A survey of Autoencoder and Convolutional Neural Network Based Methods for Fault Diagnosis

Abstract

Deep learning technology, with its exceptional nonlinear feature extraction capability, has been gradually accepted and widely adopted by the industry. This paper reviews the fault diagnosis techniques utilizing deep learning, emphasizing an in-depth analysis of two main deep learning models that are extensively applied in the current fault diagnosis domain: Autoencoder (AE) and Convolutional Neural Network(CNN). In addition, we also discuss the specific application cases of both algorithms, and introduce their specific principles of implementation. Through these in-depth analysis, it is intended to provide readers with the application overview and technical progress of deep learning technology in the field of fault diagnosis, so as to offer reference and enlightenment for related research.