Automatic control of flotation process using computer vision

In the mineral production industry, the separation of valuable material from waste material is generally carried out using the flotation process. Metallurgical parameters of the process reflect the quality and quantity of the product. Online measurement and control of these parameters is currently not possible, due to lack of scientific relationship between froth structure and various aspects of flotation process. Bubble size distribution which is regarded as the most important characteristics of froth structure, is being addressed in this thesis by using a segmentation algorithm. A marker based watershed algorithm had been adopted and improved so as to prevent the over segmentation of big bubbles and able to adapt itself with different scenario of froth images. This results in a measurement of bubble size with high precision. The performance of improved marker based watershed algorithm was validated by using several industrial and laboratory froth images. In addition, several algorithms were implemented to measure the other important image variables such as froth velocity,froth color and bubble collapse rate. A froth model correlating the image variables to process variables and a prediction system estimating the metallurgical parameters based on image variables were then developed by using a neural network structure. A control strategy based on froth model was then designed in order to optimize the visual characteristics of froth, which lead to the control of the metallurgical parameters in an indirect manner. Finally, a control strategy implementing the developed froth model and prediction system was introduced for direct optimization of metallurgical parameters. Simulation results indicated the effective performance of the designed control schemes in enhancing theM overall efficiency of the process.

Preview PDF FK 2015 76RR.pdf Download (1MB) | Preview

Actions (login required)

View Item