Surrogate model-based online quantitative prediction of pitting damage in spacecraft

To ensure structural reliability under hypervelocity impact (HVI) of the micrometeoroid and orbital debris (MMOD), Space Stations are typically equipped with Whipple shields. As a result, the secondary debris cloud is introduced, leading to a complex type of pitting damage, which usually initiates at an imperceptible scale but undermines the structural integrity and jeopardizes on-orbit safety. So far, its online quantitative evaluation has become increasingly imperative but still faces significant challenges. To address this limitation, a generative deep learning-based surrogate model (termed AE-PDSM) is proposed for the online prediction of pitting damage using in-situ acoustic emission (AE) signals. After defining two quantitative characterization indices based on a radial discretization scheme, the AE-PDSM establishes an end-to-end mapping from AE-based time–frequency spectra to two-dimensional geometric morphology of pitting damage. Then, a hybrid finite element–smoothed particle hydrodynamics (FE-SPH) framework is employed to construct a training database covering representative debris cloud impact conditions. Numerical and experimental validations demonstrate that the proposed approach identifies the perforation pattern with an accuracy of 93.3 %, and predicts the pitted morphology with low error (MSE = 0.038) and high structural similarity (SSIM = 0.942). By enabling online real-time quantification of debris cloud-induced pitting damage via in situ AE measurements, this study provides a feasible pathway toward on-orbit structural integrity evaluation for spacecraft shielding structures.