EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

End-to-End 1D-CNN Leakage Detection in Low-Density Oil-Flow Pipelines Using Raw Simulated Vibration Signals

Bernadetta Kwintiana, Ravikumar Santhanam

Equivalent: Jurnal Ilmiah Sosial Teknik · 31.10.2026 · 8 (5) · 2057-2070

DOI: 10.59261/jequi.v8i5.452 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Equivalent: Jurnal Ilmiah Sosial Teknik, online erschienen am 31.10.2026. Verfasst von Bernadetta Kwintiana und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Leakage detection in low-density oil-flow pipelines remains challenging because leakage-induced vibrations can be masked by pump harmonics, flow-dependent vibrations, and background disturbances. This study investigates an end-to-end one-dimensional convolutional neural network (1D-CNN) that learns directly from raw vibration waveforms without handcrafted feature extraction.

Objective

This study aims to develop, validate, and benchmark a compact end-to-end 1D-CNN for binary leak/non-leak detection across five flow rates under defined noise conditions using a leakage-safe, flow-stratified evaluation protocol.

Methods

A physically informed numerical vibration dataset comprising 600 windows was generated for leak and non-leak conditions at flow rates of 20, 25, 30, 35, and 40 L/min. The data were divided into training, validation, and test sets using a 70:15:15 ratio. The proposed network contains three convolutional stages, global average pooling, a fully connected layer, dropout, and a sigmoid output, with 6,265 trainable parameters.

Results

On the independent test set, the model achieved 94.44% accuracy, 97.62% precision, 91.11% sensitivity, 97.78% specificity, an F1 score of 94.25%, an MCC of 0.891, and a ROC-AUC of 0.986. Performance remained approximately 94% at 20–30 dB SNR but decreased at 15 dB SNR.

Conclusion

The compact end-to-end 1D-CNN demonstrated strong leak/non-leak discrimination under the evaluated simulated conditions. Validation using real pipeline vibration measurements and broader operating conditions is required before practical deployment.

Zitieren

APA

Kwintiana, B., & Santhanam, R. (2026). End-to-End 1D-CNN Leakage Detection in Low-Density Oil-Flow Pipelines Using Raw Simulated Vibration Signals. Equivalent: Jurnal Ilmiah Sosial Teknik, 8(5), 2057-2070. https://doi.org/10.59261/jequi.v8i5.452

Vancouver

Kwintiana B, Santhanam R. End-to-End 1D-CNN Leakage Detection in Low-Density Oil-Flow Pipelines Using Raw Simulated Vibration Signals. Equivalent: Jurnal Ilmiah Sosial Teknik. 2026;8(5):2057-2070. doi:10.59261/jequi.v8i5.452

BibTeX
@article{Kwintiana2026,
  author = {Kwintiana, Bernadetta and Santhanam, Ravikumar},
  title = {End-to-End 1D-CNN Leakage Detection in Low-Density Oil-Flow Pipelines Using Raw Simulated Vibration Signals},
  journal = {Equivalent: Jurnal Ilmiah Sosial Teknik},
  year = {2026},
  volume = {8},
  number = {5},
  pages = {2057--2070},
  doi = {10.59261/jequi.v8i5.452}
}
Kostenlos starten →