Abstract
This study proposes a vibration-based approach for detecting and quantifying sub-slab corner voids in airport cement concrete pavement. Scaled down slab models were constructed and subjected to controlled moving load simulations. Acceleration signals were collected and analyzed to extract time-frequency domain features, including power spectral density (PSD), skewness, and frequency center. A finite element model incorporating contact and nonlinear constitutive relationships was established to simulate structural response under different void conditions. Based on the simulated dataset, a random forest (RF) model was developed to estimate void size using selected spectral energy indicators and geometric parameters. The results revealed that the RF model achieved strong predictive performance, with a high correlation between key features and void characteristics. This work demonstrates the feasibility of integrating simulation analysis, signal feature extraction, and machine learning to support intelligent diagnostics of concrete pavement health.