Machine Learning Algorithm-Based Tool and Digital Framework for Substituting Daylight Simulations in Early-Stage Architectural Design Evaluation

Abstract

The aim of this paper is to examine the new method of obtaining the simulation-based results using backpropagation of errors artificial neural networks. The primary motivation to conduct the research was to determine an alternative, more efficient and less time-consuming method which would serve to achieve the results of daylight simulations. Three daylight metrics: Daylight Factor, Daylight Autonomy and Daylight Glare Probability have been used to verify the reliability of applying the latter into an early stage architectural design process. The framework was based on the computationally generated data sets build on various office models variants followed by daylight simulations. In order to predict the simulations values based on the given office parameters with artificial neural networks algorithm, a specific tool was designed as an alternative to computer simulations. The designed tool and simulations results were compared against computing time and values differences. The above findings of the research proof the reliability of the new methods as a tool during an early-stage architectural design process likewise conventional daylight simulations.