Source-free domain adaptation based on label reliability for cross-domain bearing fault diagnosis

TL;DR

This work tackles cross-domain bearing fault diagnosis under a source-free domain adaptation setting. The authors propose SDALR, a label-reliability-based framework that combines a data-augmentation-driven pseudo-label voting scheme with entropy-maximization and a cohesion/repulsion objective to leverage all target samples. The method introduces four losses—$L_{lsc}$, $L_{uem}$, $L_{im}$, and $L_{car}$—and shows that balancing discriminability with diversity yields superior adaptation without accessing source data. Experimental results on PU and JNU datasets demonstrate significant gains over SFDA baselines and competitive performance against non-source-free methods, highlighting the approach’s privacy-preserving and practical potential for industrial fault diagnosis.

Abstract

Source-free domain adaptation (SFDA) has been exploited for cross-domain bearing fault diagnosis without access to source data. Current methods select partial target samples with reliable pseudo-labels for model adaptation, which is sub-optimal due to the ignored target samples. We argue that every target sample can contribute to model adaptation, and accordingly propose in this paper a novel SFDA-based approach for bearing fault diagnosis that exploits both reliable and unreliable pseudo-labels. We develop a data-augmentation-based label voting strategy to divide the target samples into reliable and unreliable ones. We propose to explore the underlying relation between feature space and label space by using the reliable pseudo-labels as ground-truth labels, meanwhile, alleviating negative transfer by maximizing the entropy of the unreliable pseudo-labels. The proposed method achieves well-balance between discriminability and diversity by taking advantage of reliable and unreliable pseudo-labels. Extensive experiments are conducted on two bearing fault benchmarks, demonstrating that our approach achieves significant performance improvements against existing SFDA-based bearing fault diagnosis methods. Our code is available at https://github.com/BdLab405/SDALR.

Figures (12)

• Figure 1: Domain Adaptation and Source Free Unsupervised Domain Adaptation.
• Figure 2: The method of SDALR. The source domain training phase is represented by the blue section, while the target domain adaptation phase is represented by the orange section.
• Figure 3: The specific process of pseudo-label voting
• Figure 4: The specific process of cohesion and repulsion
• Figure 5: PUlessmeier2016condition experimental platform demonstration