Fault Detection
Fault detection research aims to automatically identify anomalies or malfunctions in diverse systems, from power grids and industrial machinery to satellite constellations and even large language models. Current efforts heavily utilize machine learning, employing various architectures like neural networks (including recurrent and Bayesian variants), autoencoders, and diffusion models, often coupled with techniques like attention mechanisms and knowledge distillation to improve accuracy and interpretability. This field is crucial for enhancing safety, reliability, and efficiency across numerous industries through predictive maintenance, improved diagnostics, and more robust system operation.
Papers
Explaining Deep Neural Networks for Bearing Fault Detection with Vibration Concepts
Thomas Decker, Michael Lebacher, Volker Tresp
A Robust Deep Learning System for Motor Bearing Fault Detection: Leveraging Multiple Learning Strategies and a Novel Double Loss Function
Khoa Tran, Lam Pham, Vy-Rin Nguyen, Ho-Si-Hung Nguyen