| In this work, the lattice Boltzmann method (LBM) was considered as an alternative tool for direct numerical simulation (DNS) of turbulent flows in complex geometries. The LBM is a promising numerical method for this application because it has a computational efficiency (per cell) that is comparable to that of spectral methods, and complex geometries can be implemented.; In order for the LBM to be used for DNS in complex geometries, its ability to accurately simulate wall bounded turbulent flows must be verified. To do this, an LBM code was developed and verified by solving a number of laminar wall bounded flows and comparing the results to their analytical solutions. Then, a simulation of fully-developed turbulent channel flow was performed, and the results were compared to the DNS database of Kim, Moser, and Moin [28].; Due to limitations in computational resources, the minimal channel was simulated instead of the full channel. The minimal channel represents the smallest channel (in the homogeneous directions) for which a turbulent flow can be sustained [27]. This channel accurately simulates the near-wall turbulence, but the largest scales of the flow are not properly represented due to the periodicity enforced by the boundaries. Thus, the minimal channel is a computationally efficient vehicle for verifying the accuracy of LBM for near-wall turbulence.; The grid resolution for the channel was chosen to be equal to the Kolmogorov length scale. Unfortunately, the truncation error (due to the finite difference approximation of derivatives in the LBM) reduced the maximum wavenumber that was accurately resolved. Consequently, the results show features that are characteristic of an under-resolved channel flow simulation. For this reason, the LBM could not be completely verified for this case. |
