A significant aspect of offshore pipeline engineering involves evaluating the uplift resistance and failure probability of buried pipelines in clay, which are affected by factors such as pipeline geometry, soil characteristics, material properties, and loading conditions. Subsea marine clay is generally not homogeneous, leading to variations in undrained shear strength vertically and horizontally. As a result, the stochastic analysis method is suitable for accurately modeling such soil conditions. This study addresses these challenges using the Random Adaptive Finite Element Limit Analysis (RAFELA) to analyze the mean uplift resistance factor and the probability of failure for buried rectangular box pipelines in random clay. Seven key parameters are considered in the parametric study: the embedment depth ratio (H/B = 0.5, 1, 2, 4, and 6), width-to-height ratio (L/B = 0.5, 1, 2, 3, and 4), overburden factor (γH/μc = 0, 0.5, and 1), adhesion factor (α = 0, 0.5, and 1), load inclination (β = 0°, 45°, and 90°), coefficient of variation (CoVμc = 25% and 60%), and spatial correlation length (Θc = 0.125, 0.5, 1, 2, 4, and 8). The results are presented as dimensionless uplift resistance factors (μNran), probability of failure (Pf), as well as the corresponding safety factor (FS) for designing pipelines in random clay, ensuring practical designs that are both efficient and reliable. Additionally, this study compares its findings with pullout capacity factors derived from deterministic analyses reported in the literature. This study incorporates machine learning, specifically the Random Forest (RF) algorithm, to predict Pf based on parametric data. The RF model, trained on 500 samples (70% training, 30% testing), achieves high predictive accuracy, with R2 values of 99.12% and 97.29%, respectively. The Shapley Additive Explanations (SHAP) analysis identifies FS as the most influential factor, directly contributing to the reliability of the pipeline design, while α has the least impact. The analysis emphasizes the practical significance of FS in reducing failure probabilities while contextualizing its influence alongside other factors. The integration of the RAFELA with the RF offers a robust framework to address uncertainties in soil properties, enhancing reliability and efficiency in offshore pipeline engineering.

Stochastic Analysis of Safety Factors for Buried Box Pipelines in Spatially Random Clay

期刊论文
发表日期：

2025年2月8日

查看全文

Khamnoy Kounlavong,
Jim Shiau,
Kongtawan Sangjinda,
Suraparb Keawsawasvong,
Pitthaya Jamsawang,
Bounhome Chansavang

摘要

A significant aspect of offshore pipeline engineering involves evaluating the uplift resistance and failure probability of buried pipelines in clay, which are affected by factors such as pipeline geometry, soil characteristics, material properties, and loading conditions. Subsea marine clay is generally not homogeneous, leading to variations in undrained shear strength vertically and horizontally. As a result, the stochastic analysis method is suitable for accurately modeling such soil conditions. This study addresses these challenges using the Random Adaptive Finite Element Limit Analysis (RAFELA) to analyze the mean uplift resistance factor and the probability of failure for buried rectangular box pipelines in random clay. Seven key parameters are considered in the parametric study: the embedment depth ratio (H/B = 0.5, 1, 2, 4, and 6), width-to-height ratio (L/B = 0.5, 1, 2, 3, and 4), overburden factor (γH/μ_c = 0, 0.5, and 1), adhesion factor (α = 0, 0.5, and 1), load inclination (β = 0°, 45°, and 90°), coefficient of variation (CoVμ_c = 25% and 60%), and spatial correlation length (Θ_c = 0.125, 0.5, 1, 2, 4, and 8). The results are presented as dimensionless uplift resistance factors (μN_ran), probability of failure (P_f), as well as the corresponding safety factor (FS) for designing pipelines in random clay, ensuring practical designs that are both efficient and reliable. Additionally, this study compares its findings with pullout capacity factors derived from deterministic analyses reported in the literature. This study incorporates machine learning, specifically the Random Forest (RF) algorithm, to predict P_f based on parametric data. The RF model, trained on 500 samples (70% training, 30% testing), achieves high predictive accuracy, with R² values of 99.12% and 97.29%, respectively. The Shapley Additive Explanations (SHAP) analysis identifies FS as the most influential factor, directly contributing to the reliability of the pipeline design, while α has the least impact. The analysis emphasizes the practical significance of FS in reducing failure probabilities while contextualizing its influence alongside other factors. The integration of the RAFELA with the RF offers a robust framework to address uncertainties in soil properties, enhancing reliability and efficiency in offshore pipeline engineering.