Application of machine learning (ML) and genetic algorithm (GA) to optimize window wing wall design for natural ventilation

The recent global pandemic has shown the importance of natural ventilation not only in reducing energy consumption but also in reducing airborne contamination. Long-established studies have demonstrated the advantages of window wing walls – vertical projections attached to windows that create a pressure change near the openings - to increase indoor wind flow. However, these conventional strategies to improve natural ventilation rarely take into account site-specific conditions such as dynamic wind conditions throughout the year. This research aims to find a design configuration for window wing walls to improve air circulation through careful interventions for the whole year in classrooms with conventional one-sided openings. With data-driven design, natural ventilation can be maximized to reduce the risk of contamination by insufficient fresh air. The paper utilizes the Computational Fluid Dynamics (CFD) simulations and Artificial Neural Networks (ANN) to predict indoor air movement with less computational time and load. Coupled with Genetic Algorithm (GA), the paper develops an approach that would enable increased natural ventilation in rooms with one-sided windows. Results: show an increase in indoor wind speed in the optimized case compared to the baseline conditions. The main contribution of this study is to demonstrate the use of advanced CFD simulation to provide designers and users with a site-specific configuration for installing fixed wing walls that can maximize indoor wind flow from different wind directions over the whole year, not just for one wind direction.