IFACMMM 2009. Viña del Mar, Chile, 14 -16 October 2009. Bubble Size Analysis and Boundary Conditions for Automatic Control of Industrial Flotation Cells L. E. Vinnett, F. A. Contreras and J. B. Yianatos Automation and Supervision Centre for Mining Industry, CASIM Santa María University, Valparaíso, Chile. (e-mail: juan.yianatos@usm.cl). Abstract: The gas dispersion has a high impact on flotation processes performance. To evaluate this effect, the bubble surface area flux (S b ) must be evaluated from superficial gas rate (Jg) and Sauter mean bubble diameter (D 32 ) local measurements at industrial scale. Gas rate sensors and visual bubble caption techniques (video, photograph) have been developed successfully, but the commercial software for automated image analysis still has troubles when irregular big size bubbles and bubble clusters appears in industrial measurements, which prevents an accurate and more reliable bubble size estimation. Also, a criteria for bubble size distribution (BSD) diagnostic and control at industrial scale, has not been developed yet. In this work, measurements of BSD were carried out in large flotation cells, of different size, operating in several flotation circuits in Chile. Also, new semi-automatic software for image analysis was developed to overcome the troubles before mentioned and permitting more reliable bubble size estimations. Experimental results showed that under normal operating conditions BSD matched well with a lognormal distribution. From this reference two types of operating problems has been identified and the corresponding diagnosis criteria were established, which is a powerful tool for automatic control purposes. Keywords: Bubble Size Estimation, Image Analysis, Flotation Process, Lognormal, Bubble Clusters. 1. INTRODUCTION A key variable in two and three phase reactors performance is the dispersion of the non-continuous phase to maximize the mass transfer and/or the kinetic reaction rate, increasing the concentration and specific area of disperse phase (bubbles, drops). The flotation process can be considered a reaction plus mass transfer process (Finch et al., 1995, Laplante et al. 1989), where the gas (non-continuous phase) is dispersed into small bubbles in the pulp volume (continuous phase), which contain hydrophobic particles that are captured by bubbles forming the bubble-particle aggregates, which after reaching the top of cell are recovered as the concentrate product. To evaluate the gas dispersion in large flotation cells, the bubble surface area flux (S b ) must be estimated according (1) (Finch and Dobby, 1990) where Jg is the superficial gas rate and D 32 is Sauter mean bubble diameter. All these variables can now be measured in plant operations. 32 6 g b J S D   (1) The bubble diameter D population is commonly represented by the Sauter mean diameter D 32 (Barigou and Graves, 1992, Gorain et al, 1997). D 32 is defined by (2), and represents the diameter of a bubble population, of constant bubble size, having the same total volume and total surface as the actual bubble distribution. 3 2 32 1 1 n n i i i i D D D    ¦ ¦ (2) The significant role of bubble size on flotation efficiency has been discussed in various works (Dobby and Finch, 1986; Gorain et al., 1997; Heiskanen, 2000), where a strong relationship between bubble diameter D 32 and S b with the flotation rate constant k was shown. 1.1 Bubble size distribution The bubble size distribution (BSD) depends of the bubble generation mechanism. In mechanical flotation cells, where specific power is 0.6–1.3KW/m 3 , the bubble generation is carry out in the rotor-stator zone, where due to the external inertial stress from turbulent pressure fluctuations, bubble (gas) breakage is the prevailing bubble formation mechanism. Under this condition, the resulting BSD can be represented by a lognormal type distribution. Several tests in pilot and industrial equipments for bubble and droplet distributions generated under breakage conditions confirm the good fitting of the experimental data with a lognormal distribution (Ruiz et al. 2002; Grau and Heiskanen 2005).