Adaptive Wavelet Kernel Attention Network for Cross-Device Smartphone-Based Induction Motor Fault Diagnosis
IEEE Sensors Journal · 2026-11 · 26 (21) · 32908-32916
DOI: 10.1109/jsen.2026.3733586 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, 2026-11. Verfasst von Mert Sehri und 1 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben.
Abstract (Originalsprache)
Smartphone vibration monitoring offers a cost-effective alternative to industrial accelerometers for induction motor fault diagnosis. However, the low sampling rates, high noise floors, and device-dependent sensor characteristics of consumer-grade smartphones present significant challenges that existing methods fail to address. This article proposes the adaptive wavelet kernel attention network (AWKANet), a novel deep learning framework that integrates learnable Morlet wavelet convolution kernels with a frequency-band attention module (FBAM) to adaptively decompose and weight vibration signals for fault classification. The wavelet kernel parameters, including center frequency and bandwidth, are optimized end-to-end via backpropagation, enabling the network to automatically adapt its signal decomposition to each device’s unique sensor profile. The proposed method is validated on two independent smartphone induction motor datasets with four different smartphone models (iPhone 13, Samsung Galaxy S6, Samsung Galaxy A50, and iPhone 15) under multiple fault conditions, speeds, and loads. Cross-device experiments show that AWKANet has better generalization ability than traditional methods.
Zitieren
APA
Sehri, M., & Yan, T. (2026). Adaptive Wavelet Kernel Attention Network for Cross-Device Smartphone-Based Induction Motor Fault Diagnosis. IEEE Sensors Journal, 26(21), 32908-32916. https://doi.org/10.1109/jsen.2026.3733586
Vancouver
Sehri M, Yan T. Adaptive Wavelet Kernel Attention Network for Cross-Device Smartphone-Based Induction Motor Fault Diagnosis. IEEE Sensors Journal. 2026;26(21):32908-32916. doi:10.1109/jsen.2026.3733586
BibTeX
@article{Sehri2026,
author = {Sehri, Mert and Yan, Tongtong},
title = {Adaptive Wavelet Kernel Attention Network for Cross-Device Smartphone-Based Induction Motor Fault Diagnosis},
journal = {IEEE Sensors Journal},
year = {2026},
volume = {26},
number = {21},
pages = {32908--32916},
doi = {10.1109/jsen.2026.3733586}
}
EUVIMED