基于VMD-GRU-Transformer的电梯轴承故障诊断方法
Method of fault diagnosis for elevator bearings based on VMD-GRU-Transformer
盛宇 1李维武 2杨吴奔 3李武 4巩浩5
作者信息
- 1. 林肯电梯(中国)有限公司,浙江嘉兴 314400
- 2. 杭州新马电梯有限公司,浙江 杭州 311600
- 3. 温州市特种设备检测科学研究院,浙江温州 325000
- 4. 丽水市特种设备检测院,浙江 丽水 323000
- 5. 北京理工大学 机械与车辆学院,北京 100081
- 折叠
摘要
随着城市化进程的加速,电梯作为高层建筑中不可或缺的设施,其安全运行受到高度关注.电梯轴承的故障诊断对于保证电梯安全运行至关重要.传统的故障诊断方法主要依赖于人工经验和简单的信号处理技术,难以应对复杂和非线性的故障模式.文中提出了一种基于变分模态分解(VMD)、门控循环单元(GRU)和Transformer的电梯轴承故障诊断方法.该方法首先采用VMD对电梯轴承的振动信号进行自适应分解,提取出关键的模态成分;然后,通过GRU网络深入学习每个模态成分的时间序列特征,捕捉故障信号中的动态信息,引入Transformer模型的多头自注意力机制,实现对特征的全局分析和融合,从而提高故障诊断的准确性和效率;最后,公共轴承数据集的试验结果验证了所提方法在电梯轴承故障诊断方面的有效性.
Abstract
With the acceleration of urbanization,elevators,as the essential facilities in high-rise buildings,become a public focus safety concern.Fault diagnosis of elevator bearings is crucial to ensure safe operation of elevators.The traditional methods of fault diagnosis mainly rely on manual experience and simple signal-processing techniques,which are difficult to cope with com-plex and nonlinear fault patterns.In this article,a method of fault diagnosis for elevator bearings is proposed based on Variational Mode Decomposition(VMD),Gated Recurrent Unit(GRU),and Transformer.Firstly,VMD is used to adaptively decompose the vibration signals of elevator bearings and extract key modal components.Then,the GRU network is employed to deeply ex-plore the features of each modal component's time series;the dynamic information within the fault signals is captured.The intro-duction of the Transformer's multi-head self-attention mechanism ensures global analysis and fusion of features,thereby enhan-cing accuracy and efficiency of fault diagnosis.Finally,the experimental results obtained from a public bearing dataset show that this method is effective in fault diagnosis of elevator bearings.
关键词
轴承/故障诊断/电梯/信号分析/深度学习Key words
bearing/fault diagnosis/elevator/signal analysis/deep learning引用本文复制引用
出版年
2024
NETL
NSTL
万方数据