🙏🏻
Building design

Synthetic Machine Learning: Real-time Spatial Daylight Autonomy Prediction

We present a robust and real-time system to assess the daylighting of architectural floor-plan designs.

Abstract

Ensuring optimal “daylighting” — the controlled admission of natural light, direct sunlight, and diffuse skylight — is a core principle of sustainable building design. By reducing the requirement for electric lighting, daylighting reduces energy consumption and ultimately will reduce the carbon impact of a building. Furthermore, spaces with excellent daylighting are associated with improved mood, enhanced morale, less fatigue, reduced eye strain, and several other occupant benefits (Edwards and Torcellini).

Presently, building designers must use advanced simulation software to assess the daylighting of a particular design. This includes using the existing libraries available in tools, like Rhinoceros-Grasshopper's Ladybug plugin and manual setup of the correct simulation model. The setup is not straightforward and includes a tedious creation of geometry for simulated building and selecting proper parameters. When the setup is ready, the simulation can be performed, which might take several minutes based on the input parameters and the complexity of geometry.

In this article, we present a new approach to assessing building efficiency in the use of daylight to illuminate its interior. Using a data-driven approach based on ML to predict daylighting for building layouts, we have created a method offering real-time user feedback on the massing solar performance. Our system allows determining the best fenestration and building layout to maximize daylighting for a given project. Compared to other simulation-based methods, our model significantly speeds up the design process while being sufficiently accurate.

Download the whitepaper