Research on Precise Retrieval Methods for Three-Dimensional Bushing Components Using SolidWorks Sketch Profile Features

DOI: https://doi.org/10.62517/jes.202602102

Author(s)

Xin Shao1, Yawen Fan2,*, Jingfeng Shen1,*, LiangWei Zhong1

Affiliation(s)

1School of Mechanical Engineering, University of Shanghai for Science and Technology, Shanghai, China 2School of Engineering and Computing, University of Shanghai for Science and Technology, Shanghai, China *Corresponding Author

Abstract

To facilitate efficient reuse of bushing components in mechanical design, an accurate 3D retrieval method based on SolidWorks sketch profile features is proposed to overcome limitations of existing approaches, including inadequate capture of local design features, dependence on sketch or view data, and poor format compatibility. The proposed method extracts the core sketch for rotationally formed bushing parts via the SolidWorks API and represents it as an ordered sequence of line segments obtained through sketch decomposition. A feature descriptor is constructed using a five-tuple consisting of length, direction angle, line segment type, diameter, and keyway type. Based on this representation, a main contour matching strategy that integrates bidirectional angle mapping with sliding-window matching is employed to evaluate the contour sequence similarity between the query sketch and candidate models in the part library. A special scenario verification mechanism is further introduced to refine the similarity score, supplemented by engineering correction rules that enforce consistency in keyway configuration and outer diameter. The final output is a normalized similarity value within the range of [0, 1]. Experimental validation was conducted on an initial dataset of 500 industrial-grade bushing parts, from which 466 valid models were retained after data screening. The proposed method was evaluated in comparison with CADFind3D, a mainstream industrial CAD retrieval tool based on global shape features. When the top 10 ranked retrieval results were considered for performance evaluation, the proposed five-tuple feature retrieval method based on SolidWorks sketch profiles achieved a precision of 84.5%, a recall of 78.67%, an F1-score of 80.97%, and a mean average precision (MAP) of 87.99%, representing improvements of 172.5%, 174.4%, 177.6%, and 259.88%, respectively, over CADFind3D. These results demonstrate that the proposed method can effectively capture fine-grained design differences, such as local groove types and shaft-section transitions, thereby providing robust technical support for design intent–driven part reuse.

Keywords

Bushing Parts; SolidWorks Sketch Profile; Five-Tuple Feature Description; Bidirectional Angle Mapping; Sliding Window Matching; 3D Model Retrieval

References

[1] Manda B,Bhaskare P, Muthuganapathy R. A convolutional neural network approach to the classification of engineering models. IEEE Access, 2021, 9:22711-22723. [2] Yang Z Y, Bai J, Li W J, et al. Triple hierarchical metric network for 3D model sketch retrieval. Journal of Computer-Aided Design & Computer Graphics, 2024, 36(11):1791-1804. [3] Han X F, Diao Z Y, Zhang C Y, et al. Single-view 3D model retrieval based on attention and view information. Journal of Shandong University (Engineering Science), 2025, 55(04):48-55. [4] Zhou Y, Li W J, Ye D W, et al. Universal retrieval and classification algorithm for 3D models based on internal and external views. Journal of Computer-Aided Design & Computer Graphics, 1-12 [2025-12-06]. [5] Zhu T. Research on 3D mechanical part model retrieval technology and software development based on PointNet neural network. Dalian Maritime University, 2024:1-57. [6] Ji H. Research on classification and retrieval method of 3D medical rehabilitation aids based on point cloud features. Sichuan University, 2024:1-65. [7] Cheng P. GNMR: 3D neuron geometric shape retrieval based on graph neural network. Computer & Digital Engineering, 2024, 52(04):1131-1136. [8] Fan H Y and Bo P B. 3D parametric tire pattern model retrieval based on geometric features. Journal of Zhejiang University (Science Edition), 2023, 50(06):803-810+819. [9] Tai T Y. Research on video retrieval technology based on 3D convolutional neural network. North China University of Technology, 2024:1-59. [10] Zhang G Y and Shao H. Establishment of standard part library based on SolidWorks and VBA. Development & Innovation of Machinery & Electrical Products, 2025, 38(01):124-126. [11] Chang Q Q, Li M X, Zhou W H, et al. The Application of Tolerance Database in the Secondary Development of SolidWorks. Applied Mechanics and Materials, 2014, 3360(599-601):1910-1913. [12] Zhao X, He F, Li K, et al. Introducing an efficient method for feature extraction in image retrieval systems. Scientific Reports, 2025, 15(1):43664-43664. [13] Zhang K, Yu Y Y, Yang X H, et al. Exploration of ideological and political teaching practice in "Mechanical Drawing and CAD" based on craftsman training camp—Taking the teaching project "Mapping of Shaft-Bushing Parts" as an example. Guangdong Vocational and Technical Education and Research, 2024, (06):128-133. [14] Duchnowski R, Wyszkowska P. Laser scanning data processed using Msplit estimation and sliding window algorithm. Reports on Geodesy and Geoinformatics, 2025, 120(1):67-74. [15] Yang Y, Zhang P, Wang Y, et al. An attention-based parallel model with sliding window decomposition algorithm for water quality prediction. Journal of Water Process Engineering, 2025, 78:108751. [16] Xu A M, Da G X, Yu L, et al. Research on rail corrugation defect detection algorithm based on YOLOv8 and multi-dimensional feature enhancement module. Modern Urban Transit, 2025, (11):97-103. [17] Li D S, Wang G, Yu J, et al. Research on insulator defect detection algorithm based on YOLO11-PDL. Intelligent Computer and Applications, 1-12 [2025-12-07].