DATA-DRIVEN AEROACOUSTOELASTIC ASSESSMENT OF LAUNCHER PAYLOADS

This work presents a data-driven vibroacoustic methodology for the prediction of the internal acoustic environment of launcher payload fairings during atmospheric ascent. The proposed framework combines machine-learning-based estimation of external aeroacoustic loads with a reduced-order vibroacoustic model of the fairing–payload system. The external wall-pressure fluctuations acting on the launcher fairing were reconstructed using supervised machine-learning models trained on experimental data acquired during transonic wind-tunnel tests on a scaled VEGA-type launcher model. The dataset includes pressure measurements collected under different Mach numbers and incidence angles and processed in terms of onethird octave band sound pressure levels. Several machine-learning approaches were evaluated, and the bagged-tree ensemble method demonstrated the best predictive performance, thus being selected for the reconstruction of the aeroacoustic loads. The predicted aerodynamic loads were used as input for a one-way vibroacoustic model based on modal formulations for both structural dynamics and cavity acoustics. Three fairing configurations with increasing payload complexity were analyzed: an empty fairing, a simplified payload configuration, and a detailed satellite model including antennas and solar panels. The results demonstrate that payload geometry significantly affects the acoustic modal characteristics and the internal noise environment. In particular, the presence of the payload generally reduces low-frequency acoustic levels while modifying the spatial distribution of the acoustic field inside the fairing.