Impact-induced damage mechanisms in PPS-based laminates: experimental characterization and ANN prediction of residual mechanical properties

Abstract Quasi-isotropic laminates with two stacking sequences ([0, ± 60] s and [0, ± 45, 90] s ) were subjected to low-velocity impacts at varying energies. Damage characterization via C-scan and microscopy identified key failure modes, including delamination, fiber breakage, and matrix plasticization. The effect of this damage was quantified through compression-after-impact (CAI) and four-point bending tests, which confirmed progressive strength degradation with increased impact energy. The core innovation of this study is the development of a feedforward neural network (FNN) model that uses impact energy and stacking sequence as inputs to predict residual mechanical properties. The model demonstrated exceptional accuracy, with a maximum deviation of only 1.23% from experimental results. This high-fidelity predictive tool offers a significant advancement over traditional methods, enabling efficient and accurate structural integrity assessment for impacted composite structures.