Abstract:
Froth flotation models that are developed for circuit design applications are often not suitable for model-based dynamic control and optimisation applications. For real-time control and optimisation applications dynamic models of the key flotation mechanisms are required, as these use real-time measurements to update internal model states and estimate model parameters in real-time.
The development of a dynamic froth flotation model is described, based on a combination of fundamental mass and volume balances, fundamental steady-state froth models and empirical models for bubble size and air recovery. The model outputs are defined to correspond with real-time measurements that are commonly available on industrial flotation circuits, including measurements from froth imaging devices in combination with measurements of levels, flow rates, densities and grades.
The flotation model is analysed for state observability and controllability, and it is shown that the model states and parameters can be estimated from real-time process measurements that are commonly available on industrial flotation circuits. The ability to estimate process parameters in real-time opens up opportunities for improved process control and optimisation by compensating for a specific flotation mechanism rather than the combined effect of multiple flotation mechanisms. The speed of response can also be improved when more accurate models are maintained by continuously updating model parameters.
The flotation model, a state and a parameter estimator and model predictive controller are combined to simulate the potential benefits of using a non-linear model-based approach with state and parameter estimation capabilities in a dynamic control and optimisation application on flotation circuits. The strategy is shown to reject typical process disturbances effectively in the presence of process noise and outperforms a linear non-model based control strategy by a significant margin.
