In this doctoral study the effect of surface roughness on the turbulent flow in a channel is investigated using Large eddy simulation (LES) and Direct numerical simulation (DNS) techniques. The roughness is in the form of elongated, transversely placed square ribs. The ribs are spaced at a uniform pitch along the stream. The fully developed turbulent flow in this channel configuration is simulated at Reynolds numbers Reb in the range of 1200 - 4800. Here, Reb refers to the Reynolds number based on the bulk velocity and half-height of the channel. Three case studies each with different rib arrangements are considered to understand the vertical and lateral effects of surface roughness. In the first case study, turbulent flow in a channel with symmetrically roughened walls is investigated using near-wall resolving LES. LES is performed with different grid resolutions and are validated with in-house DNS data. The accuracy of LES in capturing several features of turbulent flow is discussed in detail.

In the second case study, only one wall is roughened and two configurations of the channel are considered based on the spanwise extent of the ribs-- two-dimensional configuration resulting from full-span-width ribs and a three-dimensional (3D) configuration, where the ribs extend only up to half the spanwise extent. The 3D configuration thus produced a unique problem of coflowing rough and smooth turbulent channel flows. A pair of secondary roll cells has been observed, exhibiting a strong updraft in the smooth half of the channel. Comparisons were drawn with DNS of a smooth channel flow and it was found that the roll cells on the smooth half not only affect the bulk flow negatively but also attenuate the turbulence significantly. In the third case study, both the channel walls are symmetrically roughened, and similar to the second case study the ribs extend to only half the spanwise extent of the channel leaving the other half smooth. DNS is used to study this coflowing rough and smooth turbulent channel flows, where two pairs of secondary roll cells are observed occupying the entire smooth half of the channel. The results are compared with the DNS data of the first case study. In both the case studies 2 and 3, in spite of having a higher bulk velocity, the rough half of the 3D configuration has been found to be more turbulent than the 2D configuration; this has been attributed to the momentum transfer from the smooth half to the rough half.