Building airflow simulations with fast fluid dynamics
Fast airflow simulation is a demanding technique in building industry for improving the building design and operation, especially in such applications as emergency management, fast design of sustainable buildings, and integrated tool of building energy and airflow simulation. The review of airflow simulation models finds that current models are not sufficient for performing both fast and informative airflow simulations. For instance, multi-zone models are fast but cannot provide detailed airflow information; computational fluid dynamics (CFD) models on the other hand can offer accurate airflow information but is too time-consuming; and those intermediate models such as zonal models, coarse-grid CFD models and coupled CFD and multi-zone models have their limitations on some applications. But the fast fluid dynamics (FFD) model can be potential for fast airflow simulations because it is unconditionally stable and can provide detailed airflow simulations with a speed much faster than CFD. Therefore, this study is to develop FFD for performing fast building airflow simulations. Previous studies only validated FFD for two-dimensional airflow that has building airflow features, but building airflow is complex and always three-dimensional. Thus this study developed FFD into three-dimensional and validated its performance for simulating airflow in buildings. The validation results showed that FFD could successfully capture three dimensionality of the airflow and provide reliable and reasonably accurate simulations for airflow in buildings with a much faster speed than current CFD models. To enhance the conservativeness of FFD model, this study proposed a conservative semi-Lagrangian scheme to solve the advection equation in FFD model. The numerical tests proved that the proposed scheme was indeed conservative with negligible impact on the accuracy of the airflow prediction. FFD with the conservative semi-Lagrangian scheme can effectively enforce the mass, energy and species conservation for building airflow simulations. To demonstrate the capability of FFD for simulating more complex airflow in real applications in buildings, this study further applied FFD to simulate airflow in and around buildings with natural ventilation. The performance of FFD was evaluated for simulating different types of natural ventilation. The results showed that FFD was capable of predicting main airflow features and the ventilation rate with reasonable accuracy for wind-driven or buoyancy-driven natural ventilation. FFD simulation can reflect the influence of wind direction and surrounding buildings on natural ventilation. Since computing speed is a major concern of FFD, this study further applied two approaches to speed up FFD simulations. By reducing the grid number, coarse-grid FFD can significantly reduce simulation time, but it may also encounter issues of inaccurately predicting buoyancy-driven airflow when using a very large mesh cell to represent a heat source that could have a much smaller physical size than the cell. This study integrated a thermal plume model into the coarse-grid FFD to address this problem. Through the tests it was found that the coarse-grid FFD with the plume model could predict correctly the mean air temperature stratification in the rooms with displacement ventilation and calculated reasonably accurate ventilation rate for buoyancy-driven natural ventilation. Compared with fine-grid FFD, the coarse-grid FFD with the plume model used only a small fraction of computing time while the errors associated were comparable. For FFD on the fine grid, the study proposed a coarse-grid projection (CGP) scheme to reduce the computing time for the pressure equation, which consumes the major part of total computing effort of FFD. Through the tests, it was found that CGP scheme would not cause any negative impact on the accuracy of FFD for simulating building airflow while it could significantly reduce the fluctuations occurred within the simulation. CGP scheme could accelerate FFD by around 1.5 times, so FFD with CGP scheme could achieve a computing speed of 30 to 50 times faster than the CFD models. In the future work, FFD will be further improved by implementing simplified but robust turbulence models and by integrating models to better represent boundary conditions. The structure of FFD will be modified to support high performance computing techniques and advanced mesh systems. To demonstrate the performance of FFD in real applications, it is necessary to further develop and validate FFD for various applications.
Chen, Purdue University.
Civil engineering|Mechanical engineering|Environmental engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our