AI in flotation is one of the most mature uses of industrial AI in mineral processing. Closed-loop reinforcement-learning controllers already write setpoints directly to plant control systems, rather than just recommending them to an operator. This category covers that control layer and the plant-side work around it: predicting performance, keeping the feed steady and reconciling the plant with its design.
What AI Does in Flotation
Flotation is continuous, sensor-rich and quick to respond, so models can learn from it fast. AI in flotation does three jobs. It senses the process through froth cameras and plant instruments. It predicts recovery and grade before lab assays arrive. Finally, it controls the circuit by adjusting reagent dosage, airflow and pH in real time.
Where to Start
Each page below answers one “how do I do this with AI?” question.
- Control the circuit. AI flotation control pairs froth cameras with closed-loop platforms such as Imubit that write setpoints directly. If you want predictions before automation, build a flotation soft sensor first.
- Steady the feed. Ore blending optimization keeps grade and hardness within the range the plant handles best.
- Watch the whole plant. Turn periodic mill surveys into continuous ones, and use digital twins to track plant variability against the design model.
- Plan for tailings. Acid rock drainage prediction forecasts how flotation tailings and waste rock will behave over decades.
What’s Mature and What’s Emerging
Closed-loop flotation control is proven, with reported gains of up to about 15% in recovery and 20% in reagent use. Soft sensors for recovery and throughput are well established too. By contrast, concentrate grade prediction, AI mill surveys and machine learning for acid rock drainage are still emerging. Treat vendor claims in those areas with more scrutiny.