Comparing Fast Numerical Methods and Conventional CFD for Mixing Ventilation Flows

Computational fluid dynamics (CFD) provides detailed information on the flow inside a room and can thus be used for detailed analyses of the influence of design variables, such as the placement of ventilation openings. As a result, CFD is well-suited to the optimization of ventilation at room level. However, the high computational cost and level of expertise required for implementation constitute bottlenecks in the engineering sector. In this research, a number of fast numerical techniques such as voxel-based CFD (vCFD) and coarse grid CFD (cCFD) are implemented as alternatives to conventional CFD. The methods are used to predict the velocity field for a generic mixing ventilation case and their accuracy and speed are compared. Reynolds-averaged Navier-Stokes simulations using RNG 𝑘−ϵ closure are performed, using ANSYS Fluent, in series and in parallel. The vCFD simulation is executed using the commercially available software ANSYS Discovery 2021 R2, which utilizes a proprietary algorithm that runs on GPU. The results show that, for the isothermal case, the fast numerical methods (FNMs) are two orders of magnitude faster than CFD. The accuracy of vCFD, cCFD and CFD is very similar: All cases yield RMSE and FAC1.3 values in a similar range. Current results show that vCFD and cCFD offer accelerated performance when compared to CFD, while maintaining similar accuracy. FNMs offer a distinct advantage over engineering tools, in the form of spatial information, which decreases the uncertainty of local comfort calculations prescribed by building standards. Ongoing thermal simulations for, among other things, a displacement ventilation case, using CFD, vCFD and cCFD are expected to offer additional insight into the feasibility of FNMs in the context of ventilation design optimization.