Empirical correlations in B-C transition model were calibrated with experimental values of zero pressure gradient plates. Calibrated B-C transition model predicts reasonably transition positions with low inlet turbulence intensity. However, due to limitations of RANS method, B-C transition model’s accuracy diminishes for massively separated flows. A turbulence closure named transitional Delayed Detached-Eddy Simulation (BC-DDES) method is proposed which combines traditional SA-DDES method and B-C transition model. This method has the potential for accurately capturing massively separated boundary layers in transitional Reynolds number range. Numerical simulation of three-dimensional S-K flat plate shows that BC-DDES method obtains transition prediction consistent with baseline transition model. Comparisons are evaluated on circular cylinder in crossflow. It shows that pressure distribution and drag coefficient of cylinder calculated with BC-DDES method agree well with experimental data, with less computational costs than tHRLES method.

Spatial and time resolution of thermal scalar field and velocity fields are different in turbulence thermal convection. A multiple-resolution parallel direct method for DNS with two different girds is presented to solve the problem of huge computational cost in simulation of turbulence thermal convection with very high Rayleigh numbers. A constant translation interpolation method of speed was designed to meet continuous equations at each finer gird in data transformation between coarser gird and finer gird. Simulations of 2D turbulence thermal convection with very high Rayleigh numbers show that computational cost is reduced an order of magnitude by using this method. Small-scale eddy-like plumes, that move rapidly in thermal convection, are well performed in instantaneous temperature fields. It is consistent with results with a single gird. Difference of Nusselt number obtained by two methods is below 1%.

Temperature fields in two-dimensional (2D) Rayleigh-Bénard (RB) convection with no-slip and free-slip walls were simulated with parallel direct method of direct numerical simulation (PDM-DNS) at different Rayleigh numbers (Ra) and aspect ratio numbers (Γ). Different from no-slip boundary thermal convection with random plume, free-slip boundary thermal convection eventually forms a large-scale circulation without turbulent features and temperature distributes only on four walls. Temperature distribution characteristics of time-average field near floor changes gradually for no-slip boundary but overshoot occurs for free-slip boundary. Dependences of Nusselt numbers (Nu) with Ra at Γ=1 have same scale index Nu~Ra^0.3. Free-slip boundary thermal convection enhances heat transfer. Changes of Nu with Γ are obvious in free-slip boundary thermal convection. They are in two stages. At Γ=0.5, Nu_max≈250, it is five times of Nu in no-slip boundary thermal convection.

A parallel direct methods of DNS (PDM-DNS) for two-dimensional turbulent thermal convection was created with PDD algorithm for Poisson equation direct solving. DNS calculations of turbulent convection with high Ra were completed on "Tianhe-2" supercomputer, in which iteration time steps are over one hundred million. Effect of calculations is surprised. A calculation scale, which was difficult to achieve a year before, is realized. It shows that turbulent convection flow with different high Ra number are completely different. This work provides a valuable method for massively parallel computing and numerical simulation of turbulent convection with high Ra number.