基于卷积神经网络的滚动轴承故障诊断方法
2023,31(9):9-15
摘要：为了简单、准确地进行轴承故障诊断，结合深度学习理论，对基于卷积神经网络的滚动轴承故障诊断方法进行了研究。首先，选用了结构相对简单的LeNet5卷积神经网络；然后，对轴承振动信号原始数据进行截取和归一化处理后直接生成生成二维矩阵作为神经网络输入；接着，优选卷积核大小、批大小、学习率及迭代次数等网络模型参数；最后，应用sigmoid函数进行多标签分类。实验结果表明，该方法能有效识别正常状态及不同损伤程度下的内圈、外圈、滚动体故障状态，识别准确率达到99.50%以上水平。基于卷积神经网络的滚动轴承故障诊断方法不仅在一定程度上可以简化故障诊断的过程，而且可以充分利用卷积神经网络模型的优势实现高效准确地故障诊断。
关键词：深度学习，卷积神经网络，特征自动提取，轴承故障诊断
Bearing fault diagnosis algorithm based on convolutional neural network
Abstract：In order to diagnose the bearing fault simply and accurately, combined with deep learning theory, a mode based on convolutional neural network(CNN) was proposed. Firstly, the LeNet5 CNN with simple model architecture was chosed;Secondly, using the raw data of the bearing vibration signal which is intercepted and normalized, a two-dimensional matrix is generated directly as the input of CNN; Thirdly, the convolution, kernel batch, learning ratesize and the iterations was optimized. Finally, the sigmoid function was choiced to classify. The experimental results show that the method can identify the normal, inner ring fault, outer ring fault and rolling fault effectively and the recognition accuracy can reach a level of over 99.50%. Bearing fault diagnosis algorithm based on convolutional neural network not only simplifies the process of fault diagnosis to a certain extent, but also fully utilizes the advantages of CNN models to achieve efficient and accurate fault diagnosis.
Key words：deep;learning, convolutional;neural network, feature;extraction automatic, bearing;fault diagnosis
收稿日期：2023-03-01
基金项目：国家部委预研基金资助(9140A27020214JB1446)
