Data-Driven Optimization of Geotechnical Parameters Using Machine Learning with Numerical Validation: Pile-Anchor-Soil Nailing Structures in Coastal Composite Strata

Coastal infrastructure development in soil–rock composite strata faces unique challenges due to marine geohazards such as tidal fluctuations, storm surges, and saltwater intrusion. Based on a deep foundation pit support system in a coastal area, this study used ABAQUS finite-element software in conjunction with Python’s automation capabilities for modeling, simulation, and postprocessing to establish a three-dimensional numerical model of a pile-anchor and soil nail combined support structure in a soil–rock composite stratum. On this basis, the orthogonal test method was employed to optimize and analyze the support structure parameters that affect the stability of the foundation pit. The preprocessed deep foundation pit horizontal displacement dataset was divided into training set and test set in a ratio of 7∶3, and a deep foundation pit deformation situation prediction model based on a genetic algorithm–long short-term memory network was constructed and verified. The results show that the horizontal and vertical displacements of the pile top obtained from the numerical simulation differ from the field monitoring data by 1.05 and 0.55 mm, respectively, indicating that the finite-element model established in this study has high accuracy and reliability. The orthogonal test results reveal that the factor with the greatest influence on foundation pit stability is the pile spacing, followed by the horizontal anchor distance, and finally the anchor inclination angle. The optimized support parameters are a horizontal anchor spacing of 1,500 mm, a pile spacing of 1,700 mm, and an anchor incident angle of 17°. The optimized support structure significantly reduced the horizontal and vertical displacements of the pile top, with reductions of 1.82 and 1.88 mm, respectively. This parametric approach provides predictive insights for coastal geohazard prevention.