Drop size measurement in a two-phase swirling flow using image processing techniques

Abstract

An experimental study of two-liquid swirling flows in a cylindrical hydrocyclone was conducted using flow visualization and digital image processing. The flow pattern was recorded using a video camera and the resulting flow images at cross-sectional planes were digitized in a personal computer. A technique was developed to measure drop sizes and areas using image enhancement and segmentation techniques. The image was cleaned of noise using a set of image processing algorithms. These algorithms included background noise subtraction, contrast enhancement, median and smoothing filters and gray level thresholding, and had to be applied in that order for accurate drop size measurement. The errors and the uncertainties associated with the technique have been addressed. For the range of oil and water flow rates studied, it was seen that the concentration of oil in any plane of cross section of the cyclone reaches a constant value. For the low flow rates studied, the cyclone behaved more as a mixer than a separator due to a weaker radial pressure gradient. Hence, a simple first order model was used to predict time constants for the two water flow rates used. The possibility of introducing this time constant as a suitable time scale was investigated. An attempt was also made to normalize the data using oil-water flow rate as a parameter. There was not sufficient evidence to show that such a ratio could be a flow parameter in this problem.