Exploring wind load effects on structures: An insight into machine learning applications

The NIST-UWO database has pressure coefficient time-history data, encompassing various roof slopes, eave heights, terrain exposures, and wind angles. Utilizing SAP2000 to obtain the influence coefficients (IC) for eave and ridge moments and displacements, corresponding critical moment and displacement coefficients were computed for three different gable roof pitch (1/4:12,1:12, and 3:12) models each having three different eave heights of 7.32 m, 9.75 m, and 12.19 m, in two terrain types – open country and suburban. The study utilized Decision Tree (DT), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) to predict these load effect coefficients for potential missing wind angles. Additionally, the study compared these machine learning models' performance in handling exposure categories as numerical values (roughness length) and categorical variables (represented via one-hot encoding). The results showed that all models performed consistently well, regardless of exposure category representation, with XGBoost demonstrating better performance compared to RF and DT.