Page 102 - 《振动工程学报》2026年第5期
P. 102
第 39 卷第 5 期 振 动 工 程 学 报 Vol. 39 No. 5
2026 年 5 月 Journal of Vibration Engineering May 2026
一 种 结 合 shapelets 和 模 式 距 离 的 图 注 意 力 网 络
旋 转 机 械 故 障 诊 断
郭海宇 , 李 帆 , 张晓光 2,3,4 , 方 佳 , 陆凡凡 , 徐清晨 2
2
1
1
2
(1. 沈阳工业大学电气工程学院,辽宁 沈阳 110870; 2. 上海智质科技有限公司,上海 201801;
3. 中国科学技术大学计算机科学与技术学院,安徽 合肥 230026; 4. 长三角信息智能创新研究院,安徽 芜湖 241000)
摘要:针对传统卷积网络提取数据结构特征与相互关系能力有限,并且图神经网络处理图数据时泛化性能不足的问题,本文
提出一种结合 shapelets 和模式距离的图注意力网络。对原始数据进行节点划分,并对每个节点提取多个 shapelets 特征序列,
突出局部特征,提升数据在低噪声情况下的特征提取能力,将每个节点所取得的 shapelets 特征序列通过 K-最近邻(K-nearest
neighbor,KNN)方法构建 KNN 图模型,进而捕捉节点之间的关系与结构信息,将构建好的图模型输入图注意力网络,对相邻
节点自适应分配权重,提高了模型提取故障特征信息的能力,增强了故障分类的性能。为了验证所提方法的有效性,使用多
个公共数据集进行了试验验证,准确率均大于 99%,并在 3 种水泥生产设备数据上进行验证,平均准确率达 99.5%,结果表明
该方法对比现有的传统故障检测模型拥有更好的检测精度和泛化性能。
关键词: 轴承故障诊断;图注意力网络;shapelets;模式距离
中图分类号:TH113.1;TH133.3 文献标志码:A DOI:10.16385/j.cnki.issn.1004-4523.202403047
Graph attention network combining shapelets and pattern distance
for rotating machinery fault diagnosis
1
2
1
2
GUO Haiyu ,LI Fan ,ZHANG Xiaoguang 2,3,4 ,FANG Jia ,LU Fanfan ,XU Qingchen 2
(1.School of Electrical Engineering,Shenyang University of Technology,Shenyang 110870,China;
2.Shanghai Intelligent Quality Technology Co.,Ltd.,Shanghai 201801,China;
3.School of Computer Science and Technology,University of Science and Technology of China,Hefei 230026,China;
4.Yangtze River Delta Information Intelligence Innovation Research Institute,Wuhu 241000,China)
Abstract:Aiming at the problems of limited ability for traditional convolutional network to extract structural features and interrelationships of
data,as well as insufficient generalisation performance of graph neural network when dealing with graph data,this paper proposes a graph
attention network based on shapelets and pattern distance. The original data is divided into nodes and multiple shapelets feature sequences are
extracted for each node to highlight local features and improve the feature extraction ability of the data under low noise,and the shapelets
feature sequences obtained from each node are passed through the K-nearest neighbor method. The KNN graph model is constructed to capture
therelationship and structural information between the nodes,and the constructed graph model is input into the graph attention network to
adaptively assign weights to the neighbouring nodes,which improves the ability of the model to extract fault feature information and enhances
the performance of fault classification. In order to verify the effectiveness of the proposed method,experimental validations are carried out
using multiple public datasets, with accuracy greater than 99%. Additionally, validations are performed on data from three types of cement
production equipment, achieving an average accuracy of 99.5%. The results show that the proposed method exhibits better detection accuracy
and generalization performance compared to existing traditional fault detection models.
Keywords:bearing fault diagnosis;graph attention network;shapelets;pattern distance
旋转机械是机械系统的重要组成部分,广泛应 失甚至人员伤亡 。研究显示,大多数旋转机械的故
[1]
用于能源、工业、航天、运输等领域,且大多数工作 障都是由滚动轴承的损坏引起的,因此,对滚动轴承
环境比较恶劣,一旦发生故障会造成重大的经济损 进行故障诊断对于降低机械维修成本、提高系统的
收稿日期:2024-03-20;修订日期:2024-07-10
基金项目:国网辽宁省电力有限公司科技项目(2023YF-21);中国博士后科学基金资助项目 (2024M753116)

