(554a) Separation of Fine Oil Droplets from Oil-in-Water Mixtures By Dissolved Air Flotation

Mohamad Radzi, A. R. B. - Presenter, University of Surrey

Dissolved air flotation (DAF) is a separation technique, often used after a primary gravity separator or filters, to enhance the quality of the wastewater before it can be released to streams, rivers and the sea. The main aim of the DAF investigations reported here was to measure the removal efficiency of oil droplets in the ranges of 15-80 μm from the oil-in-water mixtures. Several operating parameters were varied and two kinds of oil were used. It was decided that vegetable oil should be tested as it is relevant to the food industries. To compare the vegetable oil droplet removal efficiencies, investigations were also carried out using a mineral oil. Saline water in this experiment was made by adding 3.5% by concentration of NaCl, to mimic the average salinity of water produced from the oil wells. An oil-in-water measuring technique (FastHEX) and a droplet counting method (Coulter Counter) were used to measure the concentration and numbers of oil droplets in the water, respectively. The best oil droplet removal efficiency obtained from these experiments was found to be 96 % and it was for the removal of vegetable oil droplets from a saline oil-in-water mixture. The vegetable oil has a large and positive spreading coefficient (14.6 x 10-3 N/m), which facilitates the attachment of bubble and oil droplet and therefore produced higher removal efficiency than lamp oil that has a negative spreading coefficient (-7.11 x 10-3 N/m). The experimental results were subjected to dimensional analysis and the removal efficiency was found to be a function of several dimension and dimensionless groups. The experimental data tabulated in dimensionless form has been subjected to multivariable linear regression. The resulting correlation was found to have a root mean square error of 11.1 % but predicted oil droplet removal efficiencies greater than one and less than zero. An alternative mathematical formulation was devised that cannot predict efficiencies outside the range. Regression of this formulation, which had the same number of adjustable parameters as the linear regression, was successful with a root mean square error of 10.9 %.