Hybrid artificial neural network models for bearing capacity evaluation of a strip footing on sand based on Bolton failure criterion

This paper employs the Bolton failure criterion, incorporating strength-dilatancy relationships, to analyze the bearing capacity factor of a strip footing on dense sand. Utilizing finite element limit analysis (FELA) based on the lower and upper bound theorems, the study presents the results as average bound solutions. By using the Bolton model, the b parameter is first calibrated and found that it should be about 3.50 to align the ultimate bearing capacity (q_u) from FELA to have a good agreement with that from experimental test results from previous studies. The influence of parameters relevant to the Bolton failure criterion is analysed, showing that an increase in relative density (D_R) significantly affects the variation in the bearing capacity factor (N_γ) at higher Q values, while lower Q values inhibit dilatancy due to soil crushing. The width of the strip footing (B) has a decreasing effect on N_γ at higher Q values, and the unit weight (γ) changes minimally impact N_γ within the range of 16–22 kN/m³. Additionally, an increase in the critical state friction angle (ϕ_cv) consistently increases N_γ, highlighting its direct correlation with soil shear strength. A hybrid artificial neural network (ANN) model integrates machine learning with four optimization algorithms: Imperialist Competitive Algorithm (ICA), Ant Lion Optimization (ALO), Teaching Learning Based Optimization (TLBO), and New Self-Organizing Hierarchical Particle Swarm Optimizer with Jumping Time-Varying Acceleration Coefficients (NHPSO-JTVAC). Comparative rank analysis of hybrid ANN models based on the selection of the optimal number of hidden neurons demonstrates that the ANN-TLBO model excels in predicting the bearing capacity factor, achieving a score of 48. This conclusion is corroborated by an error heatmap matrix, which indicates a minimized percentage of error relative to other hybrid ANN models. Importance analysis identifies particle crushing strength (Q) as the most significant factor influencing the bearing capacity factor (N_γ).