A meta-learning method for few-shot bearing fault diagnosis under variable working conditions

• DOI: 10.1088/1361-6501/ad28e7
• Publication Date: 2024-02-13T22:26:25Z
• Authors: Liang Zeng; Junjie Jian; Xinyu Chang; Shanshan Wang
• Keywords: One shot

Abstract Intelligent fault diagnosis in various industrial applications has rapidly evolved due to the recent advancements data-driven techniques. However, scarcity of data and a wide range working conditions pose significant challenges for existing diagnostic algorithms. This study introduces meta-learning method tailored classification motor rolling bearing faults, addressing limited diverse conditions. In this approach, deep residual shrinkage network is employed extract salient features from vibration signals. These are then analyzed terms their proximity established prototypes, enabling precise categorization. Moreover, model’s generalization few-shot scenarios enhanced through incorporation paradigm during training. The approach evaluated using two well-known public datasets, focusing on varying speeds, loads, high noise environments. experimental results indicate superior accuracy robustness our compared with those studies.