Channels, Fall 2020

Channels • 2 020 • Volume 5 • Number 1 Page 13 surface grooves in conjunction with flow around the ends of the cylinder, a truly 3D computational domain was used where the ends of a 150 mm long cylinder were away from the computational domain boundaries. Following a mesh convergence study, 9 million elements were found to be sufficient for the 3D simulations. The drag coefficients for the 3D simulations compared to the experimental results are in Table VI. Fig. 18 Velocity (m/s) contour of flow detachment for 2D elliptical (a) smooth cylinder (b) Model A TABLE VI 3D D RAG C OEFFICIENTS FOR C IRCULAR C YLINDERS Numerical Experimental Smooth 0.699 0.768 Model A 0.642 0.672 The cut plots shown from the 3D simulations in this study were cross sections in the middle of the cylinder axially. Fig. 19 shows cut plots of the turbulent kinetic energy for the smooth cylinder and Model A. Fig. 19 is zoomed in around the separation point of the flow on an amplified scale to highlight flow detachment features. Early groove turbulence was observed in Fig. 19b, showing a gradual transition into turbulence. The model with early induced turbulence also correlated with a decreased angle of detachment of the flow. Fig. 20 shows the velocity profile of the flow pattern and shows (as well as Fig. 19) that the flow detaches more smoothly in the grooved model than the smooth model. The smooth detachment was possibly due to the “roller - bearing” effect of turbulence proposed in [10]. The decreased angle of boundary layer separation in Model A compared to the smooth case resulted, as expected, in a smaller wake region and decreased drag. Fig. 21 shows the velocity wake regions for both cases. Although the two profiles are similar, the wake region for Model A (Fig. 21b) is smaller and the decreased wake size is a key element in the drag reduction observed.