Probabilistic analysis of active earth pressures in spatially variable soils using machine learning and confidence...

Area of Science:

• Geotechnical Engineering
• Computational Mechanics
• Risk Analysis

Background:

• Soil properties like friction angle and unit weight often exhibit spatial variability.
• Accurate assessment of active earth pressures is crucial for safe and economical geotechnical designs.
• Existing methods may not fully capture the impact of soil property randomness on stability.

Purpose of the Study:

• To develop a probabilistic framework for active earth pressure assessment considering soil spatial variability.
• To integrate advanced computational techniques and machine learning for efficient analysis.
• To provide practical design tools for engineers dealing with geotechnical uncertainties.

Main Methods:

• Modeling soil properties using random fields with log-normal distributions and spatial correlation.
• Employing Monte Carlo simulations (MCS) integrated with finite element limit analysis (FELA).
• Utilizing Multivariate Adaptive Regression Splines (MARS) for failure probability prediction and a two-phase optimization for hyperparameter tuning.
• Incorporating confidence intervals for prediction reliability and adaptive finite element meshes for stochastic failure mechanisms.

Main Results:

• A probabilistic framework effectively assesses active earth pressures under spatial soil variability.
• Machine learning models (MARS) accurately predict failure probabilities, improving computational efficiency.
• Adaptive meshes capture complex stochastic failure mechanisms, revealing insights into spatial variability impacts.
• Parametric contour design charts are generated for practical engineering application.

Conclusions:

• The combined approach enhances geotechnical design by quantifying uncertainties associated with soil variability.
• The developed framework and tools enable more reliable and cost-effective geotechnical solutions.
• This research offers a robust methodology for engineers to optimize safety margins in variable soil conditions.