Wind-Induced Gravel Blow-Off: Experimental Investigations and Predictive Modeling for Resilient Roof Design

Roof gravel blow-off during severe windstorms is a major hazard, leading to roof damage and significant downwind impacts. Understanding the initiation mechanisms for gravel blow-off is critical for improving building design and mitigating windborne debris risks. Existing experimental data on the wind speed required to initiate gravel blow-off have been collected from high-speed, full-scale tests at wind tunnels. While these tests form the basis of current design guidelines, they are expensive and can lack generalizability. As a result, rooftop gravel is frequently blown off at wind speeds below the design thresholds. This dissertation investigates the onset of rooftop gravel motion, integrating full-scale wind tunnel data with advanced predictive modeling to provide a comprehensive framework for assessing blow-off initiation criteria. The experimental program consisted of three phases. In the first phase, pressure and shear stress measurements were conducted on the surface of a smooth, impermeable roof of a mid- to low-rise square-plan building. These tests were performed at the FIU Wall of Wind Experimental Facility (WOW-EF) using Irwin probes. Measurements were taken for seven parapet heights and nine wind angles. The collected pressure data aligned well with previously published data for similar building geometries, confirming the reliability of the Irwin probes, even in highly separated flow conditions. Classic V-shaped patterns were observed in both the pressure and shear stress data for cornering flows. The second phase extended these measurements to a gravel-covered roof, revealing that both pressure and shear stress magnitudes were consistently lower on the gravel-covered roof compared to the smooth roof. Notably, the highest mean shear stress values occurred in the zero-parapet condition and decreased steadily with increasing parapet height. This trend aligns with the observed blow-off wind speeds, which were lowest for zero parapet height and increased as the parapet height increased. In the third phase, destructive blow-off tests were conducted to determine the wind speeds required to initiate continuous gravel scour. During these tests, the roof was covered with gravel, and wind speeds were incrementally increased until sustained blow-off was observed. Although the tests were conducted for a limited range of cases, analysis of the data identified a critical non-dimensional shear stress value that governed blow-off initiation across various parapet heights and wind angles. This critical value enables the prediction of gravel blow-off wind speeds using surface shear stress coefficients and gravel properties, eliminating the need for costly and time-intensive destructive testing. To further advance predictive capabilities, a convolutional neural network (CNN) model was developed to estimate gravel roof shear stress distributions from smooth roof pressure data. Validated using small-scale blow-off tests at the Clemson University Boundary Layer Wind Tunnel (CUBLWT), the model demonstrated strong predictive capabilities, providing a scalable and cost-effective tool for engineering applications. This integration of empirical and computational approaches has the potential to transform predictive modeling for wind-driven debris phenomena, paving the way for safer and more cost-effective design practices.