Turbulent flow data preprocessing

Methods for extracting randomly delayed signals from noise are investigated. The technique involves estimating the delay for each realization of an ensemble of signals, shifting each realization by the negative of the estimated delay and forming the ensemble average. Three methods for estimating the time delay are examined. The well known method based on correlation techniques gives the best results but at a large cost in computational effort as compared to the others. The other methods are based on determining the "center" of the energy of the signal, the center being defined in terms of the mean or median of the time distribution of the energy. The three methods are applied to data from a wind tunnel experiment on the propagation of a turbulent spot, the transit velocity of which depends on the state of the fluid medium at the instant of creation of the spot. The experiment was performed in the Department of Mechanical Engineering at the University of Houston.