The Relationship between Froth Image Features and Platinum Flotation Grade
MetadataShow full item record
Many endeavors have been made attempting to relate froth appearance to key performanceindicators such as grade and recovery. The platinum industry have questioned the possibility of doing this based on the results of simple models trying to relate physical features (e.g. bubble size) to grade.In this study a fractional factorial design of laboratory experiments were conducted using platinum ore.The flowrate (x1), pulp level (x2), collector (x3), activator (x4), frother (x5) and depressant (x6) werevaried for each experiment and videos were recorded of the froth. The analysis of the videos rendered 20texture features from each image.The non-linear relationship between the froth texture features and grade were investigated using kernelFisher Discriminant Analysis (kFDA). The results indicate that these features could be sufficient as input into advanced model based control systems for PGM flotation plants. Multivariate linear regressionconfirmed the hypothesis and performed very well when the kFDA features are used as predictors.
Showing items related by title, author, creator and subject.
The estimation of platinum flotation grade from froth image features by using artificial neural networksMarais, C.; Aldrich, Chris (2011)The use of machine vision in the monitoring and control of metallurgical plants has become a very attractive option in the last decade, especially since computing power has increased drastically inthe last few years. The ...
Marais, C.; Aldrich, Chris (2011)Features extracted online from froth images on flotation plants are potentially useful to the developmentof advanced control for flotation systems, provided that these features can be related to the key indicatorvariables ...
Horn, Z.; Auret, L.; McCoy, J.; Aldrich, Chris; Herbst, B. (2017)Image-based soft sensors are of interest in process industries due to their cost-effective and non-intrusive properties. Unlike most multivariate inputs, images are highly dimensional, requiring the use of feature extractors ...