PIV Measurements of turbulence statistics and near-wall structure of fully developed pipe flow at high Reynolds number

Fully developed turbulent pipe flow at Re=11300 as well as Re=5300 is measured by PIV using water as working fluid. Excellent agreement with DNS (Direct Numerical Simulation) by Satake (2000) is achieved by choosing optimum measurement parameters such as spatial resolution, time separation of two images, and number of samples. The test section is 6.7m long stainless steel pipe with transparent PIV measurement section attached seamlessly at the end. The measurement section consists of transparent acrylic circular pipe of 1mm thick and water jacket with square cross section. The water jacket reduces the image distortion due to the pipe curvature in almost entire area of the pipe cross section to negligible level except for the region very close to the pipe wall. For turbulent flow measurement using PIV, the requirement for spatial resolution is very severe in order to resolve the fluctuating velocity field especially at high Reynolds number. Therefore, higher magnification is chosen at the cost of the size of measurement area to assure enough spatial resolution. Consequently, the size of the measurement area is half of the pipe radius and measurement is performed at two regions which are called near-wall region and pipe center. For image analysis, the iterative multi-step cross-correlation algorithm with spatio-temporal gradient method for subpixel refinement is used. Detail of the algorithm can be found in Scarano (1999) and Sugii (2000). The smallest interrogation window is 8x16 pixels which corresponds to 249 x 585μm. The near-wall structure is observed from vector map of fluctuating in-plane velocity components and out-of-plane component of vorticity fluctuation. The shear layer structure is clearly shown by vorticity fluctuation and ejection and sweep motion is also observed. The mean velocity profile and statistics are calculated from 4000 samples of vector map by first taking temporal statistics and then spatial average along streamwise direction and compared with DNS in non-dimensional form using friction velocity. Since accurate measurement of friction velocity is essential for the comparison with DNS, using pressure gradient is not appropriate due to the relatively small pressure drop along the pipe. Therefore, curve-fitting method using near-wall turbulent model developed by McEligot (1984) is applied to calculate friction velocity from the velocity profile in buffer region.