Paper Conference

Proceedings of Building Simulation 2017: 15th Conference of IBPSA

     

Evaluation of Impact of Turbulence Model Choices in CFD for Cp-Values used in Airflow Networks

Junji Yi1, Andrea Frisque2, Kendal Bushe1
1Department of Mechanical Engineering, University of British Columbia, Vancouver, BC, Canada
2Stantec Consulting, Vancouver, BC, Canada


DOI: https://doi.org/10.26868/25222708.2017.650
Abstract: Pressure coefficient values are incorporated into airflow network (AFN) models and are used to calculate the bulk airflow into the building. Given that Cp values are so important and at the same time are so difficult to get, accurate CFD simulations can be and are used as an easy and inexpensive way to obtain Cp values compared to wind tunnel testing. However, the accuracy of these values is not well understood. The literature on the magnitude of error to expect in Cp values is very limited. This is particularly important in building design, as the typical building design professional is not an expert in fluid dynamics, turbulence modeling, numerical mathematics and high-performance computing. It is unrealistic to expect that building professionals would become experts in this area, rather it is important to provide practitioners with clear guidance on best practices. In this paper, we evaluate the impact of selection of turbulence model on Cp values. We compare the kepsilon model, which is by far the most widely used model in the building design community, to the bettersuited Spalart-Allmares model. We conduct a 3D RANS simulation for an isolated box-shaped building using the standard k-epsilon model and the Spalart-Allmaras model and then compare the results to data from wind tunnel experiments. Finally, a comparison is made between the Cp values obtained from the simulation and those used in airflow network models. The intent is to evaluate the appropriateness of the airflow network default values and demonstrate that CFD can be used as a tool in obtaining reliable Cp values, while also drawing attention to potential sources of errors.
Pages: 2315 - 2322
Paper:
BS2017_650