Low-velocity impact positioning on composite honeycomb panels using a DPL-R regression

This paper proposes a low-velocity impact localization method for carbon fiber/aluminum honeycomb panels using a regression framework based on Discriminative Projective Dictionary Pair Learning (DPL-R). Multi-channel FBG responses are converted into statistical features and mapped through a discriminative dictionary pair to obtain sparse representations, which are used for coordinate regression with a residual-weighted strategy. The method requires no wave-velocity modeling or finite-element simulation, making it suitable for complex panels under low sampling rates. Experiments on a 14×14 training grid and 20 free impacts show an average error of 3.64 mm, demonstrating its effectiveness for satellite-borne composite panel monitoring.