Material-aware rapid prediction of low-velocity impact damage in laminated composites via surrogate modeling

Abstract

Low-velocity impact (LVI) can leave laminated composites with internal damage that is difficult to observe visually but may still degrade structural integrity. Finite-element simulation can resolve such damage fields, yet repeated case-specific analyses become costly when impact direction, ply orientation, stacking sequence, and material system vary together. Existing surrogate models often treat material information as a simple label and may overlook low-intensity regions in finite-element-generated fiber damage fields. This study develops a material-aware surrogate framework for rapid prediction of finite-element-generated fiber damage fields. The framework learns compact representations of damage images and maps physical parameters, including impact angles, ply position, ply orientation, and material type, to fiber tensile and fiber compressive damage fields. Validation shows that the model preserves global damage morphology, captures fiber orientation-induced rotation, represents symmetry loss under oblique impact, and adjusts damage intensity across material systems under the controlled lay-up setting considered in this work. With an inference time of approximately 5 ms, the framework provides a fast alternative to repeated finite-element analysis for preliminary impact damage evaluation and design screening.