Research on the Method of Rub-Impact Fault Recognition Based on the Conditional Generative Adversarial Nets and Acoustic Emission
- Resource Type
- Conference
- Authors
- Li, Jing; Deng, Aidong; Yang, Yong; Zhu, Jing; Deng, Minqiang
- Source
- 2020 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD) Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD), 2020 International Conference on. :633-637 Oct, 2020
- Subject
- Aerospace
Components, Circuits, Devices and Systems
General Topics for Engineers
Robotics and Control Systems
Signal Processing and Analysis
Gallium nitride
Vibrations
Generators
Training
Fault diagnosis
Machinery
Generative adversarial networks
Acoustic Emission(AE)
Conditional Generative Adversarial Nets(CGAN)
Rub-Imapct Fault
- Language
A large number of effective annotated data is the key support for learning a fault diagnosis model of mechanical equipment. However, the existing practical samples used for training fault classifiers are usually small and interfered by noise .According to this problem the paper presents a rub impact fault recognition method based on the Conditional Generative Adversarial Nets (CGAN) and Acoustic Emission (AE) technology. The data are from the Wind Turbine Train test bed. AE features are extracted from various views such as time, frequency and energy intensity under different operation state. The proposed CGAN model adds useful auxiliary information into each layer of GAN generation model to improve the quality of generated pseudo-samples. It is further to evaluate the probability of samples from the training set or real set. The experimental results show that this model can effectively identify the rub impact fault and have strong robustness. It is an effective way to solve the problem of inadequate sample and improve the recognition performance of rub impact fault.