Numerical simulation and machine learning framework approach for ballistic performance evaluation of aluminium structures

• A comprehensive finite element analysis of high-velocity impacts on aluminum plates was conducted for blunt, hemispherical, and ogival projectile geometries. • Distinct fracture mechanisms—adiabatic shear plugging, petaling, and ductile hole enlargement were identified based on projectile nose shapes. • Numerical results using the Johnson–Cook material model showed excellent agreement with experimental absorbed energy measurements. • Machine learning models (SVR, RF, DT, MLP) were developed to accurately predict absorbed energy, significantly reducing computational cost. • The hybrid FEM-ML approach provides a rapid and reliable predictive framework for optimizing lightweight protective structures under ballistic loading. Understanding and accurately predicting the ballistic response behavior of structural materials is critical for the design and optimization of lightweight, high performance protective structures subjected to high velocity impact. This study presents a comprehensive numerical analysis of the ballistic perforation of aluminum plates impacted by rigid steel projectiles with blunt, hemispherical, and ogival nose geometries. Finite Element Method (FEM) simulations incorporating the Johnson–Cook constitutive and failure models were conducted in ABAQUS/Explicit scheme to evaluate absorbed energy during impact. The simulation results closely match experimental data, with an average deviation of less than 7% in numerically predicted absorbed energy values. To overcome limitations associated with computational cost and model complexity, supervised Machine Learning (ML) models which includes Support Vector Regression (SVR), Random Forest (RF), Decision Tree (DT), and Multi-Layer Perceptron (MLP), were developed to predict absorbed energy based on projectile geometry, impact velocity, and target thickness. Comparative assessments demonstrated that ML-based models maintained strong predictive accuracy. Detailed analysis of fracture patterns provide additional insights into failure mechanisms under different impact scenarios. The integrated FEM–ML framework proposed herein demonstrates enhanced predictive accuracy and computational efficiency, offering a promising tool for rapid ballistic performance evaluation and impact resistant design optimization.