Jet-air excitation-based deep acoustic sensing for vehicle leakage detection

This study presents an acoustic sensing framework for vehicle leakage detection based on jet-air excitation. Instead of relying on passive leakage signals, compressed air is actively directed toward potential leak regions, generating characteristic acoustic responses through fluid–structure interactions. A theoretical model is developed to predict resonance frequencies as functions of orifice geometry and jet parameters, including diameter and velocity. The model is validated through a series of controlled experiments using aluminum plate specimens with machined orifices, confirming its ability to accurately capture frequency-domain characteristics associated with varying leak sizes. The approach is further applied to vehicle body-in-white components, where location-specific resonance patterns are observed under jet-air stimulation. These results demonstrate the sensitivity of the method to structural complexity and geometric variability. Full-vehicle experiments are conducted by introducing artificial leaks at representative regions, such as the windshield and trunk area, and measuring internal acoustic responses. The observed spectral features consistently distinguish leak conditions from non-leak baselines. A data-driven classification model is trained using spectral features extracted from the measured signals, enabling automated identification of leakage conditions. Overall, the proposed technique offers a physically grounded, non-contact, and scalable alternative to conventional inspection methods, with strong potential for integration into automated vehicle quality assurance processes.