In this study, we have introduced a novel approach termed as ClassBD for bearing fault diagnosis under heavy noisy conditions, integrating cascaded neural filters and deep learning classifiers.
Specifically, the time BD filter incorporates quadratic convolutional neural networks (QCNN), and we have mathematically proved its superior capability in extracting periodic impulse features in the time domain.
We have devised a physics-informed loss function composed of kurtosis, ℓ2/ℓ4 norm, and cross-entropy loss to facilitate the joint learning, transforming traditional unsupervised BD into supervised learning.
Comprehensive experiments conducted on three public and private datasets reveal that ClassBD significantly outperforms conventional methods under various noise conditions, showcasing its effectiveness in feature extraction and classification.
#fault-diagnosis #neural-networks #quadratic-convolutional-networks #data-analysis #machine-learning
Collection
[
|
...
]