基于 MF1-FedAvg 和 MSRANet 的改进型联合学习方法用于机械故障诊断

IF 1.5 4区 工程技术 Q3 ENGINEERING, MECHANICAL
Xiuyan Liu, Chunqiu Pang, Tingting Guo, Donglin He
{"title":"基于 MF1-FedAvg 和 MSRANet 的改进型联合学习方法用于机械故障诊断","authors":"Xiuyan Liu, Chunqiu Pang, Tingting Guo, Donglin He","doi":"10.1007/s12206-024-0806-5","DOIUrl":null,"url":null,"abstract":"<p>Current fault detection methods for rolling bearings suffer from insufficient data, which limits the generalizability of the models. Typically, conventional approaches train models with a significant amount of labeled data to improve reliability. However, centralized training poses potential risks of data privacy leakage. To address this issue, we propose a federated learning-based fault diagnosis model. In this method, fault diagnosis models for different clients are collaboratively trained by multiple entities with distinct fault characteristics, eliminating the need for third-party aggregation and thereby reducing the risk of data leakage. Specifically, we design a multiscale residual neural network with the ability to perform direct feature extraction from fault data. This proposed network integrates attention units for various scales, emphasizing key features of bearing faults and enhancing the fault recognition capability of local models. Moreover, to address the inherent problem in traditional federated learning frameworks—disparities in client contributions, leading to suboptimal model quality and prolonged training times—this research introduces an innovative weighted strategy based on multiclass F1 scores. This strategy assigns higher weight to high-quality local clients, thereby enhancing both model quality and training speed. Experiments were conducted on two authentic bearing datasets, and the results demonstrate that the proposed method can achieve an average reduction of approximately 15 % in training iterations compared to the federated averaging algorithm, coupled with an average enhancement of approximately 5 % in fault diagnosis accuracy. The experimental results indicate that the proposed method exhibits outstanding accuracy and robustness.</p>","PeriodicalId":16235,"journal":{"name":"Journal of Mechanical Science and Technology","volume":"61 1","pages":""},"PeriodicalIF":1.5000,"publicationDate":"2024-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"An improved federated learning method based on MF1-FedAvg and MSRANet for machinery fault diagnosis\",\"authors\":\"Xiuyan Liu, Chunqiu Pang, Tingting Guo, Donglin He\",\"doi\":\"10.1007/s12206-024-0806-5\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>Current fault detection methods for rolling bearings suffer from insufficient data, which limits the generalizability of the models. Typically, conventional approaches train models with a significant amount of labeled data to improve reliability. However, centralized training poses potential risks of data privacy leakage. To address this issue, we propose a federated learning-based fault diagnosis model. In this method, fault diagnosis models for different clients are collaboratively trained by multiple entities with distinct fault characteristics, eliminating the need for third-party aggregation and thereby reducing the risk of data leakage. Specifically, we design a multiscale residual neural network with the ability to perform direct feature extraction from fault data. This proposed network integrates attention units for various scales, emphasizing key features of bearing faults and enhancing the fault recognition capability of local models. Moreover, to address the inherent problem in traditional federated learning frameworks—disparities in client contributions, leading to suboptimal model quality and prolonged training times—this research introduces an innovative weighted strategy based on multiclass F1 scores. This strategy assigns higher weight to high-quality local clients, thereby enhancing both model quality and training speed. Experiments were conducted on two authentic bearing datasets, and the results demonstrate that the proposed method can achieve an average reduction of approximately 15 % in training iterations compared to the federated averaging algorithm, coupled with an average enhancement of approximately 5 % in fault diagnosis accuracy. The experimental results indicate that the proposed method exhibits outstanding accuracy and robustness.</p>\",\"PeriodicalId\":16235,\"journal\":{\"name\":\"Journal of Mechanical Science and Technology\",\"volume\":\"61 1\",\"pages\":\"\"},\"PeriodicalIF\":1.5000,\"publicationDate\":\"2024-09-04\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Mechanical Science and Technology\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://doi.org/10.1007/s12206-024-0806-5\",\"RegionNum\":4,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"ENGINEERING, MECHANICAL\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Mechanical Science and Technology","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.1007/s12206-024-0806-5","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENGINEERING, MECHANICAL","Score":null,"Total":0}
引用次数: 0

摘要

目前的滚动轴承故障检测方法存在数据不足的问题,这限制了模型的通用性。通常情况下,传统方法使用大量标注数据来训练模型,以提高可靠性。然而,集中式训练存在数据隐私泄露的潜在风险。为了解决这个问题,我们提出了一种基于联合学习的故障诊断模型。在这种方法中,不同客户的故障诊断模型由具有不同故障特征的多个实体协同训练,无需第三方聚合,从而降低了数据泄漏的风险。具体来说,我们设计了一种多尺度残差神经网络,能够直接从故障数据中提取特征。该网络整合了不同尺度的注意单元,强调了轴承故障的关键特征,增强了局部模型的故障识别能力。此外,为了解决传统联合学习框架的固有问题--客户贡献不均,导致模型质量不理想和训练时间延长,本研究引入了基于多类 F1 分数的创新加权策略。该策略为高质量的本地客户端分配更高的权重,从而提高了模型质量和训练速度。实验在两个真实的轴承数据集上进行,结果表明,与联合平均算法相比,所提出的方法平均减少了约 15% 的训练迭代次数,同时平均提高了约 5% 的故障诊断准确率。实验结果表明,所提出的方法具有出色的准确性和鲁棒性。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
An improved federated learning method based on MF1-FedAvg and MSRANet for machinery fault diagnosis

Current fault detection methods for rolling bearings suffer from insufficient data, which limits the generalizability of the models. Typically, conventional approaches train models with a significant amount of labeled data to improve reliability. However, centralized training poses potential risks of data privacy leakage. To address this issue, we propose a federated learning-based fault diagnosis model. In this method, fault diagnosis models for different clients are collaboratively trained by multiple entities with distinct fault characteristics, eliminating the need for third-party aggregation and thereby reducing the risk of data leakage. Specifically, we design a multiscale residual neural network with the ability to perform direct feature extraction from fault data. This proposed network integrates attention units for various scales, emphasizing key features of bearing faults and enhancing the fault recognition capability of local models. Moreover, to address the inherent problem in traditional federated learning frameworks—disparities in client contributions, leading to suboptimal model quality and prolonged training times—this research introduces an innovative weighted strategy based on multiclass F1 scores. This strategy assigns higher weight to high-quality local clients, thereby enhancing both model quality and training speed. Experiments were conducted on two authentic bearing datasets, and the results demonstrate that the proposed method can achieve an average reduction of approximately 15 % in training iterations compared to the federated averaging algorithm, coupled with an average enhancement of approximately 5 % in fault diagnosis accuracy. The experimental results indicate that the proposed method exhibits outstanding accuracy and robustness.

求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
Journal of Mechanical Science and Technology
Journal of Mechanical Science and Technology 工程技术-工程:机械
CiteScore
2.90
自引率
6.20%
发文量
517
审稿时长
7.7 months
期刊介绍: The aim of the Journal of Mechanical Science and Technology is to provide an international forum for the publication and dissemination of original work that contributes to the understanding of the main and related disciplines of mechanical engineering, either empirical or theoretical. The Journal covers the whole spectrum of mechanical engineering, which includes, but is not limited to, Materials and Design Engineering, Production Engineering and Fusion Technology, Dynamics, Vibration and Control, Thermal Engineering and Fluids Engineering. Manuscripts may fall into several categories including full articles, solicited reviews or commentary, and unsolicited reviews or commentary related to the core of mechanical engineering.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信