چكيده به لاتين
Studying the dispersion of pollutants can help to better understand their behavior and provide more effective solutions to eliminate or decrease their concentration. It is imperative to examine the performance of natural ventilation as the simplest and most economical way to remove or decrease pollutants from the indoor, as well as to realize the behavior of pollutants released from external sources into the buildings. Although the foremost dangerous type of pollutants for human health are suspended particles with a diameter of less than 10 microns, previous research has focused more on gaseous pollutants. The complexities of natural ventilation flow including transient changes, the interaction between the inside and outside flow, the presence of small-scale vortices inside, the presence of large-scale vortices outside, as well as the phenomenon of separation and reattaching of the flow to the surface, are the reasons for choosing large eddy simulation (LES) as a turbulent solver method. The main challenge in this method is the computational cost of simulating natural ventilation geometry, which requires parallel and fast processing. As of late, the Lattice Boltzmann method (LBM) has gained interest as an explicit approach with parallel processing capability to simulate complex flows, including multiphase flows. So, in this thesis, a parallel algorithm has been developed using the combined LBM-LES method in the OpenLB software environment to predict and simulate the behavior of natural ventilation flow with airborne particles. For efficient parallel processing, domain decomposition for both flow and particle phases, equal distribution of particles on processing cores, and a single-step collision step have been applied in LBM-LES. Additionally, SSLM and WALE subgrid models have been utilized in the internal flow and natural ventilation flow, respectively, and the Eulerian-Lagrangian tracking approach has been used to track pollutant particles. The intended model is initially used to simulate airflow containing suspended particles inside a model building, and its output is then compared quantitatively and qualitatively with the experimental and numerical results obtained using the finite volume method (FVM-LES). The results showed that the LBM had an appropiate level of accuracy. Despite requiring almost twice the number of computational grid cells, the LBM method had an error percentage of is 13% for the flow field results, while the FVM had an error percentage of 21%. For the particle concentration results, the error percentage is 23% and 25% for LBM and FVM, respectively. Furthermore, when compared to the numerical and experimental results, the LBM method accuracy predicted the main vortex, flow lines, contour, velocity vectors and particle concentration contour. Additionally, using a single-stage collision step in the LBM method resulted in a 17% reduction in memory consumption. The use of the integrated model in the internal flow showed that despite the need for higher grid accuracy, the LBM calculation time is significantly reduced so that the processing speed can increase up to four times in the flow phase and eight times in the particle simulation. In the next step, a model building placed in the flow of natural ventilation has been simulated and the velocity profiles on vertical and horizontal lines inside the building, velocity vectors and contours along with flow lines have been compared with experimental and numerical results. As expected, LBM-LES revealed the presence of coherent structures of horseshoe vortex and standing vortex in front of the building, roof edge vortex, wake region behind the building, and hairpin vortices along the wake behind the building. The results of the concentration field showed that while the concentration of the gaseous pollutant in the areas far from the pollutant source is 0.02-0.05, the concentration of suspended particles is at a higher level and in the range of 0.1-0.2. Also, the terminal distance of the particles in the inlet jet is equal to 531-797 micrometers, which has led to the deviation of the particles from the path of the flow lines and as a result, the accumulation of particles on the top and bottom of the walls and corner areas. While FVM-LES results obtained from previous studies provide an average of 32% for the average error percentage on the vertical center line of the building, this value is 34% for LBM-LES with 2.7x grid resolution. Finally, despite the need for higher mesh accuracy, the LBM-LES model increases the simulation speed of the flow field by almost 3 times.