Global-focal adaptation with information separation for noise-robust transfer fault diagnosis.

Abstract

Existing transfer fault diagnosis methods typically assume either clean data or sufficient domain similarity, which limits their effectiveness in industrial environments where severe noise interference and domain shifts coexist. To address this challenge, we propose an information separation global-focal adversarial network (ISGFAN), a robust framework for cross-domain fault diagnosis under noise conditions. ISGFAN is constructed upon a novel information separation architecture, which integrates adversarial learning with an improved orthogonal loss to decouple domain-invariant fault representations from noise interference and domain-specific characteristics. Furthermore, ISGFAN integrates a global-focal domain-adversarial mechanism that jointly constrains the model from both the conditional and marginal distribution perspectives. Specifically, the focal domain-adversarial module is, for the first time, designed to adaptively address category-specific transfer obstacles caused by noise in unsupervised scenarios, while the global domain classifier ensures the alignment of the overall distribution. Experiments on three public benchmark datasets demonstrate that the proposed method outperforms other prominent existing approaches, maintaining the best cross-domain diagnostic performance under severe noise. Datasets and code are available at https://github.com/JYREN-Source/ISGFAN.