Application of Advanced Machine Learning Models for Uplift and Penetration Resistance in Clay-Embedded Dual Interfering Pipelines

This study investigated the uplift and penetration resistance of dual interfering pipelines buried in clay using advanced regression machine learning models, specifically the group method of data handling (GMDH), genetic programming (GP), extreme gradient boosting (XGBoost), and random forest (RF). The dataset comprises 256 numerical FELA data points for uplift conditions and 384 numerical FELA data points for penetration conditions, marking the first application of these models in this context. To train the models, three input parameters are considered: the spacing ratio (S/D), the embedded ratio (w/D), and the normalized unit weight and increasing strength (γ/ρ). The models predict two output parameters: the vertical uplift resistance ( \(\frac{{q}_{t}}{\rho D}\) $\frac{q_t}{\rho D}$ ) and the vertical penetration resistance \((\frac{{q}_{c}}{\rho D}\) $(\frac{q_c}{\rho D}$ ). Performance metrics were employed to evaluate and compare the effectiveness of each model. The study revealed that the GP model is particularly effective in predicting the uplift and penetration resistance of pipelines. Both external validation and literature validation confirmed the predictive capabilities of the proposed models. Furthermore, the influence of each input parameter was analyzed, resulting in the development of empirical equations for both uplift and penetration conditions. The resulting empirical equations provide dimensionless output parameters, offering practical utility for design practitioners in real-world field conditions. Detailed study results, including comprehensive tables and empirical equations, are presented to facilitate practical applications and enhance the understanding of pipeline-soil interactions in clay environments. These contributions underscore the potential of advanced regression machine learning models in geotechnical engineering and pipeline design.