PHYSICS-INFORMED NEURAL NETWORKS FOR HIGH-VELOCITY PENETRATION IN MULTI-LAYER COMPOSITE
Abstract
This study presents a physics-informed neural network (PINN) framework for predicting projectile penetration depth, residual velocity, and erosion in multi-layer composite targets. The governing penetration equation is embedded into the learning process, ensuring physically consistent solutions. The depth-dependent field networks capture velocity decay and progressive projectile mass loss through nose and shank erosion. The model is trained on numerical data and validated on experimental data across a wide range of impact velocities for a layered structure comprising soil, a boulder, steel, HDPE, and reinforced concrete. The results show accurate predictions of penetration and erosion trends, with strong generalization to unseen conditions. The proposed framework provides a fast, physics-consistent alternative to conventional numerical simulations for layered protective structures.
DOI
10.12783/ballistics26/37737
10.12783/ballistics26/37737
Refbacks
- There are currently no refbacks.