Robust impact localisation on composite aerostructures using kernel design and Bayesian fusion under environmental and operational uncertainties

Abstract

Impact localisation on composite aircraft structures remains a significant challenge due to operational and environmental uncertainties, such as variations in temperature, impact mass, and energy levels. This study proposes a novel Gaussian Process Regression framework that leverages the order invariance of time difference of arrival (TDOA) inputs to achieve probabilistic impact localisation under such uncertainties. A composite kernel function, combining radial basis function and cosine similarity kernels, is designed based on wave propagation dynamics to enhance adaptability to diverse conditions. Additionally, a task covariance kernel is introduced to enable multitask learning, facilitating the joint prediction of spatial coordinates while capturing interdependencies between outputs. To further improve robustness and accuracy, Bayesian model averaging is employed to dynamically fuse kernel predictions, assigning adaptive weights that account for varying conditions. Extensive experimental validation on a composite plate, including scenarios with large-mass drop tower impacts and small-mass guided drop mass impacts, demonstrates the proposed method's robustness and generalisability. Notably, the framework achieves accurate localisation without requiring compensation strategies for variations in temperature or impact mass, highlighting its suitability for real-world applications. The study also highlights the critical role of sample standardisation for preprocessing TDOA inputs, demonstrating its superiority over feature standardisation by preserving TDOA order invariance and enhancing model compatibility. These advancements establish the proposed method as a reliable and effective solution for structural health monitoring in complex and uncertain operational environments.