Skip to main content

ORIGINAL RESEARCH article

Front. Mech. Eng.
Sec. Digital Manufacturing
Volume 10 - 2024 | doi: 10.3389/fmech.2024.1494890
This article is part of the Research Topic Transformative Impact of AI and ML on Modern Manufacturing Processes View all articles

Attention-based process monitoring by multi-rate sampling mask fault diagnosis network

Provisionally accepted
Li Meng Li Meng 1*Wang Yijun Wang Yijun 2
  • 1 Wuxi Institute of Technology, Wuxi, China
  • 2 Central South University, Changsha, Hunan Province, China

The final, formatted version of the article will be published soon.

    In industrial process monitoring, traditional models struggle with the incompleteness of ubiquitous multi-rate sampling data, often fragmenting it and neglecting inter-subset correlations, leading to reduced efficacy and high computational costs. Drawing inspiration from the principles of continuous learning, this paper introduces an innovative attention-based multi-rate sampling mask fault diagnosis network (AMSMFDN). This network is designed to circumvent the aforementioned limitations by reframing the multi-rate sampling fault diagnosis as a continuous learning problem. The model unifies the diagnosis of multi-rate sampling data as a continuous learning challenge, introducing a novel network architecture adept at recognizing and processing diverse task data. Empirical evaluations demonstrate the model's proficiency, achieving an average diagnostic accuracy of 77.14%, close to the 80.46% accuracy of data reconstruction methods, and peaking at 93.5% in a three-phase flow system case, surpassing all compared approaches.

    Keywords: Fault diagnosis, Multi-rate Sampling Data, Multi-task Continual Learning, deep learning, attention mechanism

    Received: 11 Sep 2024; Accepted: 15 Oct 2024.

    Copyright: © 2024 Meng and Yijun. 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: Li Meng, Wuxi Institute of Technology, Wuxi, China

    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.