A combine approach of soft computing and system reliability using sequential compounding method for geogrid reinforced retaining wall analysis

This study integrates machine learning (ML) algorithms and system reliability analysis to assess the stability of geotextile-reinforced retaining walls. The research utilized ensemble-based ML techniques such as Random Forest (RF), Gradient Boosting (GBM), and Extreme Gradient Boosting (XGB), alongside the First-Order Second Moment (FOSM) method, the study evaluates multiple failure scenarios sliding, overturning, bearing capacity, rupture, and pullout with soil parameters treated as random and geometric parameters as deterministic. Results indicate that RF and GBM outperform XGB, achieving R2 accuracy, while XGB exhibits variability, particularly in sliding and rupture conditions, suggesting sensitivity to data distribution. Lower RMSE and RSR values for RF and GBM confirm minimal errors, while a Willmott Index (WI) above 0.99 reflects strong agreement between predicted and actual values. The Bias Factor remains close to 1.0, ensuring unbiased estimations, while low TIC and sMAPE values highlight superior generalization between training and testing datasets. System reliability analysis reveals a system reliability index (B𝑠𝑦𝑠𝑡𝑒𝑚 = 1.083), lower than the minimum component reliability index (B𝑚𝑖𝑛 = 1.213) from FOSM, emphasizing the need for a comprehensive probabilistic assessment. The study demonstrates that machine learning, particularly GBM and RF, provides robust predictions, improving the reliability evaluation of reinforced retaining walls.