摘要
针对变转速工况下滚动轴承振动信号变化大、故障诊断准确率低的问题,提出基于改进深度卷积神经网络(IDCNN)和门控循环单元(GRU)的诊断模型。使用原始振动信号作为输入,避免因人为提取特征而损失信息;引入批标准化(BN)和小卷积核对DCNN进行改进,加深网络深度,增强网络的辨别能力和稳定性;引入在处理时间序列信号上有着独特优势的门控循环单元(GRU),通过将GRU与IDCNN相结合来提高网络模型的性能。试验验证该模型效果显著且性能稳定。
Aiming at the problems that the vibration signals of rolling bearing vary greatly and the fault diagnosis accuracy is low under variable speed conditions,a diagnosis model based on improved deep convolution neural network(IDCNN)and gated recurrent unit(GRU)is proposed in this paper.Firstly,it uses the original vibration signal as input to avoid information loss caused by artificially extracting features.Then,it introduces the Batch normalization(BN)and Small convolution kernel to improve DCNN,so as to deepen the depth of the network and strengthen the distinguishing ability and stability of the network.Finally,it introduces GRU,which has unique advantages in processing time series signals to improve the performance of the network model by combining GRU with improved IDCNN.The results show that the proposed method is effective and the network performance is stable.
作者
唐衡
夏均忠
白云川
金灵
TANG Heng;XIA Junzhong;BAI Yunchuan;JIN Ling(Army Military Transportation University,Tianjin 300161,China)
出处
《军事交通学报》
2023年第2期32-38,共7页
Journal of Military Transportation University
关键词
滚动轴承
故障诊断
改进深度卷积神经网络
门控循环单元
变转速工况
rolling bearing
fault diagnosis
improved deep convolution neural network(IDCNN)
gated recurrent unit(GRU)
variable speed condition
