Novel hybrid XGBoost-based soft computing models for predicting penetration resistance of buried pipelines in cohesive soils

Created by W.Langdon from gp-bibliography.bib Revision:1.8414

@Article{Kounlavong:2024:oceaneng,

• author = "Khamnoy Kounlavong and Laith Sadik and Suraparb Keawsawasvong and Pitthaya Jamsawang",
• title = "Novel hybrid {XGBoost-based} soft computing models for predicting penetration resistance of buried pipelines in cohesive soils",
• journal = "Ocean Engineering",
• year = "2024",
• volume = "311",
• pages = "118948",
• keywords = "genetic algorithms, genetic programming, Penetration resistance, Pipeline, Limit analysis, XGBoost, Machine learning",
• ISSN = "0029-8018",
• URL = "https://www.sciencedirect.com/science/article/pii/S0029801824022868",
• DOI = "doi:10.1016/j.oceaneng.2024.118948",
• abstract = "In this paper, both the finite element limit analysis (FELA) and soft computing techniques of four hybrid XGBoost (XGB) models, namely, GA-XGB, optimised with Genetic Algorithms; SMA-XGB, optimised with Slime Mould Algorithms; PSO-XGB, optimised with Particle Swarm Optimisation; ACO-XGB, Ant Colony Optimisation; and one Genetic Programing model (GP), are employed to develop surrogate models for predicting the undrained penetration resistance of buried pipelines embedded in clays. The penetration resistance of pipelines is caused by internal and external pressures, resulting in a force that acts on the pipeline in the downward and inclined directions at the same time. The penetration resistance factor (N) of a buried pipeline is determined based on four dimensionless variables, namely, the buried depth ratio (H/D), the inclination angle of the applied load (beta), the soil strength or overburden ratio (gammaH/c), and the adhesion factor at soil-pipeline interfaces (alpha). The findings of this study are presented and summarised in the form of charts for dimensionless penetration resistance factors (N) and failure mechanisms of pipelines. Furthermore, the results from this study are compared with the results from previous studies and agrees well with those of previous studies. According to the results of the four developed hybrid XGBoost-based models, SMA-XGB outperforms GA-XGB, PSO-XGB, and ACO-XGB with no symptoms of overfitting, while the GP model resulted in an adequately accurate model with the advantage of a closed-form equation; its determination correlation coefficient of training and testing equals 0.98 (R2 = 0.980)",

}

Genetic Programming entries for Khamnoy Kounlavong Laith Sadik Suraparb Keawsawasvong Pitthaya Jamsawang

Citations