EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Centrifugal Pump Fault Diagnosis Based on a Novel SobelEdge Scalogram and CNN

Wasim Zaman iD, Zahoor Ahmad iD, Muhammad Farooq Siddique iD, Niamat Ullah, Jong-Myon Kim iD

Sensors · 01.06.2023 · 23 (11) · 5255

DOI: 10.3390/s23115255 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 01.06.2023. Verfasst von Wasim Zaman und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Electrical, Electronic and Computer Engineering, University of Ulsan, Ulsan 44610, Republic of Korea. Hauptthemen laut MEDLINE: Neural Networks, Computer, Signal Processing, Computer-Assisted. Die Arbeit wurde bisher 40-mal zitiert (Crossref) und nennt 45 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: University of Ulsan, Ministry of Trade, Industry and Energy (MOTIE), Korea Evaluation Institute of Industrial Technology (KIET).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper presents a novel framework for classifying ongoing conditions in centrifugal pumps based on signal processing and deep learning techniques. First, vibration signals are acquired from the centrifugal pump. The acquired vibration signals are heavily affected by macrostructural vibration noise. To overcome the influence of noise, pre-processing techniques are employed on the vibration signal, and a fault-specific frequency band is chosen. The Stockwell transform (S-transform) is then applied to this band, yielding S-transform scalograms that depict energy fluctuations across different frequencies and time scales, represented by color intensity variations. Nevertheless, the accuracy of these scalograms can be compromised by the presence of interference noise. To address this concern, an additional step involving the Sobel filter is applied to the S-transform scalograms, resulting in the generation of novel SobelEdge scalograms. These SobelEdge scalograms aim to enhance the clarity and discriminative features of fault-related information while minimizing the impact of interference noise. The novel scalograms heighten energy variation in the S-transform scalograms by detecting the edges where color intensities change. These new scalograms are then provided to a convolutional neural network (CNN) for the fault classification of centrifugal pumps. The centrifugal pump fault classification capability of the proposed method outperformed state-of-the-art reference methods.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Neural Networks, ComputerSignal Processing, Computer-AssistedVibration

Stichwörter der Autor:innen

centrifugal pumpconvolutional neural networkfault diagnosisrotating machinerystockwell transformvibrational signals

Autor:innen und Institutionen

  • Department of Electrical, Electronic and Computer Engineering, University of Ulsan, Ulsan 44610, Republic of Korea.
  • PD Technology Cooperation, Ulsan 44610, Republic of Korea.

ORCID-Profile: Wasim Zaman, Zahoor Ahmad, Muhammad Farooq Siddique, Jong-Myon Kim

Förderung

  • University of Ulsan
  • Ministry of Trade, Industry and Energy (MOTIE)
  • Korea Evaluation Institute of Industrial Technology (KIET)
  • Technology Infrastructure Program funded by the Ministry of SMEs and Startups (MSS, Republic of Korea)

Zitieren

APA

Zaman, W., Ahmad, Z., Siddique, M. F., Ullah, N., & Kim, J. M. (2023). Centrifugal Pump Fault Diagnosis Based on a Novel SobelEdge Scalogram and CNN. Sensors, 23(11), 5255. https://doi.org/10.3390/s23115255

Vancouver

Zaman W, Ahmad Z, Siddique MF, Ullah N, Kim JM. Centrifugal Pump Fault Diagnosis Based on a Novel SobelEdge Scalogram and CNN. Sensors. 2023;23(11):5255. doi:10.3390/s23115255

BibTeX
@article{Zaman2023,
  author = {Zaman, Wasim and Ahmad, Zahoor and Siddique, Muhammad Farooq and Ullah, Niamat and Kim, Jong-Myon},
  title = {Centrifugal Pump Fault Diagnosis Based on a Novel SobelEdge Scalogram and CNN},
  journal = {Sensors},
  year = {2023},
  volume = {23},
  number = {11},
  pages = {5255},
  doi = {10.3390/s23115255}
}
Kostenlos starten →