Generative Machine Learning Models for Airflow Prediction of Architectural Spaces

Search for this publication on Google Scholar
Claudio Rodriguez, F. (2021). Generative Machine Learning Models for Airflow Prediction of Architectural Spaces. Unc Charlotte Electronic Theses And Dissertations.
Download PDF

Analytics

82 views ◎
60 downloads ⇓


Abstract

Computer Fluid Dynamics (CFD) airflow simulations are not used as often in architectural settings primarily due to time constraints. The proper use of CFD airflow simulations involves a complex setup and run-time process that requires a large amount of expertise on the different stages. This study aims to apply existing generative machine learning algorithms to compute CFD wind velocity simulations to significantly shorter run times while maintaining a relatively high accuracy level. In order to test the proposed hypothesis that machine learning can be used as a method to produce rapid and acceptable results for airflow CFD simulations in the early design stages, multiple machine learning models were created, trained, and tested. The evaluation metrics consisted of using different image similarity methods to evaluate the accuracy of the images produced by the machine learning model compared to their CFD engine counterparts. In addition, run times between the CFD engine and the trained machine learning model were recorded and compared. These results obtained indicate that GAN application for CFD airflow predictions can produce acceptable results showing a significant run time difference of over a minute between the CFD simulation and the machine learning model.

Details
Author

Claudio Rodriguez, Fernando

Title

Generative Machine Learning Models for Airflow Prediction of Architectural Spaces

Physical Description

1 online resource (48 pages) : PDF

Date

2021

Degree Granting Institution

University of North Carolina at Charlotte

Abstract

Computer Fluid Dynamics (CFD) airflow simulations are not used as often in architectural settings primarily due to time constraints. The proper use of CFD airflow simulations involves a complex setup and run-time process that requires a large amount of expertise on the different stages. This study aims to apply existing generative machine learning algorithms to compute CFD wind velocity simulations to significantly shorter run times while maintaining a relatively high accuracy level. In order to test the proposed hypothesis that machine learning can be used as a method to produce rapid and acceptable results for airflow CFD simulations in the early design stages, multiple machine learning models were created, trained, and tested. The evaluation metrics consisted of using different image similarity methods to evaluate the accuracy of the images produced by the machine learning model compared to their CFD engine counterparts. In addition, run times between the CFD engine and the trained machine learning model were recorded and compared. These results obtained indicate that GAN application for CFD airflow predictions can produce acceptable results showing a significant run time difference of over a minute between the CFD simulation and the machine learning model.

Genre

masters theses

Subjects--Topics

Computer science
Information technology
Architecture

Degree

M.Arch.

Keywords

Airflow Simulations
Architectural Spaces
Computer Fluid Dynamics
Cross Ventilation
Generative Adversarial Network
Machine Learning

Subject Area

Architecture

Advisor(s)

Ellinger, Jefferson

Committee Members

McCormick, Elizabeth
Wu, Chengde
Lee, Minwoo

Degree Note

Thesis (M.Arch.)--University of North Carolina at Charlotte, 2021.

Rights Statement

This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.

Rights Holder Information

Copyright is held by the author unless otherwise indicated.

Identifier

ClaudioRodriguez_uncc_0694N_12846

Permalink

http://hdl.handle.net/20.500.13093/etd:2218