摘要
针对旋转电机轴承在高频冲击和负载多变环境下容易出现机械损伤的问题,传统的深度学习故障诊断方法往往依赖大量的训练数据,导致其在故障诊断精度上存在一定的局限性,难以及时对早期故障进行有效的预警和监测。结合半监督学习模型、多尺度特征融合、一致性正则化和伪标签等技术,提出一种基于多尺度卷积的半监督跨工况电机轴承机械损伤检测方法。该方法通过多尺度特征提取与伪标签生成,有效提高模型对少量带标签数据的利用率。实验结果表明:该方法在训练数据稀缺的条件下仍具有较强的故障识别能力,且相较传统方法优势较为明显。
Rotating motor bearings in vibration equipment are prone to mechanical damage under high-frequency impacts and variable load conditions,and traditional deep learning-based fault diagnosis methods often rely on large amounts of training data,limiting their diagnostic accuracy and the ability to provide timely early warning and monitoring of faults.To address these challenges,this paper proposes a motor bearing mechanical damage detecting method based on multi-scale convolution in semi-supervised cross working conditions,by integrating the techniques of semi-supervised learning model,multi-scale feature fusion,consistency regularization and pseudo-labeling,which improves the utilization of limited labeled data through multi-scale feature extraction and pseudo label generation.Experimental results show that the proposed method achieves strong fault recognition performance in spite of scarce training data,and demonstrates significant advantages over traditional methods.
作者
赵兴华
张智博
朱国勤
吴拼
方元政
姚运仕
于树源
ZHAO Xinghua;ZHANG Zhibo;ZHU Guoqin;WU Pin;FANG Yuanzheng;YAO Yunshi;YU Shuyuan(The Fourth Engineering Co.,Ltd.,CCCC First Highway Engineering Co.,Ltd.,Nanning 530031,China;School of Construction Machinery,Chang'an University,Xi'an 710064,China)
出处
《机械制造与自动化》
2025年第4期68-72,共5页
Machine Building & Automation
基金
广西揭榜制科技项目(桂科JB23023008)。
关键词
多尺度
半监督
跨工况
伪标签
故障诊断
multi-scale
semi-supervised
cross working condition
pseudo label
fault diagnosis
