摘要
为提升旋转机械故障诊断的水平和效率,提出一种基于谱图和声学特征的旋转机械故障检测方法。CNN网络中输入声发射信号谱图,得到声发射信号的全局特征;短时能量、过零率、峭度等信息输入BiLSTM网络中,提取声发射信号的声学特征,最后将CNN网络和BiLSTM网络提取到的特征融合起来,采用Softmax实现信号识别,通过实验验证了模型的有效性。
To improve the level and efficiency of fault diagnosis of rotating machinery,a fault detection method for rotating machinery based on spectrum and acoustic characteristics is proposed.The global features of acoustic emission signals are obtained by inputting spectrum of acoustic emission signals into convolutional neural network(CNN);short-term energy,zero-crossing rate,kurtosis and other information are input into bi-directional long short-term memory network(BiLSTM)to extract acoustic features of acoustic emission signals.Finally,the features extracted from CNN and BiLSTM are fused,and signal recognition is realized by using Softmax.The effectiveness of the model is verified by experiments.
作者
梁小康
LIANG Xiaokang(Shaanxi Energy Fengjiata Mining Operation Co.,Ltd.,Xi’an Shaanxi 710021)
出处
《电子器件》
CAS
北大核心
2021年第3期737-740,共4页
Chinese Journal of Electron Devices
关键词
卷积神经网络
声发射
故障诊断
双向长短期记忆网络
convolutional neural network
acoustic emission
fault detection
bi-directional long short-term memory