RANS-based fast computational methods for indoor flows: a framework-driven performance assessment for a simple benchmark

Reliable indoor climate prediction requires reliable tools. Complex airflow predictions often involve the use of physical or numerical models. Computational fluid dynamics (CFD) is a popular, high-fidelity method that can produce reliable predictions. Conventional CFD methods like steady Reynolds-averaged Navier-Stokes (RANS) are time-consuming. Certain methods, called fast computational methods, may be faster and easier to use. However, a framework for addressing the speed, accuracy and software trade-off of these methods is lacking. In this paper, a unified framework is used to compare conventional, two-equation RANS (RANS-2) to four, RANS-based fast computational methods: coarse-grid modelling (c-RANS-2), zero-equation modelling (RANS-0), one-equation modelling (RANS-1), and GPU-based modelling (gpu-RANS-2). This framework is applied to a simple case, as a preliminary step towards understanding how fast computational methods may be applied in practice. This research concludes that gpu-RANS-2 and c-RANS-2 offer the best trade-off for simple mixing ventilation, and makes suggestions for simulating complex flows.