EUVIMEDEUVIMED entdecken
Crossref

Physics-Informed Neural Networks for the Prediction of Critical Sand Transport Velocity in Oil and Gas Pipelines

A. B. Ehinmowo, B. Ofum, J. O. Olaide

Journal of Engineering Research · 2024 · 29 (4) · 31-40

DOI: 10.52968/72016982 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Engineering Research, 2024. Verfasst von A. B. Ehinmowo und 2 weiteren Autor:innen. Die Arbeit wurde bisher 4-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Poorly consolidated reservoir formations make the generation of sand alongside hydrocarbons unavoidable. The condition of petroleum pipelines can be seriously compromised by the rapid generation of sand or by sand deposition at low velocities, leading to degradation and reduced capacity. Consequently, it is essential to investigate the minimum transport velocity needed to prevent pipeline sand accumulation. This study employed two predictive modelling tools, a physics-informed neural network (PINN) and a multilayer perceptron (MLP) regressor, to estimate the minimum transport conditions in a solid-liquid-gas pipeline. The models were developed using 182 experimental data sets, which included variables such as superficial velocity, pipe diameter, particle diameter, sand density, sand concentration, liquid density, viscosity, pipe angle of inclination, and critical velocity parameters. The results indicated that the physics-informed neural network outperformed the MLP regressor, achieving an R² coefficient of 0.9999 and a root mean square error (RMSE) of 0.00465. In comparison, the MLP regressor attained an R² coefficient of 0.9992 and an RMSE of 0.0295. Both models were evaluated alongside existing empirical and data-driven models and demonstrated superior performance in predicting minimum transport velocity, with the PINN yielding the best results. A sensitivity analysis revealed that the superficial velocity is the most influential parameter for predicting minimum transport velocity, followed by pipe diameter, sand concentration, and pipe angle. This research highlights the potential of effectively integrating physical laws into the machine learning training process to estimate minimum transport velocity in multiphase pipelines better.

Zitieren

APA

Ehinmowo, A. B., Ofum, B., & Olaide, J. O. (2024). Physics-Informed Neural Networks for the Prediction of Critical Sand Transport Velocity in Oil and Gas Pipelines. Journal of Engineering Research, 29(4), 31-40. https://doi.org/10.52968/72016982

Vancouver

Ehinmowo AB, Ofum B, Olaide JO. Physics-Informed Neural Networks for the Prediction of Critical Sand Transport Velocity in Oil and Gas Pipelines. Journal of Engineering Research. 2024;29(4):31-40. doi:10.52968/72016982

BibTeX
@article{Ehinmowo2024,
  author = {Ehinmowo, A. B. and Ofum, B. and Olaide, J. O.},
  title = {Physics-Informed Neural Networks for the Prediction of Critical Sand Transport Velocity in Oil and Gas Pipelines},
  journal = {Journal of Engineering Research},
  year = {2024},
  volume = {29},
  number = {4},
  pages = {31--40},
  doi = {10.52968/72016982}
}
Kostenlos starten →