Federated meta-learning with transformer fusion and adversarial training for few-shot multi-condition fault diagnosis

Jie Zhang

doi:10.1016/j.knosys.2026.116739

Federated meta-learning with transformer fusion and adversarial training for few-shot multi-condition fault diagnosis

2026-07-31

0

PRE

AI

J

Jie Zhang

Y

Yuqing Chang *

王

王福利 (Fuli Wang)

DOI：10.1016/j.knosys.2026.116739

Original

Original request for help

Save

Abstract

En 中文

• A unified framework is proposed for fault diagnosis under changing conditions. • Transformer attention enhances feature fusion and improves diagnostic accuracy. • Adversarial training encourages learning of condition-invariant fault patterns. • Decentralized training enables knowledge sharing while preserving data privacy. • A memory-augmented network supports fast adaptation to new fault types.

Keywords:

Few-shot learning

Transformer-based feature fusion

Federated meta-learning

Adversarial training

Memory-augmented neural network

Federated meta-learning with transformer fusion and adversarial training for few-shot multi-condition fault diagnosis

Abstract

Journal

Organization

Cited Papers

Citing Papers