基于多尺度CNN和BiLSTM的船舶推进永磁同步电机故障诊断
Fault diagnosis of permanent magnet synchronous motors for ship propulsion based on multi-scale CNN and BiLSTM
闫国华 1胡以怀2
作者信息
- 1. 宁波工程学院机械与汽车工程学院,浙江宁波 315336
- 2. 上海海事大学商船学院,上海 201306
- 折叠
摘要
鉴于船舶推进永磁同步电机(permanent magnet synchronous motor,PMSM)的匝间短路和永磁体不可逆均匀退磁故障可能导致严重的船舶航行事故,提出一种基于多尺度卷积神经网络(multi-scale convolutional neural network,MCNN)和双向长短期记忆网络(bi-directional long short-term memory,BiLSTM)的多信号融合的故障诊断方法(MCNN-BiLSTM),用于诊断PMSM故障.该方法以振动和三相电流信号为输入,采用三列并行的不同尺度的CNN结构来提取信号的全局和局部特征;使用BiLSTM进一步提取特征并识别故障类型.在一台PMSM试验台架上进行多种工况下的故障模拟试验,结果表明与采用单一信号和其他深度学习算法的故障诊断方式相比,本文提出的故障诊断方法具有很好的抗噪声干扰能力和泛化能力.
Abstract
Given that the inter-turn short circuit and irreversible uniform demagnetization fault of permanent magnet synchronous motors(PMSM)for ship propulsion may lead to serious ship navigation accidents,a multi-signal fusion fault diagnosis method based on multi-scale convolutional neural network(MCNN)and bi-directional long short-term memory(BiLSTM),called MCNN-BiLSTM,is proposed for diagnosing PMSM faults.The method takes vibration and three-phase current signals as inputs,and employs three parallel CNN structures with different scales to extract global and local features of the signals;BiLSTM is used to further extract the features and identify the fault types.Fault simulation tests are carried out in a PMSM test bench under various operating conditions,and the results show that the fault diagnosis method proposed in this paper has excellent noise interference resistance and generalization ability compared with fault diagnosis methods using single signals and other deep learning algorithms.
关键词
永磁同步电机(PMSM)/匝间短路/均匀退磁故障/多尺度卷积神经网路(MCNN)/双向长短期记忆(BiLSTM)/故障诊断/信号融合Key words
permanent magnet synchronous motor(PMSM)/inter-turn short circuit/uniform demagnetization fault/multi-scale convolutional neural network(MCNN)/bi-directional long and short-term memory(BiLSTM)/fault diagnosis/signal fusion引用本文复制引用
出版年
2024
NETL
NSTL
万方数据