Utilizing froth phase behaviour and machine vision to indicate flotation performance

Flotation is one of the most well-known mineral separation methods. In flotation process hydrophobicity of the solids is manipulated in order to separate the valuable minerals from the gangue. It is highly complex process because many simultaneous sub-processes and interactions occur within the system. It is essential to have a good understanding and representation of the flotation phenomena in order to design control strategies. Modern physical froth stability measures have an intrusive nature and therefore they are not practical to provide a continuous online froth stability measurement. Unlike these measures, machine vision is able to measure the key aspects of the froth owing to its non-intrusive nature. Several physical, statistical and dynamical features of froth surface are possible to measure with the machine vision techniques. The objective of this work was to understand the froth phase behavior better which indicates the flotation performance. Literature review of flotation and measurements used in flotation were performed. Image analysis methods were listed and, regarding to the importance of the froth stability, dynamical features of the froth image analysis were investigated more closely. A primary batch flotation test-work was carried out in University of Oulu. The main batch flotation experiments were executed in University of Cape Town with the wide range of operating conditions. Video captures were analyzed with statistical methods and dependencies between FrothSense™ data and concentration data were discovered. Furthermore, PLS model was formed from FrothSense™ data and process measurements in order to predict water recovery, copper grade and copper recovery. Online measurements obtained from FrothSense™ with wide range of operating conditions can be used for soft sensors. Soft sensors can estimate the stability of the froth with the robust predictions