Font Size: 

Natural convection in a horizontal concentric annulus has raised interests of many researchers due to its wide applications in heat exchangers, electronic devices, and aerospace systems. It presents great challenges for numerical methods to simulate natural convection with high Rayleigh numbers. As a Lagrangian particle method, smoothed particle hydrodynamics (SPH) [1] has many advantages in modeling different problems associated with moving interfaces or heat transfer. However, due to the larger computational cost, the 3D natural convection at high Rayleigh numbers have not been well simulated by SPH so far. In this work, an improved SPH model is presented to deal with this problem, where the particle shifting technique and δ-SPH model are integrated to improve the computational stability and alleviate numerical oscillations respectively [2,3]. The graphics processing unit (GPU) through multi-threading is employed to enhance the computational performance in terms of efficiency and scale. With the GPU-accelerated SPH model, some underlying mechanisms of natural convection in a horizontal concentric annulus are investigated, which is of great helpful to engineering applications.