摘要
Recently, deep learning-based methods are widely applied in intelligent machinery fault diagnosis due to the rising industrial demands. Particularly, the cross-domain fault diagnosis problem is well addressed by transfer learning techniques. However, recent literatures have shown that the robust performance of features extracted by traditional deep learning models are not well between different working conditions and result in low accuracy. To solve this problem, a multiscale adversarial convolutional neural network (MACNN) is proposed for fault diagnosis of rotating machinery under different working conditions. Firstly, a multiscale convolutional neural network is pre-trained to form the multiscale feature extractor and sample classifier module. Secondly, an adversarial training scheme between the multiscale feature extractor and additional domain discriminator is introduced to improve the ability of feature extractor and obtain the optimal feature representation. Finally, the trained multiscale feature extractor and sample classifier module are applied to construct the fault diagnostic model. The well-known CWRU dataset and a more practical transmission dataset are applied to demonstrate the proposed MACNN. The results indicate that it is feasible to extract the domain invariant features and the proposed MACNN provide a promising approach for knowledge generalization between the different working conditions.
学科与关键词
- Class-Aware Adversarial Multiwavelet Convolutional Neural Network for Cross-Domain Fault DiagnosisZhao, Ke; Liu, Zhenbao*; Zhao, Bo; Shao, HaidongIEEE Transactions on Industrial Informatics, 2023.被引 4
- Domain Adversarial Transfer Network for Cross-Domain Fault Diagnosis of Rotary MachineryChen, Zhuyun; He, Guolin; Li, Jipu; Liao, Yixiao; Gryllias, Konstantinos; Li, Weihua*IEEE Transactions on Instrumentation and Measurement, 2020, 69(11): 8702-8712.被引 150
- Deep Adversarial Capsule Network for Compound Fault Diagnosis of Machinery Toward Multidomain Generalization TaskHuang, Ruyi; Li, Jipu; Liao, Yixiao; Chen, Junbin; Wang, Zhen; Li, Weihua*IEEE Transactions on Instrumentation and Measurement, 2021, 70: 3506311.被引 65
