摘要
为了提高电梯牵引机齿轮泵的典型故障诊断精度,提出一种基于经验模态分解(EEMD)和双向长短时记忆网络(BLSTM)的行星齿轮泵故障诊断方法。总共设置4种行星齿轮故障类型,综合验证检测性能。通过EEMD方法完成信号分解,对网络实施训练来提升故障类型的分辨精度。研究结果表明:本故障诊断网络模型损失<1%,具有良好稳定性。断齿、正常齿的轮识别率都达到了93%以上,齿根裂纹故障识别率达到了87.2%,可以实现精确识别齿面故障。经过EEMD处理的网络稳定性与精度显著提升。到达后期迭代阶段时,BLSTM网络拟合速度开始变快,精度也获得提升。
To improve the typical fault diagnosis accuracy of elevator tractor gear pump,a fault diagnosis method of planetary gear pump based on ensemble empirical mode decomposition(EEMD)and bidirectional short and long time memory network(BLSTM)is proposed.Four kinds of planetary gear fault types are set up to verify comprehensively the detection performance.EEMD method is used to complete signal decomposition,and network training is carried out to improve the resolution accuracy of fault types.The results show that the loss of the fault diagnosis network model is less than 1%and has good stability.The wheel recognition rate of broken teeth and normal teeth is more than 93%,and the tooth root crack fault is 87.2%,which can accurately identify tooth surface fault.After EEMD processing,the network stability and accuracy are significantly improved.At the later iteration stage,the BLSTM network fitting speed becomes faster with higher accuracy.
作者
王长华
蒋云刚
李保
吴珂
朱凯
WANG Changhua;JIANG Yungang;LI Bao;WU Ke;ZHU Kai(Zhejiang Institute of Mechanical and Electrical Engineering Co.,Ltd.,Hangzhou 310051,China;Center for Balance Architecture,Zhejiang University,Hangzhou 310007,China;Hangzhou Canal Group Construction Management Co.,Ltd.,Hangzhou 310000,China;College of Quality and Safety Engineering,China Metrology University,Hangzhou 310018,China)
出处
《机械制造与自动化》
2022年第3期57-60,共4页
Machine Building & Automation
基金
国家重点研发计划项目(2021YFF0602201)
浙江省重点研发项目(2018C03029)。
关键词
齿轮泵
故障诊断
经验模态分解
双向长短时记忆网络
分类精度
gear pump
fault diagnosis
ensemble empirical mode decomposition
bi-directionallong short-time memory
classification accuracy
作者简介
第一作者:王长华(1978-),男,浙江临海人,高级工程师,本科,主要从事机电设计工作;通信作者:朱凯(1988-),男,江苏苏州人,副教授,博士,研究方向为交通安全与可靠性。
