Data-driven modelling of daylight redirecting fenestration at variable directional resolution

Daylight Redirecting Fenestration (DRF) aims at the optimal utilisation of daylight in buildings striving for high visual comfort standards. Daylight simulation allows to assess whether this objective is met in architectural context, and guides decisions in building design as well as the development of DRF. The daylight simulation suite Radiance allows to employ data-driven models of variable resolution to accurately replicate the irregular light scattering by DRF. In this context, this research provides methods to improve DRFs’ integration in daylight assessments. The thesis consolidates a series of publications that address particular problems in the generation and application of data-driven models, with a focus on accurate image synthesis for visual comfort assessments. First, the parametrisation of model generation from gonio-photometric measurements is tested. Second, a novel extension of the instrumentation allows to characterise and subsequently model retro-reflection by an innovative coating. Applied in DRF, the coating controls solar gains and glare, while maintaining a view to the outside. Third, to assemble accurate data-driven models of fenestration layers into descriptions of the entire DRF, an approach employing matrix calculations is adapted and tested. Finally, the Photon Map implementation in Radiance is modified for efficient image synthesis with data-driven models, and employed in a simplified but accurate approach to Climate-Based Daylight Modelling that demonstrates the potential of retro-reflection to efficiently control glare and maintain view with static DRF. The research contributes to the applicability of data-driven models, and confirms the potential of DRF to reconcile diverging daylight performance targets such as glare control and view.