基于MLKFE-LSTM的滚动轴承电流信号故障诊断  

作　　者：曾胜亮 吴补汐 Zeng Shengliang;Wu Zhixi

机构地区：[1]华中科技大学

出　　处：《变频器世界》2025年第2期85-91,共7页The World of Inverters

摘　　要：提出了一种基于多尺度大核特征提取和长短时记忆网络的故障诊断方法(MLKFE-LSTM)。为充分挖掘原始电流信号中的多尺度信息,设计了多尺度大核特征提取模块(MLKFE),旨在提取长时依赖特征与细粒度特征,从而有效捕获上下文信息。基于学习到的特征,综合考虑效率与性能,采用长短时记忆网络(LSTM)进行故障类型的识别。为验证所提方法的有效性,在PU数据集上进行了实验测试。实验结果表明,该方法能够有效识别滚动轴承故障类别,分类结果的准确率、召回率及F1分数分别达到了99.75%、99.85%和97.75%。这些结果表明,所提方法在滚动轴承故障诊断中具有良好的应用前景。A fault diagnosis method based on multi-scale large-kernel feature extraction and long short-term memory network(MLKFE-LSTM)is proposed.In order to fully mine the multi-scale information in the original current signal,a multi-scale large-kernel feature extraction module(MLKFE)is designed to extract long-term dependent features and finegrained features,thereby effectively capturing contextual information.Based on the learned features,considering both efficiency and performance,a long short-term memory network(LsTM)is used for fault type recognition.To verify the effectiveness of the proposed method,experiments were conducted on the PU dataset.The experimental results show that the method can effectively identify the fault categories of rolling bearings,with the accuracy,recall and Fl score of the classification results reaching 99.75%,99.85%and 97.75%respectively.These results indicate that the proposed method has a good application prospect in rolling bearing fault diagnosis.

关 键 词：轴承故障检测 电流信号 多尺度特征提取 LSTM 

分 类 号：U226.5[交通运输工程—道路与铁道工程]