A Data-Driven Smart Morphing Façade System Controller for Wind-Induced Vibration Mitigation in Tall Buildings

The wind vortex system that forms around a tall building when attacked by the wind causes the forces affecting the building to fluctuate. As a result, Wind-Induced Vibration (WIV) is a concern for tall buildings even when subject to steady, unidirectional wind speeds. This study presents a data-driven control system that utilizes smart morphing facades (Smorphacades) to effect changes to the external geometry of the building. By choosing geometries that generate a vortex system to which the building is less sensitive, the severity of the WIV is shown to be reduced. The Smorphacades are currently deployed on many buildings to provide shading and/or solar energy harvesting, which reduces the investment costs of the system. The study also shows that only a subset of the façade modules need to be controlled, which leaves the rest of them to be used for their original purpose. The façade is reduced in this study to a set of plates that have a single active orientation angle to simplify the parametric study. A 2-dimensional (2D) Fluid-Solid-Interaction (FSI) simulation technique for individual floors of the building is developed to estimate the relationship between the plate angles and the WIV amplitudes. The floor is considered to be a rigid body and the model estimates the planar floor deflections using an uncoupled second order 3 Degree-of-Freedom (DOF) system. The resulting floor motions are implemented using a dynamic mesh. The plates are placed inside mesh bubbles that expose interpolation interfaces with the rest of the mesh, which enables them to be rotated while the simulation is running. The simulation model is validated using a wind tunnel experiment for a scaled Commonwealth Advisory Aeronautical Research Council (CAARC) standard building from the literature. Additionally, an experimental apparatus consisting of scaled CAARC building floors at the top of an oscillating tube is designed and implemented. The apparatus is setup inside a wind tunnel and subjected to speeds that mirror the dynamic behavior of the full-scale building at severe conditions (i.e. resonance). The simulation as well as the experimental results are used to learn a predictor Artificial Neural Network (ANN). A novel ANN learning procedure guided by Genetic Algorithms (GAs) is also presented to speed up the learning of the ANN as well as improve its prediction in the vicinity of plate angle combinations that reduce the WIV. For online operation, another ANN (optimizer) is used as a look up for the best plate angle combinations for different wind conditions. Both ANNs are continuously updated after deployment based on wind and vibration measurements. This allows the system to automatically adapt to changes in the structure or the environment, which is an advantage in comparison to Tuned Mass Damper (TMD) or Tuned Liquid Damper (TLD). The results show a major reduction in WIV amplitudes when the controller is utilized.