Integrated machine learning for modeling bearing capacity of shallow foundations

Area of Science:

• Geotechnical Engineering
• Civil Engineering
• Computational Mechanics

Background:

• Footing stability analysis is crucial in geotechnical engineering.
• Two-layered soil masses present complex challenges for traditional analysis methods.
• Accurate prediction of bearing capacity is essential for structural safety.

Purpose of the Study:

• To develop and evaluate novel predictive tools for footing stability on two-layered soil.
• To employ artificial neural networks (ANN) optimized by backtracking search algorithm (BSA) and equilibrium optimizer (EO).
• To approximate the stability value (SV) and bearing capacity of footings.

Main Methods:

• Utilized artificial neural networks (ANN) for predictive modeling.
• Employed backtracking search algorithm (BSA) and equilibrium optimizer (EO) for ANN training.
• Conducted finite element analyses to determine system stability based on settlement criteria (5 cm threshold).

Main Results:

• Both BSA and EO significantly improved ANN training, reducing root mean square error (RMSE) by up to 17.46% (EO).
• Prediction accuracy, measured by area under the receiving operating characteristics curve (AUROC), increased to 94.3% (BSA) and 94.1% (EO).
• The equilibrium optimizer (EO) demonstrated higher accuracy, while the backtracking search algorithm (BSA) proved more time-efficient.

Conclusions:

• The developed ANN models, particularly the EO-ANN, are effective tools for analyzing footing stability in two-layered soils.
• An explicit mathematical formula derived from the EO-ANN model offers a practical method for stability value prediction.
• The study highlights the potential of optimization algorithms in enhancing geotechnical engineering analyses.