Aerodynamic Shape Optimization of Tall Buildings
By: Xinwen He
Department: Civil Engineering
Faculty Advisor: Dr. Zhaoshuo Jiang
This study proposes an innovative approach to use Gaussian process-based active learning to develop surrogate models for identifying the most efficient shape of high-rises. The study is part of a recent NSF-funded project that addresses the grand challenge of wind engineering in tall building designs by leveraging the efficiency of numerical simulations and the reliability of large-scale experimental validation through a cyber-physical approach. This project intends to leverage wind tunnel testing, machine learning, and advanced manufacturing to advance shape optimization of high-rises. In order to establish a transformative cyber-physical aerodynamic shape optimization framework that enables the automated exploration of the design space as a continuum, the first and crucial step is to develop a robust algorithm-guided surrogate (i.e., approximation of a complex physical system) modeling strategy that (1) enables reliable and efficient exploitation search for optimal design candidates; (2) allows fast and accurate design space exploration that gives designers more flexibility to rapidly investigate alternative designs.