LARGE EDDY SIMULATION OF TURBULENT ATMOSPHERIC BOUNDARY LAYER THROUGH RECYCLING-RESCALING-REVISING METHOD

The recycling and rescaling method was firstly proposed by Lund for generating three-dimensional, time-dependent turbulent inflow data for simulations of complex spatially developing flat-plate boundary layers. The heart of Lund's method is estimating the mean and fluctuating velocity at the inlet plane by recycling and rescaling the solution of downstream. The accuracy and efficiency of this method was validated by a relatively low Reynolds number flow (Re-delta approximate to 20000). In some extent, the atmospheric boundary layer of earth could be seen as a flat-plate boundary layer in planetary scale. Hence, the methodology of recycling-rescaling is adopted in this paper and extended to a larger Reynolds number boundary layer. The height of computational domain is 3.6m which is same as the wind tunnel of Shenzhen wind environmental engineering laboratory. For a U-infinity =10m/s incoming flow, the boundary layer thickness Reynolds number Res is 820000, which is approximately 100 times larger than Lund's simulated flow. Firstly, the effects of several key parameters in recycling-resealing procedure are compared and discussed, such as the initializing condition, the location of recycle plane, the reintroduced pattern of fluctuating velocity, the rescaling coefficients for both mean and fluctuating velocities and the boundary condition for top surface. A most suitable configurations are concluded for the subsequent numerical simulation. Furthermore, considering the mesh and time step size used in this paper, the solution is also proved to be independent. Secondly, in order to realize the prescribed profiles of turbulence intensity defined in China loading code (GB 50009-2012), the fluctuating velocities extracted at recycle plane is not only resealed based on the development of boundary layer thickness, but also revised according to the objective turbulence intensity. So, the rescaling coefficient y, originally defined as the friction velocity ratio of inlet plane to recycle plane by Lund, is modified to a function with a variable of height z. To achieve the aimed turbulent flow, the value of y is calculated using the Proportion-Integration-Differentiation (PID) control strategy. To stabilize the numerical iteration and time stepping, only the proportional parameters of PID control strategy are selected in this study. Finally, the characteristics of large eddy simulation wind flow corresponding to the two exposure area: A and B are analyzed and compared with the empirical formula, which include the mean wind speed profile, turbulence intensity profile, wind spectrum and turbulence integral length.