.png){kind=icon}
.png){kind=icon}
Return
A multi-source data fusion anomaly detection method for new energy vehicle bearings based on electric drive-adaptive heterogeneous feature collaborative learning
.png){kind=icon}
PRE
AI
Q
DOI：10.1177/10775463261424425.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Bearings in new energy vehicles (NEVs) are critical components in drive systems, directly affecting vehicle safety, energy efficiency, and operational reliability. To address the challenges of incomplete single-source data and heterogeneous data fusion in NEV bearing health monitoring, this paper presents an Electric Drive-Adaptive Heterogeneous Feature Collaborative Learning Framework (ED-HFCLF). This framework is designed for NEV operating conditions, including electromagnetic interference, regenerative braking impacts, and wide-range speed variations. The framework employs a dual-stream architecture to separately process time-frequency spectrograms and multivariate time series. The image stream incorporates a ResNeXt-based multi-scale spatial feature extractor with electromagnetic noise suppression and a pyramid feature fusion module. The time-series stream utilizes a hierarchical LSTM encoder-decoder with a drive-cycle-aware mechanism. A cross-modal alignment mechanism with torque-compensation bridges semantic features across streams. A multi-task learning strategy jointly optimizes fault classification, severity estimation, and remaining useful life prediction. Experiments on the CWRU dataset and real vehicle data demonstrate 95.8% fault detection accuracy and 86.7% early fault detection rate, achieving better performance than existing deep models by 3.1% and 8.5%, respectively.
Keywords:
NEV bearing fault diagnosis
electric drive-adaptive learning
multi-source data fusion
electromagnetic interference suppression
cross-modal alignment
deep learning
Journal
J
IF:
2.4
Papers:
349
Citations:
8.8K
Organization
Cited Papers