Machine Learning Models for Predicting Deformation in RC Slabs Shielded by Auxetic Sandwich Panels Under Impact

Abstract

The integration of artificial intelligence (AI) into structural engineering is accelerating, with machine learning (ML) techniques showing strong potential for predictive modelling tasks. This study introduces a data-driven framework that systematically predicts the deformation response of a protective system made up of an auxetic sandwich panel resting on a reinforced concrete (RC) slab impacted vertically. Traditional finite element (FE) analysis of such impact scenarios is computationally intensive and may involve manual processes prone to human error. To overcome these limitations, this study combines FE simulations with machine learning (ML) to be able to efficiently predict the structural deformation under varying impact conditions. A series of simulations were conducted in ANSYS, systematically varying the impactor parameters in order to generate a training dataset. The main objective is to develop a predictive model for the deformation of both the auxetic core and the underlying RC slab. Various machine learning regression models—involving ensemble tree-based models (Extra Trees, Random Forest), kernel-based models (Kernel Ridge), and linear models (Linear Regression, Ridge, Lasso)—were trained and evaluated. The performance of the model was measured using the coefficient of determination (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}^{2}$</tex>), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). The Extra Trees model yielded the greatest accuracy predicting core deformation (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}^{2} = 0.98$</tex>), while the Kernel Ridge model was the best for predicting slab deformation (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}^{2} = 0.94$</tex>). Overall, these results confirm the promising aptitude of machine learning for the rapid prediction of structural response for applications in impact protection design, while providing an actionable framework for realizing optimal materials and geometry for layered protective systems.