Impact Load Reconstruction for Composite Structures Using a Gated Temporal Convolutional Network

Techniques for identifying impact forces are crucial for improving the operational reliability and safety of aerospace and railway composite structures. To address the challengesin impact load reconstruction arising from the complex nature of composite materials, thispaper proposes a novel data-driven method based on a Gated Temporal Convolutional Network (GTCN). The proposed framework establishes a direct inverse mappingfromstructural vibration responses, captured by a network of piezoelectric (PZT) sensors, totheimpact force's time history. The GTCN architecture leverages dilated convolutionstoexponentially expand the receptive field for capturing long-term dependencies insensorsignals, while a gating mechanism is employed to selectively filter critical features. Theefficacy of the proposed method was validated through a comprehensive finite element simulation dataset, encompassing various impact scenarios on a T700/BA9916CFRPlaminate. Simulation results demonstrate that the GTCN method achieves robust impact forcereconstruction with an average relative error of 5.8% across different impact types (Gaussian, half-sine, and sawtooth). The proposed method demonstrates considerable potential forapplication in advanced structural health monitoring (SHM).