Machine learning-based prediction of hydrodynamic forces on small-diameter submarine pipelines: The influence of seabed roughness

Abstract

Submarine pipelines are vital components of offshore energy infrastructure, used for transporting oil, gas, and renewable energy resources from sub-sea production sites to processing facilities. The main aim of this study is: first, to investigate the effect of varying seabed roughness and submarine pipeline diameter on the inline and lift wave forces acting on small-diameter pipelines; and second, to propose a quantitative method to determine the wave forces on small-diameter pipelines. Understanding and accurately predicting these hydrodynamic forces is crucial for ensuring the stability and integrity of small-diameter pipelines. To solve this problem, physical model investigations were carried out in a wave flume to assess inline and lift forces on slender pipelines of varying diameters, resting on seabeds with different roughness. The experimental data were used to compute inline and lift force coefficients, which were then utilized to train machine learning models for predictive analysis. It was found that seabed roughness and pipeline diameter significantly influence wave-induced forces. Machine learning (XGBoost) outperforms statistical methods in predictive accuracy but statistical models offer better interpretability for understanding force dynamics. Selecting the optimum quantitative method to determine the wave forces will help in the cost-effective and safe design of small submarine pipelines.