Estimation of flotation rate constant and collision efficiency using regression and artificial neural networks

The effects of particle characteristics and hydrodynamic conditions on the flotation rate constant (k) and bubble–particle collision efficiency (E_c) of pyrite and chalcopyrite particles were investigated. Experimental results showed that k increases with increase of bubble surface area flux (S_b) and E_c. Artificial neural network (ANN) and multivariable linear regression procedures were used to predict both k and E_c based on the particle characteristics and hydrodynamic conditions. Multivariable linear regression resulted in R² of 0.6 and 0.93 for k and E_c, respectively. Using an ANN model, R² as high as 0.98 was achieved in modeling the E_c with regard to the available parameters. The proposed ANN model can be reliably used to determine both k and E_c parameters in froth flotation.