基于离散S变换和改进深度残差网络的轴承故障诊断

Bearing Fault Diagnosis Based on Discrete S-transform and Improved Deep Residual Network

为了提高滚动轴承故障诊断的准确率,提出基于离散S变换和改进深度残差网络的轴承故障诊断方法。首先,对采集到的滚动轴承数据进行二维特征挖掘;分别用时域重组(TDR),小波变换(CWT)和离散S(Stockwell)变换制作出图片数据集,充分地提取蕴含在振动数据中的时域和频域特征。其次,使用多种神经网络,如Resnet残差网络、VGG网络、AlexNet网络等多种卷积网络对图片数据进行训练,训练中采用小批量梯度下降算法。最后,通过对变载和变速的公开数据集进行不同网络不同预处理方法的训练,并计算模型在测试集的准确率和标准差指标。结果表明,基于离散S变换和改进深度残差网络的轴承故障诊断方法有着更高的识别精度和更好的鲁棒性,训练效果明显优于其他训练方法。证明了该方法在滚动轴承故障诊断中的有效性。

To improve the accuracy of rolling bearing fault diagnosis,a bearing fault diagnosis method based on discrete S-transform and improved deep residual network is proposed.firstly,two-dimensional feature extraction is performed on the collected rolling bearing data.Specifically,the data is processed using time-domain reconstruction(TDR),continuous wavelet transform(CWT),and discrete S(Stockwell)transform to create image datasets.This approach aims to extract temporal and spectral features that are inherent in the vibration data.Secondly,multiple neural networks such as ResNet residual network,VGG network,and AlexNet network are employed to train the image datasets using mini-batch gradient descent algorithm.Finally,various networks with various preprocessing methods are trained and evaluated on publicly available datasets with variable loads and speeds.The accuracy and standard deviation metrics are calculated on the test set to assess the performance of the models.The results demonstrate that the bearing fault diagnosis method based on discrete S-transform and improved deep residual network achieves higher recognition accuracy and better robustness,outperforming other training methods.This validates the effectiveness of the proposed method in rolling bearing fault diagnosis.