.png){kind=icon}
.png){kind=icon}
Return
A Novel Dynamic Sparse Convolution Residual Network for Incipient ITSC Fault Diagnosis of Electric Machines
.png){kind=icon}
PRE
AI
H
G
X
Z
陈
DOI：10.1109/tie.2026.3677581.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Existing ITSC fault diagnosis methods for electric machines still face challenges in terms of incipient fault detection, noise robustness, cross-motor type applicability and multiple operating conditions. To this aim, this article proposes a novel dynamic sparse convolutional residual network fault diagnosis model for detecting incipient ITSC fault and estimating severity levels in electric machines. First, a novel dynamic sparse convolution paradigm is proposed in the diagnostic model to extract local features and long-range periodic dependencies from time-series signals. Second, an improved squeeze-and-excitation mechanism is introduced to adaptively enhance the representation of fault-related discriminative features. Extensive experiments carried out across three distinct electric machine platforms demonstrate that the proposed method achieves superior diagnostic accuracy and noise robustness compared with the state-of-the-art models, while maintaining low computational cost. The proposed method also exhibits good generalization performance for unseen machines. Furthermore, visualization analysis is conducted to enhance the model’s interpretability.
Keywords:
Electric machines
dynamic sparse convolution
fault diagnosis
inter-turn short-circuit (ITSC) fault
interpretability analysis
Journal
IF:
7.2
Papers:
1.8W
Citations:
9.8W
Organization
Cited Papers