FALSE DISPLACEMENT PEAK BASED IMPROVEMENT OF PARTICLE IMAGE VELOCIMETRY TO DETECT FLOW FIELDS EXHIBITING LARGE VELOCITY GRADIENTS

摘要

An extension of the processing of image cross correlation based PIV is proposed in order to reduce the uncertainty of detecting local volume averaged mean velocity at relative image displacement variation higher than 3% and to overcome the problem of finding the true correlation peak centroid of velocity gradient spread and splintered correlation peaks. The method is based on a characteristic feature of false displacement correlation peak distribution, namely that paired particle images create false displacement peaks whose centroid coincides with that of the true correlation peaks. This symmetry is broken at strong velocity gradients depending on the local displacement variation of particle images. However, even in this case false displacement peak centroid still contains the accurate information of volume averaged mean velocity. Analytic analysis and Monte-Carlo synthetic PIV simulation, based on the image transmissivity function of a distortion free, nearly paraxial is applied to evaluate the behavior of false displacement peak distribution on correlation plane. The main influential parameters, problems and their solutions related to the identification of false displacement peaks, formed by paired particle images, under real image recording conditions were identified. It has been found that non-paired images due to in-plane velocity biases the false displacement correlation distribution towards zero hence contributing to the so called negative velocity bias in outliers contaminated spatial (or temporal) mean velocity. On the other hand, it is shown that non-uniform illuminating laser sheet intensity creates strong uncertainty on the location of false displacement peak centroid which can not be neglected in utilizing false correlation peaks. The superiority of false displacement correlation peak extended PIV processing technique to that of the standard tallest peak fining method at high velocity gradients, after eliminating the effect of non-paired images, is demonstrated by Monte-Carlo PIV simulation (Fig. A-1b.). Finally, a new processing algorithm which searches false displacement peak pairs around the estimated true correlation peak location and incorporates solutions to the above mentioned problems is proposed and demonstrated on a correlation plane from single exposed frames of real flow field image.