Temporally and spatially resolved simulations of turbulent flow need realistic inlet velocity fluctuations. The vortex method, described and implemented in this work, adds vortices to the inlet mean flow profile. The vortices are randomly distributed over the inlet and they move across the extent of the inlet with a finite life-time. The initial vortex sizes, strengths, as well as their developing motion and life-times, are determined by the inlet mean velocity and turbulence distributions. In contrast to previous studies of the vortex method, the inlet mean velocity and turbulence distributions are in the present work theoretically determined, and thus form an explicit part of the boundary condition. This makes the method independent of precursor RANS simulations. Special care has been taken to distribute the vortices evenly across the inlet. The tangential velocity of the vortices is randomly initialized and the spatial and temporal correlation is preserved. An edge bouncing mechanism is implemented at solid boundaries.

The boundary condition is applied to a DNS simulation of pipe flow at a Reynolds number of 5300, based on pipe diameter, aiming at reaching fully developed conditions at a short distance from the inlet. The results are compared with those using cyclic boundary conditions, and with two DNS results found in the literature. It is shown that a good agreement is reached for the mean velocity profiles five pipe diameters downstream from the inlet. Also the resolved velocity fluctuations at that location are reasonable. It is concluded that the present implementation of the vortex method boundary condition, based on theoretical mean velocity and turbulence profiles, gives an inlet flow that is sufficiently realistic to reach a well-resolved fully developed turbulent pipe flow at five pipe diameters downstream the inlet.