摘要
针对行星齿轮箱在实际工作中故障训练样本有限的问题,设计一种基于孪生卷积神经网络的故障诊断方法。采集行星齿轮箱多方向振动信号,使用卷积神经网络对多路振动信号进行特征提取,利用孪生网络架构对所提取特征样本进行距离度量,确定分类准则,完成故障诊断模型的训练。通过行星齿轮箱实验平台的数据实验,证明孪生卷积神经网络在故障训练样本有限的情况下相比传统方法具有更好的故障诊断性能。
A fault diagnosis method based on Siamese convolutional neural networks is adopted to address the problem of limited fault training samples for planetary gearboxes in practice.The multi-directional vibration signals of the planetary gearbox are collected and the convolution neural network is used for feature extraction of multiple vibration signals.With Siamese network architecture,the distance of the extracted feature samples is measured and the classification criteria is determined to accomplish the training of the fault diagnosis model.The data gained from the experiments on the planetary gearbox experimental platform demonstrates that the Siamese convolutional neural network has better fault diagnosis performance than the conventional methods in the condition of limited fault training samples.
作者
钱心筠
王友仁
赵亚磊
QIAN Xinyun;WANG Youren;ZHAO Yalei(College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 21106,China)
出处
《机械制造与自动化》
2022年第4期116-119,共4页
Machine Building & Automation
关键词
故障诊断
孪生网络
行星齿轮箱
深度学习
有限数据
fault diagnosis
Siamese network
planetary gearbox
deep learning
limited data
