Fault diagnosis of EHA with few-shot data augmentation technique

Abstract

As an emerging object in aerospace actuators, electro-hydrostatic actuator (EHA) has the advantages of heavy load capacity and high reliability. An EHA fault diagnosis method based on a few-shot data augmentation technique is proposed to diagnose and isolate possible faults. The sensitive parameters of typical failure modes are demonstrated based on the mathematical model of EHA. By converting multi-dimensional experimental data into two-dimensional grayscale data and extracting local features, the time series characteristics and correlation between different signals can be highlighted. The Wasserstein Deep Convolutional Generative Adversarial Network (WDCGAN) is used to enhance the EHA small sample data. The diagnostic model WDCGAN-SDAE combined with WDCGAN and Stacked Denoised Auto Encoder (SDAE) is proposed to differentiate between multiple types of EHA failures. Compared with the five commonly used fault classification methods, the proposed method can effectively identify the typical fault modes of EHA, with the highest accuracy of fault classification and strong feature extraction ability.

Few-Shot Learning Approaches for Fault Diagnosis Using Vibration Data: A Comprehensive Review

“…In the case of limited data, a generative model can be used to enhance sample diversity. According to the reviewed publications on the application of few-shot learning in the machine fault diagnosis field using vibration data from 2018 to 2023, the data augmentation methods are mainly GAN (generative adversarial network)-based methods [53,54].…”

“…Wan et al [56] proposed an unsupervised fault diagnosis method based on a quick self-attention convolutional generative adversarial network. Chen et al [54] utilized a Wasserstein deep convolutional generative adversarial network (WDCGAN) to improve the performance of few-shot fault diagnosis in electrohydrostatic actuators. They achieved this by transforming multidimensional experimental data into 2D grayscale data and extracting local features, effectively emphasizing the time-series characteristics and correlations among different signals.…”