摘要
针对传统图神经网络在故障诊断中使用单一尺度进行特征提取且难以在复杂工况下提取信号的弱特征问题,提出了一种基于多尺度图Transformer的滚动轴承故障诊断方法。该方法提出了一种新的图节点多尺度特征聚合模块,扩大特征提取的感受野以增强特征表示;构建了图节点的中心性编码和空间性编码,以获得图结构信息;利用多头自注意力对故障节点进行特征提取和学习,提高方法捕捉重要特征的能力。在凯斯西储大学轴承数据集和滚动轴承实验平台上分别进行实验验证,诊断准确率最高为99.86%,平均准确率也在98%以上。结果表明,提出的多尺度图Transformer网络模型在多种工况下均能准确的进行故障分类。
A fault diagnosis method for rolling bearings based on a multi-scale graph Transformer network is proposed to address the limitations of traditional graph neural networks,which extract features using a single scale and struggle to capture weak features under complex operating conditions.This method introduces a novel graph node multi-scale feature aggregation module to enlarge the receptive field and enhance feature representation.It constructs centrality encoding and spatial encoding for graph nodes to incorporate graph structure information.Multiple-head self-attention is employed to extract and learn features from fault nodes,improving the ability to capture important features.In this study,experiments were conducted on both the Case Western Reserve University bearing dataset and the rolling bearing test platform.The highest diagnostic accuracy achieved was 99.86%,with an average accuracy exceeding 98%.The results indicate that the multi-scale graph Transformer network model proposed in this paper can accurately classify faults under various operating conditions.
作者
卢浩龙
朱彦敏
Lu Haolong;Zhu Yanmin(School of Computer Science and Engineering,Anhui University of Science&Technology,Huainan 232001,China;School of Mechanical Engineering,Anhui University of Science&Technology,Huainan 232001,China)
出处
《国外电子测量技术》
北大核心
2023年第12期186-194,共9页
Foreign Electronic Measurement Technology
基金
安徽省高等学校科学研究项目(重大项目)(2022AH040113)资助。
