A particle image velocimetry was used to study shallow open channel turbulent flow over d-type and k-type transverse ribs of square, circular, and semi-circular cross sections. The ratio of boundary layer thickness to depth of flow varied from 50% to 90%. The mean velocities and turbulent quantities were evaluated at the top plane of the ribs to characterize interaction between the cavities and overlying boundary layer. It was found that the overlying boundary layer interacts more strongly with k-type cavities than observed for d-type cavities. The profiles of the mean velocities and turbulent statistics were then spatially averaged over a pitch, and these profiles were used to study the effects of rib type and cross section on the flow field. The mean velocity gradients were found to be non-negligible across the boundary layer, and the implications of this observation for momentum transport, eddy viscosity, and mixing length distributions are discussed. The results show that the skin friction coefficient, Reynolds stresses and mixing length distributions are independent of rib cross section for d-type. For the k-type ribs, significant variations in skin friction coefficient values, mean flow, and turbulence fields are observed between square ribs and circular/semi-circular ribs.

Three‐phase (G/L/S) horizontal pipe flow data collected from the literature are used to evaluate the performance of a number of correlations designed to predict the pipeline pressure gradient. In the present study, a number of popular two‐phase gas–liquid pressure loss correlations were modified for three‐phase flow predictions. The primary modification is to assume that the slurry (L/S) mixture behaves as a singlephase. The modified Dukler and the Beggs and Brill correlations did not provide accurate estimates of the three‐phase pressure gradients. When the classical Lockhart–Martinelli (L–M) correlation was used, along with a kinematic friction loss model to calculate the slurry (L/S) superficial flow pressure gradient, accurate predictions of the three‐phase (G/L/S) pressure gradient were obtained provided the slurry did not exhibit non‐Newtonian behaviour and that Coulombic (sliding bed) friction was negligible. Additional experiments should be conducted before the improved version of the L–M correlation is applied to commercial installations with pipe diameters greater than 100 mm.

A numerical investigation of two-phase solid-liquid (slurry) flow in horizontal pipes has been carried out. Simulations of concentrated slurry flows in pipes 0.0515 m and 0.15 m in diameter were performed using the two-fluid approach implemented in the commercial CFD code, ANSYS CFX. Mixtures of monosized and bimodal particle sizes were tested. Several test cases were investigated to predict particle velocity- and concentration-distributions and frictional pressure gradients. The effects of turbulence model selection, dispersed phase wall boundary conditions, and interphase force terms on model performance were evaluated. The selection of turbulence model had a significant impact on the dispersed phase velocity and concentration distributions. Comparison of simulations with benchmark experimental data shows clearly that for the relatively small particle sizes (∼100 microns), poor solids concentration profile predictions are obtained if the turbulent dispersion force is not included. In general, very good agreement between numerical and experimental results was observed.

Experiments were performed in a 265 mm diameter pipe loop with sand‐in‐water slurries (d50 = 0.371 mm). In situ solids volumetric concentrations ranging from 20–40 % by volume (0.20–0.45 L/L) and mixture velocities up to 6 m/s were tested. Measurements of the instantaneous and average solids velocity and turbulent intensity profiles across the centreline diameter of the pipe were obtained using a particle velocity probe. A CFD package, ANSYS CFX 14, was used to perform numerical simulations. Solids turbulent intensities were found to be greater near the bottom of the pipe where solids concentrations are typically higher and significant particle‐pipe wall interactions occur. The agreement between the numerical results and the experimental data was found to be concentration dependent, with relatively closer agreement at lower concentrations.

No abstract