A method for mechanical fault recognition with unseen classes via unsupervised convolutional adversarial auto-encoder. (15th December 2020)

• Record Type: Journal Article
• Authors: Pan, Tongyang; Chen, Jinglong; Qu, Cheng; Zhou, Zitong
• Is Part Of: Measurement science & technology. Volume 32:Number 3(2021)
• Publication Date: 2020-12-15
• Subjects: fault recognition -- bearing -- deep learning -- auto-encoder; Physical measurements -- Periodicals; Scientific apparatus and instruments -- Periodicals; Equipment and Supplies -- Periodicals; Science -- instrumentation -- Periodicals; Technology -- instrumentation -- Periodicals; Mesures physiques -- Périodiques; Physical measurements; Scientific apparatus and instruments; Periodicals; 502.87
• DOI: 10.1088/1361-6501/abb38c
• Languages: English
• ISSNs: 0957-0233

Abstract

Abstract: Intelligent fault recognition has been a hot topic in the area of mechanical fault detection. However, it is difficult to collect sufficient monitoring data to represent the various kinds of faults and support network training. This paper proposes a convolutional adversarial auto-encoder (AE) for mechanical fault recognition with unseen classes via one-class classification. The generator is established based on the convolutional AE, while the discriminator is a multi-scale convolutional neural network. Through unsupervised adversarial training, the model can recognize unseen faults, which are not represented in the training data. The proposed method is verified by three bearing datasets, and some related research is also introduced for comparative analysis. Results show that the RR of the proposed method arrives at 100%, 100% and 97.3% in three cases, while the AC reaches 91.4%, 90.5% and 90.8% respectively.