An Effective Impact Detection Method for Composite Curved Panel
ANNs are only accurate for the scope of the given training data which is not suitable for real life impact localisation due to the large range of possible impact variation. Impact data was collected for a variation of impact cases (angle, mass and energy) on a sensorized curved composite panel. From observation of the obtained data, a robust signal Time of Arrival (TOA) extraction method is proposed using a Normalised Smooth Envelope Threshold (NSET) which is a modification of the currently known Normalised Threshold (NT) method. Two ANNs were trained using TOA extracted with the NT and NSET method from a baseline case and tested with TOA extracted from cases having added variation of impact condition. The results show that the proposed NSET method results in more accurate results for impact cases different to the training case and thus allows for only a single impact training case to accurately predict cases with multiple variation. This enhances the applicability of ANNs for impact localisation in real life conditions.
Luis Rodríguez-Tembleque, Jaime Domínguez and Ferri M.H. Aliabadi
A. H. Seno et al., "An Effective Impact Detection Method for Composite Curved Panel", Key Engineering Materials, Vol. 774, pp. 529-534, 2018