Several attempts to measure "periodicity" in the output signal of hot wires and hot films located in the viscous sublayer have been made in recent years. The usual method is to perform an autocorrelation on the signal and to interpret the strong peaks in the correlation function as indicative of the mean period of the "turbulence bursts" during a particular sampling interval.
The data analyzed for the presence of "turbulent bursts" were from the test of a 60-foot, 8-oared racing shell at the David Taylor Model Basin of the Naval Ship Research and Development Center. The shell was instrumented with flush mounted hot film sensors at several locations along the bottom and the hot film data were recorded on magnetic tape and saved for this study.
Initial attempts to uncover periodicity directly from frequency domain analysis were unsuccessful since the random components in individual spectral estimates were too great. A two-step algorithm sequence was then developed which averaged out many of the random components by autocorrelating a 2000 point sample, and then taking the auto spectrum of the resulting 1000 point correlation function. This method is considerably more revealing than "eyeballing" a 100 point lag function for "peaks." Using this technique, the entire hot film record at each speed was analyzed, and a mean bursting frequency and the corresponding standard deviation were calculated.
The results for ten different Reynolds numbers indicated that the Kline and Black models predict burst frequencies which are much higher (burst periods much lower) than any observed values except at the lowest Reynolds numbers. The Einstein-Li model, on the other hand, predicted burst frequencies less than the observed values at the lowest Reynolds numbers. In the middle range of Reynolds numbers, the Einstein-Li prediction was within the standard deviation of the mean observed value and equal to the mean value at Reynolds number of 1.7 x 107. The observed data then dropped below the Einstein-Li prediction at the highest Reynolds numbers.
These results are clouded by the second part of the investigation in which it was discovered that it is possible to closely simulate the "periodic" autocorrelation function by analyzing filtered random noise with a frequency roll-off similar to that of the time averaged turbulent spectra.
Johnson, B. and Saylor, R., "An Attempt to Characterize the ‘Turbulence Burst Phenomena’ Using Digital Time Series Analysis" (1971). Symposia on Turbulence in Liquids. 76.
Symposium on Turbulence in Liquids (1971: Oct. 4-6, Rolla, MO)
Chemical and Biochemical Engineering
Article - Conference proceedings
Turbulent Burst Signatures
© 1972 University of Missouri--Rolla, All rights reserved.