Sheet piles, as one of the most widely used retaining systems, play a vital role in stabilizing excavations an‎d underground structures by controlling lateral earth pressures an‎d ensuring construction safety. The design of these systems, particularly the determination of embedment depth an‎d factor of safety, requires high precision, as computational errors can lead to collapse, excessive settlements, an‎d irreparable damages. In this research, to facilitate an‎d accelerate the design process, an intelligent model based on the Ran‎dom Forest algorithm was developed. The required data were generated through 150 numerical simulations in PLAXIS 2D, considering a wide range of geotechnical an‎d structural conditions for both cantilever an‎d anchored systems. The numerical model was validated by comparing results with a reputable scientific study. After data cleaning, the intelligent model was trained using 100 valid datasets, an‎d its performance was eva‎luated using R² an‎d MSE metrics, indicating acceptable accuracy, particularly in predicting the embedment depth for anchored systems. Feature importance analysis revealed that in cantilever systems, excavation depth an‎d internal friction angle are the most influential parameters, whereas in anchored systems, anchor-related parameters play a decisive role. Three-dimensional surface plots further confirmed the nonlinear relationships between input an‎d output variables. The developed model is introduced as a fast an‎d practical tool for preliminary design stages of sheet pile-supported excavations, capable of significantly reducing time an‎d cost

حميد هاشم الحسيني , عليرضا باغبانان

محمدعلي روشن ضمير , سعيد مهدوي