| Title: |
A Robust Condition-Aware Contrastive Learning Method for Cross-Condition Fault Diagnosis of Mechatronic Systems |
| Authors: |
Liu, Zhigang; Gao, Zhanbao; Liu, Baoding; Tang, Diyin; Yu, Jinsong |
| Contributors: |
National Natural Science Foundation of China |
| Source: |
Engineering Research Express ; ISSN 2631-8695 |
| Publisher Information: |
IOP Publishing |
| Publication Year: |
2026 |
| Description: |
Fault diagnosis in mechatronic systems is challenging due to non-stationary operating conditions and scarce labeled data. Although self-supervised learning has shown promise, existing methods often struggle with augmentation-induced semantic instability and overlook the systematic energy drifts caused by condition variations. To address these limitations, this paper proposes a robust contrastive learning method for cross-domain fault diagnosis using condition-induced distribution correction. Specifically, to ensure representation robustness against stochastic perturbations, we first introduce a multi-view prototype consistency mechanism that dynamically weighs view contributions based on semantic reliability. Subsequently, to address energy-induced drifts overlooked by standard covariance alignment, a statistical correction module calibrates amplitude-induced mean deviations and second-order covariances, eliminating physical discrepancies. Furthermore, an adaptive condition-aware strategy leverages operating conditions as weak supervision to progressively guide the model from local signal discrimination to condition-invariant semantic alignment. Extensive experiments demonstrate the method's superior cross-domain generalization and data efficiency. |
| Document Type: |
article in journal/newspaper |
| Language: |
unknown |
| DOI: |
10.1088/2631-8695/ae5393 |
| DOI: |
10.1088/2631-8695/ae5393/pdf |
| Availability: |
https://doi.org/10.1088/2631-8695/ae5393; https://iopscience.iop.org/article/10.1088/2631-8695/ae5393; https://iopscience.iop.org/article/10.1088/2631-8695/ae5393/pdf |
| Rights: |
https://publishingsupport.iopscience.iop.org/iop-standard/v1 ; https://iopscience.iop.org/info/page/text-and-data-mining |
| Accession Number: |
edsbas.A4A3F16B |
| Database: |
BASE |