基于DCAE-CNN的自动倾斜器滚动轴承故障诊断

针对直升机自动倾斜器滚动轴承工况复杂、噪声干扰大,造成故障诊断效果不佳的问题,提出一种基于深度卷积自编码器(Deep Convolutional AutoEncoder,DCAE)和卷积神经网络(Convolutional Neural Network,CNN)的轴承故障诊断方法。该方法首先采用小波变换方法构造不同状态下振动信号的时频图,然后使用DCAE对时频图进行图像去噪,最后利用CNN对去噪后的时频图进行故障分类。利用课题组和美国凯斯西储大学的滚动轴承故障数据开展诊断实验,并与CNN、堆叠降噪自编码器(Stacked Denoise AutoEncoder,SDAE)两种深度学习方法进行对比,结果表明,该方法在高噪声环境下具有更高的故障识别率。

Fault diagnosis for rolling bearing of swashplate based on DCAE-CNN

In order to solve the problem of large noise interference and poor fault diagnosis of the helicopter swashplate rolling bearing, we proposed a fault diagnosis method based on deep convolutional autoencoder (DCAE) and convolutional neural network (CNN). Firstly, the wavelet transform method is used to construct the time-frequency diagram of the vibration signals in different states. Then, the image denoising is performed on the time-frequency map using DCAE. Finally, the time-frequency map after denoising is classified by CNN. Diagnostic experiments were carried out using the bearing fault data of the research team and Case Western Reserve University, and the proposed method is compared with the CNN and the stacked denoise autoencoder (SDAE). The results show that the proposed method has higher fault recognition rate in high noise environment.