栈式自编码器在强噪声环境下的轴承故障诊断

由于滚动轴承的工作环境复杂,所采集的信号中通常含有大量噪声,噪声的存在会影响故障诊断的结果。为了提高噪声数据的诊断精度,采用改进的小波阈值函数结合栈式自编码器(stacked auto-encoder,SAE)对强噪声环境下的轴承数据进行故障诊断。首先通过改进阈值函数对噪声数据进行去噪,其次用小波包变换提取去噪数据的小波包能量,最后通过SAE得到故障的分类结果。通过在凯斯西储大学的轴承数据集上的实验表明,该模型能够在强噪声背景下得到较为准确的分类结果。

Bearing Fault Diagnosis of Stacked Autoencoders in Strong Noise Environment

The collected vibration signals usually contain a lot of noise due to the complicated working environment of rolling bearings,and the existence of noise will affect the results of fault diagnosis.In order to improve the diagnostic accuracy of noise data,an improved wavelet threshold function combined with a stacked auto-encoder(SAE)is used to diagnose the bearing data in a strong noise environment.First,the noise data is denoised by improving the threshold function,then the wavelet packet energy of the denoised data is extracted by wavelet packet transformation,and finally the fault classification result is obtained by SAE.Experiments on the bearing dataset of Case Western Reserve University(CWRU)show that the model can get more accurate classification results under the background of strong noise.