Smart solutions for pipeline protection: Utilizing artificial neural networks to adequate device selection against hydraulic transients

ABSTRACT Hydraulic transients pose a significant threat to pipeline integrity, leading to catastrophic failures from pressure surges. Traditional methods for selecting protection devices – such as air vessels and surge tanks – often rely on engineering judgment, potentially leading to suboptimal solutions. This study introduces a data-driven approach using artificial neural networks (ANNs) to objectively select the most suitable protection devices, overcoming the limitations of conventional engineering intuition. A comprehensive dataset, representing diverse pipeline configurations and commercial materials, was developed. Utilizing established selection criteria, we identified optimal protection devices for various scenarios. Four distinct ANN architectures were trained and assessed based on performance metrics such as accuracy and precision, with the best model validated using an independent dataset of previously unseen configurations. The trained ANN model demonstrated 91.5% accuracy in device selection, outperforming traditional methods and offering enhanced strategies for pipeline protection across diverse scenarios. By incorporating a broader range of pipeline configurations and physical factors, the proposed ANN-based approach offers a robust tool for optimizing pipeline protection strategies and transcends engineer intuition, potentially revealing unconventional yet highly effective solution.