Integrated machine learning–finite element analysis methodology for assessing the ballistic performance of ceramic/aluminum armors

Abstract

This study presents a comprehensive numerical and data-driven investigation into the ballistic performance of hybrid armor configurations consisting of alumina ceramic strike faces and aluminum alloy backing plates. Two distinct projectile types, namely the 7.62 × 51 mm P80 and the 12.7 × 99 mm AP rounds, were considered to assess the armor response under medium- and large-caliber threats. The alumina ceramic plates, each with dimension of 97.6 × 97.6 mm², were modeled with thicknesses ranging from 10 to 15 mm. Three aluminum backing materials (AA5083-H116, AA6061-T6, and AA7075-T651) were incorporated with thicknesses of 5, 6, 7, and 8 mm and dimensions of 200 × 200 mm². Ballistic impact simulations were conducted using LS-DYNA. Four impact velocities, varying from 850 to 1000 m/s in increments of 50 m/s, were employed to examine velocity-dependent penetration modes. Residual velocities obtained from the finite element analyses constituted the primary dataset for machine learning (ML) modeling. Four ML algorithms (Elastic Net regression, Multilayer Perceptron (MLP), Decision Tree (DT), and Support Vector Machine (SVM)) were trained to establish predictive frameworks capable of estimating residual projectile velocity from geometric, material and loading parameters.