Low-velocity impact (LVI) damage in hybrid fiber-reinforced polymer composites poses a critical challenge for lightweight structural applications due to its dependence on laminate architecture and impact severity. In this study, an integrated experimental–explainable artificial intelligence framework is developed to quantify, predict, and interpret LVI response of carbon/flax bio-hybrid composite laminates across range of impact energies. An energy-normalized impact damage index (IDI) is introduced to consolidate peak impact force, damage area, and damage extension into single dimensionless metric, enabling energy-wise comparison of impact damage tolerance. The framework is developed using experimentally measured LVI responses obtained from instrumented drop-weight impact tests. Independent Extreme Gradient Boosting (XGBoost) regression models are trained to predict each impact response using ply-wise laminate encoding and impact energy as inputs, achieving strong generalization with test-set R 2 values of 0.963 for peak impact force, 0.985 for damage area, and 0.901 for damage extension. To provide model interpretability, Shapley Additive exPlanations (SHAP) are applied to trained XGBoost models to quantify feature contributions to model predictions and identify response-specific governing plies. The explainable analysis reveals that impact energy dominates all responses, while peak impact force is controlled by mid-plane stiffness, damage area by sub-surface ply stiffness, and damage extension by ply-level compliance near the mid-plane. These findings provide physically interpretable explanation consistent with experimental observations of symmetric flax-distributed laminates at intermediate energies. The proposed IDI–XGBoost–SHAP framework advances composite impact analysis from empirical assessment to interpretable ply-level prediction and provides a data-driven pathway for designing sustainable, damage-tolerant bio-hybrid laminates.
Masud et al. (Sun,) studied this question.