We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
基于改进卷积神经网络及 Light GBM的 滚动轴承故障诊断.
- Authors
杨瑞双; 宁芊; 雷印杰; 陈炳才
- Abstract
In view of weak generalization ability and long training time of most of fault diagnosis models based on deep learning, a fault diagnosis method for bearings based on improved convolutional neural network and LightGBM is pro posed. Firstly, the convolutional layer is used to extract features from randomly deactivated original signal. Then the global average pooling layer is used to replace fully connected layer to improve generalization ability of the model. Finally, the extracted features are input into LightGBM for classification. The experimental results show that the average duration of training and diagnosis by CNN- LightGBM model are only 44. 64, 0. 08 s respectively, and the average classification accuracy on same load and variable load test set are up to 99.72% and 95.04% respectively. The diagnostic efficiency and classification accuracy CNN-LightGBM model are superior to those of other comparative models, and the model has stronger generalization ability.
- Publication
Bearing, 2021, Issue 6, p44
- ISSN
1000-3762
- Publication type
Article
- DOI
10.19533/j.issn1000-3762.2021.06.009