ORIGINAL RESEARCH article
Front. Mech. Eng.
Sec. Engine and Automotive Engineering
Automatic Identification of High-speed Railway Wheelset Defects by Integrating PointNet++ and Swin Transformer
Provisionally accepted
- 1School of Mechanical Engineering,Dalian Jiaotong University, Dalian, China
- 2School of Electrical Engineering,Dalian Jiaotong University, Dalian, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
To address the technical challenges in detecting high-speed railway wheelset defects under complex conditions such as dynamic lighting, foreign object occlusion, and micro-scale anomalies, this paper proposes a dual-modal deep learning framework integrating PointNet++ and Swin Transformer. The framework enhances defect recognition through cross-modal feature collaboration, incorporating a Cross-Modal Attention (CMA) mechanism for dynamic feature alignment and a geometry-guided suppression strategy to mitigate occlusion noise. Experimental results on 10,918 samples demonstrate superior performance, achieving 0.985 accuracy and 0.982 macro F1, with a recognition rate of 0.938 for defects smaller than 1 mm. The model maintains robust accuracy under varying lighting (strong/weak/reflective) and up to 40% occlusion, while optimized deployment on edge devices achieves 23 FPS with only 12M parameters. This work significantly advances the intelligence and reliability of high-speed rail wheelset inspection systems.
Keywords: Automatic defect recognition, Dynamic lighting, Foreign Object Occlusion, high-speed rail wheels, PointNet++ Model, Swin Transformer Model
Received: 23 Sep 2025; Accepted: 19 Dec 2025.
Copyright: Âİ 2025 Ma, Xu and Chen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Jun Ma
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.