Towards the automatic generation of zonal models from CFD simulations

This doctoral thesis presents a novel computational tool chain able to automatically extract zonal models from CFD simulations of the built environment in order to provide accurate and rapid predictions of indoor temperature distributions. This doctoral thesis introduces a method for automatically clustering the computational cells of CFD simulations into sub-zones that can be considered quasi isothermal and compute the mass and heat exchange between the sub-zones and the domain to extract a zonal model. The zonal model can then be solved for off-design boundary conditions at a fraction of the computational cost of CFD. This doctoral research first developed three methods for clustering CFD cells and compared their ability to capture the temperature distributions of the computational domain. These methods are: (1) the Mean Values Segmentation, (2) the Classic Watershed, and (3) the Coarse Grid method. The research presented in this thesis shows that the Mean Values Segmentation method performs best with regards to computational power and pertinence of the extracted clusters. Then, this doctoral thesis developed a method for generating zonal models from cell clusters and information extracted from the CFD simulation results such as boundary conditions. The method was applied to three case studies: an office space, a naturally ventilated meeting room, and an ideal ventilation case. This study has assessed the fidelity of the automatically generated zonal models when solved for off-design boundary conditions and has showed that the zonal models are able to predict temperature distributions with a weighted mean absolute error under 0.6 K when the temperature and mass flow rate at the boundaries are changed. The proposed methodology is able to capture and simulate local phenomena such as air stratification and thermal plumes with minimal error and is able to accommodate complex flow fields provided that the domain’s temperature is not uniform.

Funder

College of Engineering and Informatics, National University of Ireland, Galway
Horizon 2020

Publisher

NUI Galway

Rights

Attribution-NonCommercial-NoDerivs 3.0 Ireland

cbn

Collections

University of Galway Theses (PhD Theses)