Abstract
Computational fluid dynamics (CFD) is a useful tool in building indoor environment study. However, the notorious computational effort of CFD is a significant drawback that restricts its applications in many areas and stages. Factors such as grid resolution and turbulence modeling are the main reasons that lead to large computing cost of this method. This study investigates the feasibility of utilizing inherent numerical viscosity induced by coarse CFD grid, coupled with simplest turbulence model, to greatly reduce the computational cost while maintaining reasonable modeling accuracy of CFD. Numerical viscosity introduced from space discretization in a carefully specified coarse grid resolution may have similar magnitude as turbulence viscosity for typical indoor airflows. This presents potentials of substituting sophisticated turbulence models with inherent numerical viscosity models from coarse grid CFD that are often used in fast CFD analysis. Case studies were conducted to validate the analytical findings, by comparing the coarse grid CFD predictions with the grid-independent CFD solutions as well as experimental data obtained from literature. The study shows that a uniform coarse grid can be applied, along with a constant turbulence viscosity model, to reasonably predict general airflow patterns in typical indoor environments. Although such predictions may not be as precise as fine-grid CFDs with well validated complex turbulence models, the accuracy is acceptable for indoor environment study, especially at an early stage of a project. The computing speed is about 100 times faster than a fine-grid CFD, which makes it possible to simulate a complicated 3-dimensional building in real-time (or near real-time) with personal computer.
