Modeling of bubble surface area flux in an industrial rougher column using artificial neural network and statistical techniques

Previous studies in mechanical and column flotation cells have shown that bubble surface area flux (Sb) is an appropriate indicator of gas dispersion in a flotation cell which has a relatively strong correlation with flotation rate constant. In the present investigation, based on extensive tests conducted in an industrial Metso Minerals CISA flotation column (4 m in diameter and 12 m in height) in a rougher circuit, Sb as a function of the most significant operating variables which affect gas dispersion in a flotation column (i.e. superficial gas velocity, slurry density (solids%) and frother dosage/type) was modeled using artificial neural network (ANN) and statistical (non-linear regression) techniques. The models were developed taking into consideration a data set consisting of 82 experimental tests conducted in an industrial rougher column (at a copper concentrator in Iran) operating under a variety of experimental conditions. aper outlines the development of the models and validation using a number of randomly selected datasets. Limitations of the present models are discussed and comments and recommendations on further investigations are given.