Artificial Intelligence in Performance-Driven Design

4An Integrative Deep Performance Framework for Daylight Prediction in Early Design Ideation

Shermeen Yousif and Daniel Bolojan

School of Architecture, Florida Atlantic University, Fort Lauderdale, FL, USA

Introduction

By 2050, the global population will expand by 2.1 billion, posing new challenges for housing and infrastructure provision and requiring us to rethink current design and construction methods (UN‐Environment‐Programme 2022). It is also anticipated that the equivalent of New York City will need to be built every 34 days for the next 40 years (Architecture‐2030 2018) to accommodate such population growth. Intertwined with overpopulation are climate change challenges and increasing resource consumption. In the US, buildings consume approximately 40% of our total energy and contribute approximately 40% of the annual Greenhouse Gas emissions (DoE 2015). Effective energy performance evaluation and optimization are required to achieve building design resilience and adaptation to growing environmental challenges (Attia and De Herde 2011). To achieve energy efficiency across all building designs, real‐time performance feedback needs to be integrated into performance‐driven frameworks and be made accessible to a broad audience of building designers and engineers.

In recent years, there has been a shift toward developing real‐time and predictive methods for building performance, moving away from traditional simulation methods. Machine learning (ML) algorithms have proven to ...