Internal low-velocity impact damage prediction in CFRP laminates using surface profiles and machine learning

Aircraft operators must maintain the safety of aircraft structures. In order to aim for an easier maintenance of impact damage on the composite structures, the possibility of inferring low-velocity impact (LVI) information in CFRP laminates from the surface damage profiles is verified. This study conducts several low-velocity impact tests considering three factors (stacking sequence, impactor shape, and impact energy), inducing barely visible impact damage on specimens. This is followed by surface profile and internal damage measurements. Subsequently, original features that could contribute to inferring impact information were created from the surface profile. After feature engineering, the predictability of impactor shape, delamination area, and delamination length was confirmed using three machine learning models. The results indicated that the models could infer approximately 80 % of them correctly using dent depth and the volume of indentation. The proposed model enables us to infer non-visible impact information from visible one generally without a great deal of inspections.

Automated summary

This study verifies the possibility of inferring low-velocity impact information in CFRP laminates from surface damage profiles. By conducting low-velocity impact tests and utilizing machine learning models, the impactor shape, delamination area, and delamination length can be predicted from dent depth and indentation volume.