Intelligent fault diagnosis of rolling bearing based on novel CNN model considering data imbalance

The intelligent fault diagnosis method based on deep learning has become a powerful tool for analyzing mechanical big data. However, a large proportion of collected data belong to the healthy condition, which will provoke data imbalance. In this article, a novel CNN-based intelligent machinery fault diagnosis method is proposed to deal with the data imbalance. This method consists of two modules, including feature extraction and fault identification. The former is optimized by the weighted-center-label loss function to extract discriminative features. The latter uses the distance between extracted features and pattern center vectors for fault identification. Two datasets are constructed to verify the effectiveness and superiority of the proposed method under data imbalance. The experimental results from two datasets show that the proposed method can effectively deal with data imbalance by extracting separable and discriminative features automatically.