Physical latent variable and loss jointly driven auxiliary classifier generative adversarial network for few-shot rotating machinery fault diagnosis

2026-08-10

PRE

Zheng Xiao *

Chi Zhang *

张飞斌 (Feibin Zhang)

Yeliang Xia

DOI：10.1016/j.asoc.2026.116192

Original

Original request for help

Save

Abstract

En 中文

• Physical Latent Variable and Loss Jointly Driven Auxiliary Classifier Generative Adversarial Network for Few-Shot Rotating Machinery Fault Diagnosis. • A physics-guided latent variable is proposed to improve few-shot generation fidelity. • We propose a physical-feature loss to reduce low-quality GAN-generated samples. • A fine-tuning strategy is designed to improve diagnosis accuracy with limited labels. • Extensive experiments provided SOTA few-shot diagnostic performance and the necessity of substructure in our method.

Physical latent variable and loss jointly driven auxiliary classifier generative adversarial network for few-shot rotating machinery fault diagnosis

Abstract

Journal

Organization

Cited Papers

Citing Papers