its overall geometric features and support shape-based similarity retrieval. The geometric fingerprint is represented as a feature vector composed of multiple geometric statistical features, including part scale features (volume, surface area, total edge length), bounding box features, invariant moment of inertia with the centroid as the origin, and a histogram of surface distribution composed of the number and area ratio of various surface types (planes, cylinders, cones, freeform surfaces, etc.). To eliminate the influence of part spatial position, rotational attitude, and scale differences on the retrieval results, the relevant geometric quantities are constructed using rigid body transformation invariants, sorted and normalized, and scale normalized during feature extraction, ensuring that the generated geometric fingerprint can stably represent the essential geometric features of the part. Specifically, the feature vector is represented as follows: ; in, Indicates scale characteristics. Indicates bounding box features. Represents the inertial feature descriptor, Let R represent the geometric composition features, R represent the set of real numbers, and d represent the dimension of the subvector.

[0042] Scale features Defined as: ; Where V is the volume of the part and S is the total surface area. This represents the total length of the side.

[0043] Bounding box features Defined as: ; in, , , The bounding box size after sorting is normalized to... .

[0044] Inertial characteristic matrix Defined as: ; The inertial eigenvalues ​​are obtained by performing eigenvalue decomposition on the inertial characteristic matrix. The inertial feature descriptors are obtained by sorting and normalizing the feature values. : ; Geometric composition features Defined as: ; ; ; in, N represents the number of faces of the k-th type of surface, and N represents the total number of faces. This represents the percentage of faces in the k-th type of surface. Let S represent the area of ​​the k-th type of surface, and S represent the total area. This represents the area percentage of the k-th type of surface.

[0045] Secondly, three types of indexes are established on top of the geometric fingerprint and data model: (1) Geometric similarity index: The geometric fingerprint of the part is mapped to a high-dimensional feature space, and a vector index structure supporting near-nearest neighbor queries is established. During the query process, the feature space distance is calculated based on the geometric fingerprint of the target part, and the set of candidate parts that meet the similarity constraints is quickly filtered to realize model-level similarity retrieval across CAD formats. For the returned candidate results, the unique identifier of the part and the accurate geometric matching results are further combined to distinguish between completely identical models, parameter variation models and shape similar models.

[0046] (2) Structural Relationship Index: A bidirectional association index between parts and assemblies is established for the assembly hierarchy structure. The reference relationship of parts in different products, subassemblies and assembly instances is recorded to realize fast reverse query from parts to product structure, and provide data support for design change propagation analysis, impact scope assessment and version management.

[0047] (3) Attribute semantic index: An inverted index is established for engineering attributes, design semantics and product manufacturing information in the unified data model. Multi-condition combination query is realized through attribute key value mapping, and fast retrieval based on information such as part name, number, material, manufacturing requirements, size constraints and engineering annotations is supported.

[0048] S5. When importing a new version of the CAD file for the same product, compare the global unique identifiers of the old and new versions, and only perform data entry, fingerprint calculation and index update for the parts that have been added or modified. Perform surface-level geometric comparison on the modified parts and output the geometric difference positioning results.

[0049] In this embodiment, when a new version of the same product file is imported again: first, the new file is parsed, and a globally unique identifier and geometric fingerprint are calculated for each part in the new file; the globally unique identifiers of each part in the new version are compared with the set of globally unique identifiers of the previous version in the library. Parts with different globally unique identifiers are determined to be newly added or modified, and geometric layer entry into the library, fingerprint calculation and index update are directly performed; for parts with the same globally unique identifiers, they are further compared on a face-by-face basis at the B-rep level.

[0050] Specifically, to achieve face-level difference recognition in the B-rep model, this invention proposes a geometric-topological joint matching method based on Attributed Adjacency Graph (AAG). First, the B-rep model is represented as an attributed adjacency graph G=(V, E), where each node v∈V corresponds to a topological face and its geometric attributes are encoded, including face type, area, parameterized geometric feature vector, and barycentric coordinates. Each edge e∈E represents the topological connection between two adjacent faces, and topological attributes such as dihedral angle and common edge type are recorded.

[0051] Furthermore, for models A and B to be compared, the comprehensive similarity between all candidate faces is calculated, taking into account face type consistency, area ratio, and geometric parameter similarity, with weights set to 0.4, 0.3, and 0.3 respectively, thus constructing a face-level similarity matrix. Based on this matrix, the Hungarian algorithm is used to solve the global maximum weight matching to obtain the optimal face mapping relationship in a one-to-one correspondence. To further ensure the reliability of the matching results, the consistency of the mapped adjacency relationship is verified, the topological adjacency preservation rate is statistically analyzed, and a model judgment rule is constructed in conjunction with the face matching rate: when the proportion of matching faces is not less than 95% and the adjacency relationship consistency rate is not less than 90%, the two models are determined to belong to the same part, and the existing geometric layer data, LOD mesh, and index in the library are directly reused without repeated parsing and input into the library; when the proportion of matching faces is between 60% and 95%, they are determined to be design change versions of the same part, and the set of differences in faces with added, deleted, and changed geometric parameters is further output; when the proportion of matching faces is less than 60%, the two models are considered to belong to different parts. This rule considers both geometric features and topological constraints, effectively avoiding mismatches caused by relying solely on geometric similarity, and improving the robustness and accuracy of identifying local design changes in complex CAD models.

[0052] The output set of changed surfaces serves as the geometric difference localization result, which is then used for Engineering Change Notice (ECN) review and downstream simulation and process stages to reprocess only the changed areas. As a result, the parsing and indexing overhead during version iterations is reduced from being proportional to the total assembly size to being proportional to the actual number of changed parts.

[0053] In one optional embodiment, the B-rep data of the imported parts is discretized into a multi-level LOD mesh representation according to multiple preset precision levels. During the mesh generation process, a bidirectional mapping relationship is established between triangular facets and source B-rep faces: any triangular facet can trace back to its source B-rep facet, and any B-rep facet can retrieve all the triangular facet sets it has discretized. This bidirectional mapping mechanism supports precise backtracking in lightweight display scenarios on the browser side—when the user picks up any mesh position in the viewport, the system can reverse locate the corresponding precise geometry and its associated product manufacturing annotation information, thereby ensuring the consistency between the "lightweight display model" and the "precise geometric data" throughout the entire process. The LOD mesh and the original B-rep data belong to the same geometric layer and share the same globally unique identifier; their incremental generation mechanism and incremental update strategy work together to achieve this.

[0054] All calculation processes involved in this invention are deployed and executed locally on an intranet server. The geometric kernel, format parser, and index structure are all self-developed and implemented, without relying on any external cloud interface calls, fully meeting the deployment requirements of the domestic IT innovation environment for independent control and data not leaving the network.

[0055] Please see Figure 3 In one optional embodiment, the present invention provides a system for parsing and constructing a unified data model from multiple sources of CAD native formats. The system includes an input device, an output device, a processor, and a memory, all interconnected. The memory stores a computer program, which includes program instructions. The processor is configured to invoke the program instructions to execute specific steps as described in the embodiments of the method for parsing and constructing a unified data model from multiple sources of CAD native formats provided by the present invention. The system for parsing and constructing a unified data model from multiple sources of CAD native formats provided by the present invention has a complete structure and stable operation, improving the overall applicability and engineering application value of the present invention.

[0056] In summary, the present invention mainly solves the following technical problems: (1) Solve the problems of existing solutions relying on foreign commercial geometric kernels, lacking independent controllability, and cloud conversion solutions requiring network access and unable to run offline on physically isolated intranets; (2) Solve the problems of geometric precision loss, topology destruction, and PMI information loss caused by multiple format conversions; (3) Solve the problems of lack of unified data model for multi-format CAD data, repeated storage of the same part across formats and inability to search, check for duplicates and compare based on geometric content; (4) Solve the problem that the changed parts cannot be located after the model is changed, and the downstream processing must be completely redone.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. 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 parsing multi-source CAD native formats and constructing a unified data model, characterized in that, include: Receive a CAD file, perform format recognition on the CAD file, obtain the recognition result, and schedule a parser for the corresponding format according to the recognition result; The parser parses the CAD file based on a self-developed geometric kernel to obtain CAD data, and converts the CAD data from different sources into a unified intermediate geometric representation; Calculate the content hash value based on the part geometric data in the intermediate geometric representation, use the content hash value as a globally unique identifier, build a data model including a geometric layer, a structural layer and a semantic layer based on the intermediate geometric representation, and associate each layer with the globally unique identifier. Calculate the geometric fingerprint of each part, and construct various indexes based on the data model and the geometric fingerprint; When importing a new version of the CAD file for the same product, the global unique identifiers of the old and new versions are compared. Only the newly added or modified parts are entered into the database, fingerprints are calculated and indexes are updated. The modified parts are then subjected to surface-level geometric comparison, and the geometric difference positioning results are output.

2. The method for parsing multi-source CAD native formats and constructing a unified data model according to claim 1, characterized in that, The process of receiving a CAD file, performing format recognition on the CAD file to obtain a recognition result, and scheduling a parser for the corresponding format based on the recognition result includes: After receiving the CAD file to be imported, the binary signature and internal structural features of the CAD file header are read for format recognition to obtain the recognition result. The parser corresponding to the format is scheduled according to the recognition result. The parser adopts a plug-in registration mechanism.

3. The method for parsing multi-source CAD native formats and constructing a unified data model according to claim 1, characterized in that, The parser parses the CAD file based on a self-developed geometric kernel to obtain CAD data, and converts the CAD data from different sources into a unified intermediate geometric representation, including: After the data reading is completed, the parser converts the CAD data output by different CAD systems into a unified intermediate geometric representation according to a unified data mapping rule. During the conversion process, the multi-dimensional geometric data of the different CAD systems are normalized and a mapping relationship between source tolerance and global tolerance is established.

4. The method for parsing multi-source CAD native formats and constructing a unified data model according to claim 1, characterized in that, The intermediate geometric representation includes topological information, geometric information, assembly structure information, attribute information, and product manufacturing information.

5. The method for parsing multi-source CAD native formats and constructing a unified data model according to claim 1, characterized in that, The step of calculating the content hash value based on the part geometry data in the intermediate geometric representation, and using the content hash value as a globally unique identifier, includes: The part geometric data is subjected to normalized serialization processing to remove fields that are irrelevant to the geometric substance. The content hash value is calculated based on the normalized serialized part geometric data and used as the globally unique identifier.

6. The method for parsing multi-source CAD native formats and constructing a unified data model according to claim 1, characterized in that, The data model built based on the intermediate geometric representation includes a geometric layer, a structural layer, and a semantic layer, and each layer is associated with the globally unique identifier, including: The geometry layer stores multiple geometric representation data of the part, and the multiple geometric representation data corresponding to the same part are associated through the same globally unique identifier; The structural layer expresses the assembly hierarchy in a graph structure and adopts a storage method that separates part definitions from part instances. The multiple geometric representation data of the same part are stored only once in the part definition, and the part instance stores spatial transformation position information. The semantic layer storage includes part attributes, engineering semantic information, and material mapping relationships.

7. The method for parsing multi-source CAD native formats and constructing a unified data model according to claim 1, characterized in that, The calculation of the geometric fingerprint of each part, based on the data model and the geometric fingerprint, constructs various indexes, including: The geometric fingerprint of each part is calculated. The geometric fingerprint is a multi-dimensional feature vector with rigid body transformation invariance and scale invariance, including part scale features, bounding box features, moment of inertia features, and surface distribution features. Based on the data model and the geometric fingerprint, a geometric similarity index, a structural relationship index, and an attribute semantic index are constructed.

8. The method for parsing multi-source CAD native formats and constructing a unified data model according to claim 1, characterized in that, The step of performing a surface-level geometric comparison on the modified part and outputting the geometric difference location result includes: The parts to be compared are converted into attribute adjacency graphs, with each face as a graph node and the adjacency relationship of the face as a graph edge, and the geometric and topological attributes of the corresponding face are encoded. Calculate the surface similarity matrix of the parts to be compared, and solve the global maximum weight surface matching relationship using the Hungarian algorithm; By combining the surface matching ratio and the topological adjacency consistency rate, the degree of part change is determined, and the geometric difference positioning result is output. The geometric difference positioning result includes a set of difference surfaces composed of newly added surfaces, deleted surfaces, and surfaces with modified geometric parameters.

9. The method for parsing multi-source CAD native formats and constructing a unified data model according to claim 1, characterized in that, Also includes: The geometric data of the inbound parts are discretized into meshes with multiple precision levels to generate a lightweight mesh representation with multiple levels of detail. Establish a two-way mapping relationship between lightweight mesh cells and source geometry; The multi-level detail lightweight mesh representation shares the same globally unique identifier with the geometric data and is incrementally generated as the part is incrementally updated.

10. A system for parsing multi-source CAD native formats and constructing a unified data model, characterized in that, The system includes an input device, an output device, a processor, and a memory, which are interconnected. The memory stores a computer program, which includes program instructions. The processor is configured to invoke the program instructions to execute the multi-source CAD native format parsing and unified data model construction method as described in any one of claims 1-9.