Investigation Of Burst Structures In A Turbulent Channel Flow Simulation Using Conditional Sampling Techniques.

Abstract

A turbulent channel flow was simulated at a Reynolds number of 5000 based on the initial center-line velocity. The pseudo-spectral technique was used to discretize the spacial domain. The computer program was written by S. Orszag and A. Patera and the simulation was performed on a Cray-1. The flow was developed from a laminar profile with two- and three-dimensional Orr-Sommerfeld disturbances. A steady state was never completely achieved. The evolution of the mean streamwise velocity profile was followed. When the wall region, the buffer region, and the logarithmic layer were developed, data from the simulation was saved. Planar averages of the mean and fluctuating velocities were calculated and compared with published results. The Variable Interval Time Average (Vita) and the second quadrant Reynolds stress conditional sampling schemes were modified to search the data sets for burst events in planes parallel to the wall of the channel at a distance of 11.7 viscous lengths (1*). The detected events were ensemble averaged. The averages for the three velocity components, the wall pressure and the streamwise velocity near the wall were calculated. Models were developed for events detected by the three schemes. The negative slope Vita technique detects the uw)4-uw)2 sequence in time caused by a sweep (u > 0, w < 0) interacting with a pair of lifting counter-rotating vortices. A weak sweep strikes the wall upstream of the detection site and the pair of vortices lift from the wall. The uw)2 technique detects the pair of vortices at a later stage of life; after they have increased the angle of inclination and passed through the plane z = 11.7 1*. The positive slope Vita scheme detects a uw)2-uw)4 sequence in time, caused by a sweep, forcing the pair of vortices back to the wall. The three models are presented individually and then related into a single model. The ensemble averaged wall pressure and the streamwise velocity at the wall support the model. Examination of the detection locations show that about 10% of the uw)2 detections have either a positive or negative slope Vita event associated with them. There was no attempt to determine the best set of detection parameters, T(,a) and k, which gives an optimal matching of the events.