Nonparametric density estimation of bubble size distribution for monitoring mineral flotation process

The mineral separation efficiency of flotation process depends very much on the surface properties of feed ore and addition of chemical reagents. Machine vision based analysis of froth appearance is considered as an indication of flotation performance. Bubble structure obtained by watershed segmentation scheme is used to determine the amount of reagent. To explore bubble size distribution, nonparametric wavelet thresholding estimator is introduced to approximate the output Probability Density Function (PDF). With the aim of tracking the output PDFs to a target distribution shape, the output PDF model is therefore transformed into the dynamic weight coefficients model which allows a predicted reagent addition profile to be identified for controlling the flotation process.