An artificial intelligence-based hybrid method for multi-layered armour systems

The design of protective structures is a complex task mostly due to threat-related unknowns, such as the exact kinetic energy of the impactor and the dominant energy dissipation mechanisms. The design process is often costly and inefficient due to the number of these unknowns and to the cost of necessary steps such as laboratory testing and numerical modelling. In this chapter the authors propose a hybrid method that significantly increases the efficiency of the design process, and consequently decreasing its cost. The method combines an energy-based analytical approach with a set of deep learning (DL) models. Finite Element Analysis (FEA) and experimental results are used to train the artificial intelligence (AI) models and verify and validate the design process. The energy-based analytical method generates solutions for the DL algorithms, which can then be used to find optimal configurations for the protective structure. The proposed deep learning model is a neural network which is trained using experimental results and analytical data, to understand the ballistic response of a specific material, and predict the residual velocity for a given impact velocity, layer thickness and material properties. Networks trained for individual layers of the armour system are then interconnected in order to predict the residual velocity of blunt projectiles perforating multi-layered composite structures. Validation tests are done on systems including single and multi-layered targets.