An Optimal Control Model
The control model operates on a combination of AI neural network principles and fuzzy logic. This allows for the flexibility to change the array of sensors and controls used. Optimal control actions are selected for all flotation machines based on the prediction of grades and flowrates at all points of the section taking into account connections between flotation machines. The controls maximise recovery, ensuring concentrate grade does not fall below the required minimum.
How does the model work?
- Virtual Sensors. The model calculates flowrate and recovery from sensor reading through grade changes and mass balance.
- Prediction. The model predicts grades and concentrate flowrate 60 minutes in advance.
- Self-correction. The predictive model is corrected in case of discrepancies between the estimates and the sensor data.
- Selection of control settings. Controls that maximise recoveries, while meeting the specified limits on metal grades and concentrate flowrate are required based on the predictions.
- Incident handling. In case of equipment malfunctions the system notifies the operator and continues control taking into account the detected abnormal situation.
Control logic for flotation machines
1. Development of a core predictive model on one flotation machine and control of that machine at the head of a given flotation section
2. Development of links between several core models of flotation machines (including pump models, sump) and the development of joint control
3. Scaling and working out end-to-end control of the entire flotation section
The end-to-end optimisation of flotation circuit: