基于FSK-CNN的轴承故障诊断研究

Research on fault diagnosis method of ship propulsion motor bearing based on FSK-CNN

在旋转机械设备的运维保障过程中,采用基于专家经验的传统故障检测方法难以对轴承的健康状态做出实时的状态检测。针对这一问题,本文提出一种基于快速谱峭度与卷积神经网络(FSK-CNN)的故障诊断方法。首先采用快速谱峭度(FSK)法对振动信号进行特征提取,将一维时域信号转化为二维的谱峭度图;之后,采用一种结合卷积注意力模块(CBAM)的卷积神经网络模型完成故障分类。试验结果表明,快速谱峭度法可以有效提取轴承振动信号故障特征,引入卷积注意力模块对传统卷积神经网络模型具有明显的优化作用,FSK-CNN的故障诊断方法对于10种不同的轴承故障类型的诊断准确率可以达到99%。

In the process of operation and maintenance guarantee of rotating mechanical equipment,it is difficult to detect the health status of bearings in real time by using traditional fault detection methods based on expert experience.To address this problem,this paper proposes a fault diagnosis method based on fast spectral kurtosis and convolutional neural network(FSK-CNN).Firstly,the fast spectral kurtosis(FSK)method is used to extract features from the vibration signals and transform the one-dimensional time-domain signals into a two-dimensional spectral kurtosis map.After that,a convolutional neural network model combined with a convolutional block attention module(CBAM)is used to complete the fault classification.The experimental results show that the fast spectral kurtosis method can effectively extract the fault features of vibration signals of the bearing,the convolutional block attention module has a significant optimization effect on the traditional convolutional neural network model,and the FSK-CNN fault diagnosis method can reach 99%of the diagnosis accuracy for 10 different bearing fault types.