Title: Deep learning-driven parts feature extraction and surface reconstruction for efficient parts pairing

Authors: Xuezhen Li; Xiao Lu; Zhehan Chen; Ning Zhao; Lechang Yang

Addresses: School of Mechanical Engineering, University of Science and Technology Beijing, Xueyuan Road 30, Haidian District, Beijing, China ' China Energy Engineering Group Co., Ltd., Chaoyang District, Beijing, China ' School of Mechanical Engineering, University of Science and Technology Beijing, Xueyuan Road 30, Haidian District, Beijing, China ' School of Mechanical Engineering, University of Science and Technology Beijing, Xueyuan Road 30, Haidian District, Beijing, China ' School of Mechanical Engineering, University of Science and Technology Beijing, Xueyuan Road 30, Haidian District, Beijing, China

Abstract: Assembly stands as a crucial process in industrial manufacturing, but traditional manual parts pairing is often inefficient. Previous research has highlighted the potential of deep learning for feature extraction and 3D reconstruction from point clouds. We introduces an innovative method based on deep learning for high-precision feature extraction and surface reconstruction aimed at parts pairing. By defining essential assembly features and employing the random sample consensus method, geometric dimensions and surface topography data are acquired. Subsequently, deep learning is utilised to directly regress the surface distance function from point samples, enabling detailed surface modelling of parts and supporting assembly simulation within the digital twin framework. A case study for validation reveals that after optimisation, 30 shaft parts and 30 hole parts are successfully matched, with an average uniformity increase of 0.024. This demonstrates the proposed method's superior effectiveness and accuracy in feature extraction and surface reconstruction. [Submitted 26 June 2023; Accepted 17 December 2024]

Keywords: parts paring; features extracting; surface reconstruction; non-contact measuring; deep learning.

DOI: 10.1504/IJMR.2024.149060

International Journal of Manufacturing Research, 2024 Vol.19 No.4, pp.397 - 417

Accepted: 17 Dec 2024
Published online: 13 Oct 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article