A New Adaptive Intelligent DC motors Fault Diagnosis System Using R-WDCNN Classifier

Conventional diagnostic approaches for DC motors often require extensive expertise and are labor-intensive. These traditional methods also tend to be less efficient and accurate. In light of these issues, we have introduced a novel, smart diagnostic process for DC motors utilizing wide-kernel residu...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2024 3rd International Conference on Energy, Power and Electrical Technology (ICEPET) S. 1217 - 1220
Hauptverfasser: Xie, Yonghui, Qin, Huabin, Yin, Haihong, Hou, Xujie
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 17.05.2024
Schlagworte:
Accuracy
adaptive batch normalization algorithm
Classification algorithms
Convolutional neural networks
DC motors
Fault diagnosis
Feature extraction
Manuals
Residual deep convolutional neural networks
residual learning
wide first-layer kernels
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Conventional diagnostic approaches for DC motors often require extensive expertise and are labor-intensive. These traditional methods also tend to be less efficient and accurate. In light of these issues, we have introduced a novel, smart diagnostic process for DC motors utilizing wide-kernel residual deep convolutional neural networks (R-WDCNN). Utilizing residual learning techniques, this method adeptly identifies and classifies four distinct types of vibrational signals and is fine-tuned through an adaptive batch normalization algorithm (AdaBN). This classifier has demonstrated improvements in both recognition rates and convergence speeds. The findings demonstrate that our newly developed method exhibits superior adaptability and cognitive capabilities compared to existing approaches, with R-WDCNN achieving up to 94% accuracy on novel sample recognition. Thus, our proposed solution offers enhanced intelligence and precision in DC motor fault diagnosis.
DOI:10.1109/ICEPET61938.2024.10626632