Simulations of Air Distributions in Buildings by FFD on GPU
Simulations of Air Distributions in Buildings by FFD on GPU
复制标题
DOI：
10.1080/10789669.2010.10390934
复制
发表时间：
2010-11
期刊：
影响因子：
--
通讯作者：
W. Zuo;Qingyan Chen
中科院分区：
文献类型：
--
作者：
W. Zuo;Qingyan Chen
建筑设计和运行通常需要实时或比实时更快的模拟，以获得空气分布的详细信息。快速流体动力学（FFD）模型通过求解能量和物质的Navier-Stokes方程和输运方程，可以作为计算流体动力学（CFD）模型提供详细的信息。与CFD相比，FFD的速度快了50倍，但精度有所降低。但通过改进模型的数值格式，可以进一步提高模型的精度和速度。此外，发现FFD程序的计算时间可以减少多达30倍的图形处理单元（GPU）而不是中央处理单元（CPU）上执行。此外，可以通过优化GPU代码和使用多个GPU来加速FFD仿真。总体而言，使用gpu上的FFD可以对具有107个网格和Δt = 0.1s的中等大小的建筑物进行实时模拟
Building design and operation often requires real-time or faster-than-real-time simulations for detailed information on air distributions. By solving the Navier-Stokes equations and transportation equations for energy and species, Fast Fluid Dynamics (FFD) model can provide detailed information as a Computational Fluid Dynamics (CFD) model. Compared to the CFD, the FFD is 50 times faster with some compromise in accuracy. But the accuracy and speed of the FFD model can be further enhanced by improving its numerical schemes. In addition, it was found that the computing time of the FFD program can be reduced up to 30 times by executing on a Graphics Processing Unit (GPU) instead of a Central Processing Unit (CPU). Furthermore, the FFD simulation can be accelerated by optimizing the GPU code and by using multiple GPUs. As a whole, it is possible to perform real-time simulation for a moderate size building with 107 grids and Δt = 0.1s using the FFD on GPUs