Research Experiences for Undergraduates (REU), NHERI 2023: Deep Learning Based Estimation of Peak Wind Pressures on Buildings from Short Duration Measurements

This project presents a deep learning model to predict peak wind pressures experienced on vertical building surfaces from statistical features extracted from short durations of boundary-layer wind tunnel data with varying wind conditions. The multi-layer perception model yielded excellent results in training and testing and is thus considered a viable solution to determining peak wind pressures from fewer data, proving its potential in reducing costs associated with boundary-layer wind tunnel experiments and computational fluid dynamics simulations. This project is unique as it utilizes short-duration measurements to determine peak wind pressures. The intended audience is those interested in machine learning applications to wind engineering.