Image Analysis Based Control of Copper Flotation

This paper presents recent results and experiences of using real-time image analysis in the control of copper flotation at the Pyhäsalmi Mine Oy (Finland). First we introduce our measurement set-up and the image variables it produces. After that the image variables are correlated against process variables using Partial Least Squares and clustering. The results show that some of the process disturbances are seen 30-40 minutes earlier in specific image variables when compared to the on-stream X-ray analyses. Furthermore, the image variables help in deducing what has caused the process disturbance and what should be the corrective action. Hence the most important image variables were added into the rule-based controller of the cleaner bank. Comparison results then show that this addition of information helps us to run the process near its optimum in a reliable fashion.