摘要
模拟电路具有非线性、元件容差等特性,导致不同故障模式之间存在混叠现象,特别是模拟电路早期故障,这大大增加了故障诊断的难度。因此,提出了一种基于小波变换和多尺度特征注意力卷积神经网络(MS-FACNN)的模拟电路早期故障诊断方法,采用小波变换得到脉冲响应信号的多尺度分量,利用设计好的MS-FACNN网络自动提取更加全面且高可分性故障特征,并实现故障模式识别。此外,采用高效通道注意力(ECA)聚焦故障高相关性特征,过滤低相关性的冗余信息,进一步提升模型特征提取能力。实验结果表明,相比传统方法,所提方法具有更强的故障特征提取能力,对四运放双二阶高通滤波器早期故障诊断的准确率达到99.18%。
The nonlinearity and tolerance of analog circuits always lead to the aliasing phenomenon among different fault modes,which increases the difficulty of fault diagnosis,especially in scenarios with incipient faults.An incipient fault diagnosis method for analog circuits is proposed.Multi-scale convolutional neural network with feature attention mechanism(MS-FACNN)is developed,which can obtain complementary and rich diagnosis information from multi-scale components extracted by wavelet transform and identify the fault modes.Then,the efficient channel attention(ECA)module is employed to emphasize the feature regions associated with fault diagnosis and suppress the irrelevant feature regions,which can improve the representation ability of important fault features.The experimental results show that the proposed method is very effective in feature extraction for fault diagnosis,and has superior classification performance for incipient faults compared with other excellent models recently proposed.
作者
徐欣
侯成凯
XU Xin;HOU Chengkai(Zhengzhou Metro Group Co.,Ltd.,Zhengzhou He’nan 450052,China)
出处
《电子器件》
CAS
2024年第4期929-934,共6页
Chinese Journal of Electron Devices
关键词
模拟电路
早期故障诊断
小波变换
多尺度卷积神经网络
有效通道注意力
analog circuit
incipient fault diagnosis
wavelet transform
multi-scale convolutional neural network
efficient channel attention
