Structuring a flotation model for robust prediction of flotation circuit performance

The complexity of the flotation process has made the identification of an appropriate model structure, for use in the prediction of flotation circuit performance, far from obvious. Work performed under the auspices of the AMIRA P9 flotation program has indicated that the metallurgical performance achieved from a flotation cell arises from the interaction of two essentially independent property classes, namely the cell operating conditions and the properties of the cell feed stream. These observations have led to the development of a model structure that associates a constant set of flotation parameters to the streams of the circuit, which are transformed into a flotation response depending on the operating conditions within the flotation cell. A variety of alternative model structures exist within the literature which do not make this distinction between cell and stream flotation properties, and many of these models are in widespread use for flotation optimisation and design studies, both in industry and academia. In this paper, data collected from an industrial flotation circuit is used to derive the parameters associated with two different types of flotation models. Simulations performed using these models demonstrate the need to differentiate between cell and feed properties. A model that does not differentiate between these two properties fails to successfully predict the metallurgical performance of a flotation circuit after a disturbance.