737 results about "Computer Aided Design" patented technology

The invention relates to the technical field of computer aided design, in particular to a text-driven CAD modeling method and system based on a diffusion and visual language model.The method comprises the steps that natural language text description is obtained, and CAD semantic features of the natural language text description are extracted; carrying out geometric standardization on the CAD semantic features by adopting a fine-tuning diffusion model, and generating a CAD view image conforming to engineering specifications; carrying out fusion by adopting a fine-tuned visual language model to generate a parameterized CAD construction sequence; a three-mode alignment mechanism is adopted, and the semantic consistency of the CAD semantic features, the CAD view images and the CAD construction sequences is checked; performing verification and post-processing on the CAD construction sequence, and outputting an executable Python code or STEP file; the CAD modeling method disclosed by the invention performs explicit modeling based on flexible modal description, and has the characteristics of high geometric constraint and high usability.

The invention relates to the technical field of computer vision, computer aided design and artificial intelligence, and discloses a geometric feature recognition method based on a large language model, which comprises the steps of three-dimensional model natural linguistization, large language model feature reasoning, incremental learning and optimization, and feature verification and output. According to the geometric feature recognition method based on the large language model, three-dimensional geometric data are converted into structured text description, and a joint representation space of'geometric parameter-topological relation 'is constructed, so that the large language model can better understand the meaning of geometric features and the structure of parts; the method comprises the following steps: analyzing topological association among geometric features by utilizing long context understanding ability of a large language model, capturing dependency among the features through an attention mechanism, realizing hierarchical recognition of the features in a complex assembly, dynamically updating a geometric feature knowledge base of the large language model by continuously learning new geometric cases and engineering specifications, and realizing hierarchical recognition of the features in the complex assembly. Adaptive recognition of novel features is supported, and the model does not need to be retrained.

The invention discloses a 3D art model intelligent generation system and method based on multi-source data fusion, and relates to the technical field of computer aided design and graph generation, and the system comprises a multi-source data input module which is used for receiving a conceptual design drawing, text description and a reference 3D model; the feature extraction module comprises an image feature extraction unit used for extracting a style feature vector and a structure contour vector of the conceptual design drawing; and the text semantic analysis unit is used for outputting a key attribute tag vector described by the text. According to the 3D art model intelligent generation system and method provided by the invention, through multi-source data input including the concept design drawing, the text description and the reference 3D model, the design intention is comprehensively fused, the limitation of a single data source generation model is solved, and the dynamic weight distribution strategy and the closed-loop optimization mechanism of the 3D art model intelligent generation system and method are high in practicability. The geometric shape and style characteristics of the generated model are highly consistent with the design intention, and the workload of manual later correction is reduced.

The invention belongs to the technical field of computer aided design, and provides a spinning equipment design assisting method and system based on model collaboration. The method comprises the steps that a collaborative design system is constructed, a large language model receives design requirements of a user, performs task decomposition and calls corresponding professional agent models to execute design subtasks, and the professional agent models feed simulation calculation results back to a closed-loop feedback engine; the closed-loop feedback engine analyzes feedback results of the professional agent models and generates attention prompt signals pointing to potential design conflicts or optimization bottlenecks on the basis of interdisciplinary parameter coupling relations; and the large language model performs multidisciplinary coupling analysis by using the attention prompt signal, dynamically adjusts task planning and design parameters by focusing on key parameters and key disciplines pointed by the attention prompt signal, and drives related professional agent models to perform directional iterative optimization. According to the method, the blind iteration problem can be effectively avoided, and the convergence speed of the design and the scheme quality are improved in a high-quality mode.

The invention discloses a high-stability hydroxysteroid dehydrogenase mutant, a co-immobilized enzyme construction method and application. According to the invention, a site-directed mutant library is constructed based on computer-aided design, and mutants with significantly improved thermal stability and greatly improved catalytic efficiency compared with wild type mutants are obtained through screening; meanwhile, a co-immobilized enzyme construction method based on electrostatic adsorption and chemical crosslinking is established, the hydroxysteroid dehydrogenase mutant and lactic dehydrogenase are co-immobilized, the co-immobilized enzyme can still keep high activity under high substrate concentration, and the conversion rate is still maintained to be 90% or above after 30 batches of co-immobilized enzyme are repeatedly used. And an important foundation is laid for industrial application of the hydroxysteroid dehydrogenase.

The invention discloses a pulmonary nodule boundary detection method based on semantic segmentation. The method comprises the following steps: generating an edge prior image set; obtaining a multi-scale boundary alignment feature map set and a deformation offset parameter set; a post-gating deformation offset parameter set is obtained, the multi-scale boundary alignment feature image set is updated, cross-scale feature fusion is executed, and a fusion feature image set is generated; establishing a node set and an edge set by utilizing node features in the fusion feature graph set and a spatial adjacency relationship, and initializing a node feature vector set; forming a fusion feature map set enhanced by multi-hop reasoning; and outputting a pulmonary nodule region mask prediction image set and a pulmonary nodule boundary probability prediction image set. According to the method, fracture between pseudo boundaries and slices is effectively inhibited, so that a three-dimensional structure prediction result is more coherent, and high-readability and high-consistency boundary segmentation output is provided for a clinical CAD (Computer Aided Design) system.

The invention discloses a three-dimensional physical field real-time prediction method and system based on geometric deep learning and physical constraint, and relates to the technical field of computer-aided engineering and artificial intelligence. The method comprises the following steps: directly extracting native boundary representation data (B-Rep) of a three-dimensional model from a computer aided design system; constructing a heterogeneous dual graph taking a parameterized curved surface as a graph node, skipping finite element grid division, and aggregating local and global topological features by using a graph neural network; in combination with a physical information driving mechanism, a partial differential equation (PDE) residual error is introduced as a loss function for constraint training, and generalization prediction of a novel geometric structure is realized; and finally, the physical field state quantity is predicted through direct regression and is rendered in real time. An incremental reasoning mechanism based on a local topology subgraph is adopted, millisecond-level physical field real-time feedback under design modification is achieved, and the method is suitable for scheme rapid screening and trend prediction in the initial stage of design.

Methods, systems, and apparatus, including medium-encoded computer program products, for computer aided design of physical structures include: obtaining a design space, design criteria, and boundary conditions for numerical simulation; defining an application of the boundary conditions to a voxelized mesh, including specifying a distribution of a load to nodes of voxels in the voxelized mesh that correspond to a surface of preserve geometry; and iteratively modifying a generatively designed three dimensional shape in the design space in accordance with the design criteria and a physical response of the modeled object determined by the numerical simulation performed using the application of the boundary conditions to the voxelized mesh, wherein the distribution of the total loading value during determination of the physical response ensures an equivalence between the total loading value and a sum of loading values distributed to the respective nodes of the voxels that correspond to the surface.

The invention belongs to the technical field of computer aided design and structural design, and discloses a multi-scale lattice structure and a design method thereof. The design method comprises the following steps: S1, designing a cubic unit cell with eight symmetrically distributed corner nodes according to a cubic lattice structure; s2, all the lattice unit cells are zoomed, one part of the lattice unit cells are zoomed into small lattice unit cells with the size being half of the original size, and the other part of the lattice unit cells are large lattice unit cells with the original size reserved; s3, in the macroscopic lattice structure, arranging the position relation of the small lattice unit cells and the large lattice unit cells; and S4, through a computer aided design method, adjusting the macroscopic lattice structure by adopting a large and small lattice unit cell connection mode based on rod diameter transition. A multi-scale large and small lattice unit cell arrangement mode is introduced into the lattice structure, so that the lattice structure has better mechanical performance under the same compression load.

The invention discloses a natural language driven CAD (Computer Aided Design) modeling method based on multi-agent collaboration. According to the method, a collaborative mechanism of a structure planning agent, a parameter completion agent and an execution verification agent is introduced, and the method comprises the steps that firstly, a structure planning agent analyzes natural language description and generates an abstract syntax tree structure meeting a CAD modeling rule; then, on the premise that the structure of the abstract syntax tree is kept unchanged, a parameter completion agent predicts and completes all symbol and numerical parameters, and a complete CAD modeling instruction sequence is generated; and finally, the execution verification agent performs structure and parameter dual verification on the generation instruction by using the geometric kernel, and feeds back and corrects if necessary. According to the method, a strategy of combining field adaptive full-parameter pre-training and task specific fine tuning is adopted in a training stage, and a structure-parameter decoupling modeling and closed loop verification mechanism is introduced, so that the structure accuracy, the parameter matching property and the geometric performability of a generated result are effectively improved.

The invention discloses a reactor instant-slow source item radiation shielding design system and method, and relates to the field of computer aided design of reactor radiation protection. The reactor instantaneous-delayed source item radiation shielding design system comprises a model conversion module, a population construction module, an index calculation module, an individual evaluation module, a selective evolution module, a hierarchical sorting module and an iterative convergence module which are in data flow connection in sequence from front to back. The invention relates to a reactor instant-slow source item radiation shielding design method. The method comprises the following steps: model conversion; constructing a population; calculating indexes; performing individual evaluation; selecting evolution; layering and sorting; and carrying out iterative convergence. According to the invention, a radiation shielding design system covering the whole life cycle of reactor operation and shutdown is established, and collaborative breakthrough of light weight, high safety and efficient convergence is realized through a multi-objective optimization algorithm.

The invention discloses an electric engine assembly integral forming process based on precision integral casting, which comprises the following steps: designing a split type precision mold through computer aided design software, and carrying out nitriding treatment to improve the performance; selecting a specific aluminum alloy raw material, performing vacuum induction melting and refining, and injecting the aluminum alloy raw material into a preheated and sprayed mold cavity in a low-pressure casting manner; and demolding after air cooling and water cooling are combined, then after-treatment procedures such as deburring and heat treatment are conducted, and finally detection and assembly are conducted. A conformal cooling water channel design and an ultrasonic-assisted smelting technology are adopted in the process innovation, and the casting process is monitored in real time. According to the process, manufacturing procedures are reduced, compared with a traditional mode, the production efficiency is improved by 30%-50%, the cost is reduced by 20%-30%, the assembling problem is effectively avoided, the sealing performance and reliability of the electric engine are improved, the service life of the electric engine is prolonged, high integration is achieved, and the modern industrial requirement is met.

The invention provides a vision-based curved metal surface defect online detection method, which comprises the following steps of: selecting a plurality of vision scanning discrete optimization imaging points from a CAD (Computer Aided Design) model of a to-be-detected metal workpiece in a visual simulation platform, and generating a continuous vision scanning track; aiming at the metal workpiece obtained in the current processing procedure, enabling a machine tool to move according to a continuous visual scanning track, and obtaining a plurality of to-be-detected images with the same shooting angle and different incident light source conditions from the surface of the metal workpiece at each optimized imaging point; the optimized imaging points are input into a highlight removal model, and a fused image after highlight removal is output for each optimized imaging point; inputting the fused image into an interactive image segmentation model, and segmenting a to-be-detected area image of a single part from the fused image in cooperation with interactive prompt; and inputting the data into a defect detection model with an attention mechanism to obtain a defect detection result. According to the invention, real-time, efficient and high-precision defect detection of the complex curved surface metal workpiece in the machining process can be realized.

The invention provides an elastic-plastic unified description method and system for nonlinear characteristics of surrounding rock considering medium-principal stress, and belongs to the technical field of computer aided design. The method comprises the following steps: carrying out a true triaxial mechanical property test on surrounding rock to obtain a surrounding rock total stress-strain curve, extracting characteristic stress points of the stress-strain curve, and carrying out stage division on the stress-strain curve; constructing a characteristic yield surface in the stress space based on the characteristic stress points; based on the staged characteristics of the stress-strain curve, plastic deformation capacity characterization functions of different stages are constructed respectively, and the plastic deformation capacity characterization functions are used for describing the plastic deformation capacity of the stress-strain curve; and constructing a nonlinear hardening-softening equation to describe evolution characteristics of the characteristic yield surface. According to the scheme, the nonlinear characteristics of the stress-strain curve can be more comprehensively and accurately described under different minimum principal stresses and medium principal stresses.

Disclosed is a mutually generative artificial intelligence system based on multi-dimensional spatiotemporal information vector graphics, relating to the field of artificial intelligence and engineering applications. Based on a geographic information system (GIS) or a computer-aided design (CAD) platform and a data source, a multi-dimensional vector spatiotemporal large model terminal, a multi-dimensional spatiotemporal information processing agent terminal, and an intelligent information system application terminal are constructed. For multi-dimensional spatiotemporal data such as two-dimensional and three-dimensional vectors and temporal states, a multimodal spatiotemporal large model having an understanding capacity for an engineering professional knowledge system, a data processing flow, multi-dimensional vector graphics, and thematic graphics-text documents is pre-trained to achieve the mutual expression and generation of engineering multi-dimensional vector graphics and thematic graphics-text documents and to form an intelligent engineering graphic data processing application.

The invention provides an undercarriage system modeling method based on a natural language demand text, and relates to the technical field of model-based computer aided design, and the method comprises the steps: S1, constructing a SysML meta-model knowledge base and an undercarriage system model knowledge base; s2, analyzing the natural language demand text, and generating a structured demand text of the undercarriage system; s3, mapping a structured demand text entity of the undercarriage system into a node type of a semantic graph; and S4, in combination with a historical model template and real-time retrieval, generating an undercarriage system SysML model according to the semantic map obtained in the step S3, and performing undercarriage system engineering analysis through the undercarriage system SysML model, thereby improving the efficiency and quality of aviation equipment development. On the basis of a retrieval enhancement generation technology and an intention routing mechanism, the modeling efficiency and accuracy of the undercarriage system are remarkably improved; a graph neural network and a semantic vector space model are adopted to realize deep fusion and conflict resolution of cross-domain knowledge; and through an incremental learning mechanism, model continuous evolution and design experience closed-loop iteration are supported.

The invention relates to the technical field of computer aided design and BIM (Building Information Model), discloses a BIM ancient building component parameterization design method and system fused with a historical form and system, and aims to solve the problems of parameter distortion, low modeling efficiency, high rework rate and the like caused by disjunction of a historical building rule and a BIM in the prior art. The method comprises the following steps: acquiring a historical shape and system standard text data set; a component-region-geometry triple knowledge graph is constructed; outputting Bezier curve control points conforming to regional characteristics through a lightweight curvature parameter generation network; generating an initial model in the BIM; a dynamic geometric constraint engine is used for verifying whether parameters such as the roll-killing radian and curvature continuity are out of tolerance or not in real time, and visual feedback and correction suggestions are triggered; and the model is automatically updated after user confirmation. The modeling precision and efficiency are remarkably improved, the rework rate is reduced, and the requirements of ancient building protection engineering specifications are met.

The invention discloses a deformation compensation method for a 3D printing model of an oral implanting guide plate, and belongs to the technical field of computer-aided design and manufacturing. The method comprises the steps that original three-dimensional model data of the oral implanting guide plate is obtained; performing dynamic hierarchical stress accumulation analysis on the original three-dimensional model data, and generating a dynamic stress evolution field for describing the internal stress distribution and change of the model in the virtual printing process; based on the dynamic stress evolution field, generating an adaptive compensation transformation matrix for correcting geometric deformation; and performing compensation transformation on the original three-dimensional model data by using the self-adaptive compensation transformation matrix to generate high-fidelity compensation model data. A self-adaptive compensation transformation matrix is generated through dynamic hierarchical stress accumulation analysis to preprocess three-dimensional model data, complex deformation in the printing process can be accurately predicted and compensated, and therefore the forming precision of a printed piece is remarkably improved.

The invention discloses a planning permission data processing method, device and equipment based on a geographic information system, relates to the technical field of graphic data processing, and aims to solve the problem of relatively low processing and analysis efficiency of planning permission data. The method comprises the following steps: converting a planning range line and a building contour line in a computer-aided design (CAD) drawing into topologically closed geographic information system (GIS) surface data, extracting planning design layer information with a spatial association relationship from the GIS surface data, and constructing a structured planning database; based on an analysis engine of a GIS, determining an overlapping surface domain of planning data and permission data in the structured planning database, and dynamically segmenting and recombining the overlapping surface domain through a topological cracking-clustering algorithm to obtain at least one planning region; determining the planning type of each planning area according to a preset priority sequence, and generating a visual layer with a real-time conflict mark; and calculating a permission compliance index and generating an analysis report based on the spatial topological relation and the visual layer.

The invention discloses a component generation method based on anchor point parameter driving, a modeling method and a related product, and relates to the technical field of computer aided design. The component generation method comprises the following steps: creating an anchor point self-adaptive family and a component self-adaptive family, wherein the anchor point self-adaptive family comprises spatial positioning parameters for defining positioning points in a three-dimensional space; the component adaptive family comprises a position parameter associated with the spatial positioning parameter and a geometric parameter of the component; calling an anchor point adaptive family to generate an anchor point instance in a three-dimensional modeling environment, and forming a positioning point; calling a component adaptive family to generate a component instance associated with the positioning point; and when the spatial positioning parameters or the geometric parameters are adjusted, the component instances are automatically updated based on the adjusted parameters. Through adaptive family customization, space positioning control and a technical path generated through rapid connection, the core problems of low steel structure modeling efficiency and insufficient automation in the prior art are effectively solved.

The invention discloses a multi-agent system for knowledge-driven automatic software architecture design, and belongs to the technical field of computer aided design, and the multi-agent system comprises an analysis agent, a modeling agent, a design agent and an evaluation agent which are constructed based on a large language model, the analysis agent is used for extracting, screening and classifying requirements which influence the software architecture design from a software requirement specification and generating a requirement document; the modeling agent is used for generating a software architecture view according to a demand document and an external knowledge base; the design agent is used for generating a software architecture document according to the demand document and the software architecture view; and the evaluation agent is used for evaluating and analyzing the software architecture according to the demand document, the software architecture view and the software architecture document to generate an evaluation and analysis report. According to the method, through cooperation and interaction between the intelligent agents and prompt of external knowledge, the reliability, practicability and creativity of software architecture design can be effectively enhanced.

The invention discloses a non-standard goods shelf intelligent design method and system based on a parameterized model, and relates to the technical field of computer aided design, and the method comprises the following steps: collecting parameters, and carrying out abnormal value elimination and dimension unified processing to generate a standardized demand parameter set; defining core structure parameters and establishing a component association rule, and constructing a parameterized basic model through three-dimensional modeling; the storage rack structure is optimized by combining storage mechanics and material mechanics characteristics, stress concentration is reduced, and deformation under rated load is controlled; a simulation environment is built for statics and dynamics simulation, and a simulation test report is output; screening an optimal parameter combination to generate a detailed design scheme; and receiving user feedback, adjusting parameters and material selection in combination with production feasibility and cost budget, and repeating optimization simulation steps to generate a final scheme. Efficient design of the non-standard goods shelf is achieved through parametric modeling and intelligent optimization, the structure is safe and reliable, the material utilization rate is high, and the design quality and efficiency are remarkably improved.

The invention relates to the technical field of computer aided design, in particular to a method for designing a high-temperature and high-toughness titanium alloy based on machine learning and a preparation method. The method comprises the following steps: acquiring components, a heat treatment process and corresponding tensile property data of a titanium alloy, constructing a titanium alloy data set and defining an exploration space; cleaning the titanium alloy data set, dividing the titanium alloy data set into a training set and a verification set, and normalizing the training set and the verification set; training the titanium alloy tensile property prediction model through a machine learning model and a training set, and evaluating the model through a cross validation method and a validation set; the components and heat treatment process parameters of the high-temperature and high-toughness titanium alloy are obtained through the trained prediction model, the genetic algorithm and the exploration space; according to the components of the high-temperature and high-toughness titanium alloy, the high-temperature and high-toughness titanium alloy is prepared through electric arc melting, and heat treatment is conducted according to heat treatment process parameters of the high-temperature and high-toughness titanium alloy. According to the method, collaborative intelligent design of titanium alloy components and the process is achieved, and the development efficiency is remarkably improved.

The invention discloses a multi-task graph neural network fusion point cloud-attribute adjacency graph information-based processing feature identification method and system, and relates to the technical field of computer aided design, and the method comprises the steps: obtaining a three-dimensional CAD model data set which comprises a plurality of processing features and is provided with a feature tag; preprocessing the model in the data set to obtain a model attribute adjacency graph, a surface point cloud and a corresponding feature vector; learning the model data by using a multi-task graph neural network, realizing three tasks of semantic segmentation, instance segmentation and bottom surface segmentation, and comprehensively obtaining a processing feature recognition result; and inputting a part model of which the machining features need to be recognized into the trained recognition model, and vectorizing part model information to obtain a machining feature recognition result. According to the invention, the accuracy of processing feature recognition is improved.

The invention provides a 7beta-hydroxysteroid dehydrogenase mutant and application thereof in synthesis of ursodesoxycholic acid by a biological enzyme method, through a genetic engineering technology and computer-aided design, phosphoric acid coordination residues of protein are modified, the 7beta-hydroxysteroid dehydrogenase (7beta-HSDH) mutant with higher activity under an NADH (Nicotinamide Adenine Dinucleotide Hydroxide) system is constructed, and the 7beta-hydroxysteroid dehydrogenase (7beta-HSDH) mutant can be used for synthesizing ursodesoxycholic acid. The modification strategy is suitable for other proteins for modifying cofactor preference and improving enzyme activity, and has remarkable industrial value when being used for efficiently catalyzing CDCA to be converted into UDCA through a one-pot one-step method.

The invention relates to the technical field of computer aided design, in particular to a BIM family library construction method and system for tile construction parameters, and the method comprises the following steps: dividing a two-dimensional paving region into discrete grid units based on the geometric shape and obstacle coordinates of a to-be-paved region input by a user, and distributing a unique position identifier for each grid unit; according to the method, discrete grid division is carried out in the two-dimensional paving area according to the geometrical shape and the obstacle coordinate, and the unique position identifier is distributed, so that the original abstract paving area can be converted into a rasterized paving map capable of being positioned and analyzed; and binding probability distribution characteristic values of ceramic tile sizes, material unit prices, labor efficiency and material loss rates with the grids to realize a visual corresponding relationship between construction environment parameters and spatial positions. In the path generation process, through iterative calculation from the starting grid to the ending grid and total cost minimization selection, the construction sequence and time prediction are more accurate.

The invention belongs to the field of computer aided design systems, and relates to a CAD multi-modal data set automatic construction method, which comprises the following steps of: firstly, rendering a three-dimensional CAD model to be marked into a standardized multi-view two-dimensional image set through a data processing system, enhancing a multi-modal large model by utilizing a keyword vector and a low-rank adapter, and constructing a multi-modal data set; the method comprises the following steps of: extracting a geometric label from an image as a hard constraint, generating a text description of a natural language, then verifying a labeling result through a self-consistent strategy, and finally associating the multi-view two-dimensional image, the geometric label and the text description with an original model to construct a high-quality CAD multi-modal image-text pair data set. According to the method, visual feature extraction and language description generation are decoupled into two subtasks with clear purposes, so that the labeling consistency and the recognition precision are improved, and low-cost automatic construction of large-scale CAD data is realized.

A computer-implemented method for modeling thermal and electrochemical properties of lithium-ion battery packs is disclosed. A physics-based thermo-electro-chemical battery model is received comprising electrochemical reaction equations and thermal transport equations. A computer aided design (CAD) file is imported defining physical cell arrangements within a battery pack. Thermal connectivity matrices are automatically generated from the CAD file, where the matrices define heat transfer pathways between adjacent cells. A scaling algorithm transforms single-cell Single-Particle Model (SPM) equations into multi-cell coupled differential equations for the battery pack. The thermal and electrochemical behavior are simulated across the battery pack using the scaled equations to determine at least one battery performance characteristic based on the simulation.

The invention relates to the technical field of computer aided design, and discloses a multi-source data fusion CAD drawing method and system based on deep learning. The invention discloses a multi-source data fusion CAD drawing method based on deep learning. Intelligent fusion processing of a building CAD base map, laser point cloud data and a high-definition video stream is realized through multi-source space-time reference alignment, reflection adaptive point cloud enhancement, cross-modal feature fusion, progressive vector generation, engineering constraint optimization and navigation output. Performing point cloud splicing by adopting an ICP (Inductively Coupled Plasma) algorithm, establishing a spatial corresponding relation by using a Zhang Zhengyou calibration method, extracting multi-modal features through a deep learning network, and realizing cross-modal feature fusion by utilizing an attention mechanism; and finally, a high-precision vector map conforming to engineering specifications is generated and is suitable for automatic generation of a large-scale parking lot navigation map. According to the method, the drawing efficiency and precision are improved, the labor cost is reduced, and the method has good engineering application value.

The invention provides a three-dimensional part retrieval method based on graph similarity search, relates to the field of computer graphics, and solves the technical problems that in the prior art, geometric and design semantic information of CAD cannot be fully utilized, and a topological relation is difficult to capture and display, so that the retrieval efficiency is low. The method comprises the following steps: acquiring part three-dimensional data of a computer-aided design (CAD) model, and constructing a training data set; constructing an edge-surface connection diagram based on the three-dimensional data of the part; calculating a graph editing distance (GED) matrix of all edge surface connection graphs in the training data set as a supervision signal; based on the GED matrix, training a sorting model, mapping an edge-surface connection graph to a hidden space, and constructing a part vector database according to an output graph-level embedding vector; the sorting model is constructed based on a graph attention network; and inputting a to-be-queried CAD part into the trained sorting model to obtain a feature vector, carrying out nearest neighbor search in the part vector database, and returning a similar part result.