engleski

Suitability of POD for Airflow Estimation in a Room with Multiple Unidirectional Air Diffusers

Although the computational power is increasing constantly, numerical modeling of turbulent and multiscale flows, as well as multiphysics and flow separation are still computationally expensive, requiring a couple of millions of degrees of freedom, powerful CPU, and hardware architecture. In the case of design optimization, flow control, or sensitivity analysis the problem of computational power becomes even more evident. Therefore, reduced order models (ROMs) are being extensively developed in the last decade. One of the most frequently used approaches to ROM generation is proper orthogonal decomposition (POD). It is used as the optimum between accuracy and the required amount of data for ROM generation. Nevertheless, the complexity of the problem significantly affects the accuracy of the POD approach. This paper aims to determine if it is possible to reproduce multi-species flow (C02 and air mixing) with POD and to what extent the number of input parameters affects the required number of simulations for the offline part of parametric ROM generation on the example of a room with two unidirectional inlet diffusers and two exhaust bar diffusers. The results of the ROM were compared with the CFD results in terms of accuracy.

Proper orthogonal decomposition (POD), pollutant control, computational efficiency

nije evidentirano