Dissertation/Thesis Abstract

Applying Computational Fluid Dynamic Simulations and Predictive Models to Determine Control Schedules for Natural Ventilation
by Horin, Brett, M.S., Illinois Institute of Technology, 2018, 84; 10843192
Abstract (Summary)

This thesis investigates natural ventilation in building design, culminating in a final project to design optimal ventilation in an underground parking garage. The aim of this research is to explore a method combining computational fluid dynamic (CFD) simulations with neural networks as a means of performing a robust, yet computationally inexpensive simulation. The final project has the objective of simulating an annual operation schedule for louvers at the openings of the garage to achieve a desired airflow rate. Concepts in computational design and building science are explored to fully capture how the geometric domain of architectural modeling can be expressed in computational parameters to successfully perform effective simulations. It was important to make these workflows accessible to architects, so common software in the architecture industry was used. The results of this project support a coupled approach of using CFD simulations and neural networks to predict airflow parameters of interest. Validation CFD simulation results were compared to the results using the neural network and they were in good agreement. Ultimately, this project proves that using this approach is a relatively computationally inexpensive alternative to solely using CFD simulations, making design optimization possible.

Indexing (document details)
Advisor: Stephens, Brent
Commitee: Heidarinejad, Mohammad, Stephens, Brent
School: Illinois Institute of Technology
Department: Civil, Architectural, and Environmental Engineering
School Location: United States -- Illinois
Source: MAI 58/02M(E), Masters Abstracts International
Subjects: Architectural engineering, Civil engineering
Keywords: Architectural design, Building science, Computational design, Computational fluid dynamics, Natural ventilation, Neural networks
Publication Number: 10843192
ISBN: 9780438460973
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy