Unraveling seismic uplift behavior of plate anchors in frictional-cohesive soils: A comprehensive analysis through stability factors and machine learning

Unraveling seismic uplift behavior of plate anchors in frictional-cohesive soils: A comprehensive analysis through stability factors and machine learning

This paper explores the uplift behavior of horizontal anchor plates in frictional-cohesive soils subjected to seismic loading. The study focuses on three seismic stability factors: Fc, Fq, and Fγ, which depend on seismic coefficient Kh, depth-to-width ratio H/B, and soil internal friction angle φ. To evaluate ultimate uplift load capacity while considering the influence of Kh, H/B, and φ, Terzaghi's principle of superposition is introduced. Robust predictive models for these stability factors are developed using advanced computational techniques like finite element limit analysis (FELA) and Levenberg-Marquardt backpropagation neural networks (LMBPNN). These neural network models are meticulously optimized, resulting in high R-squared values (R2) and low root mean square errors (RMSE) for both training and testing datasets. Sensitivity analysis and feature importance assessments on the three stability factors indicate that the design parameter H/B has the largest SHAP value, that is followed by φ and Kh. The findings significantly enhance our understanding of uplift behavior in frictional-cohesive soils under seismic loading conditions, as well as providing precise and reliable design of anchor plate systems in diverse soil environments exposed to seismic forces.

Files associated with this item cannot be displayed due to copyright restrictions.