Skip to main content
Publicly Available Case
How academic/industry teams stabilized flotation through video-based froth prediction
Deep learning forecasts froth states seconds ahead, letting APC tune setpoints earlier. Operations become steadier with higher recovery.
Context
Delayed froth response made it difficult for operators to adjust the flotation circuit in time to prevent instability. These lags reduced recovery and required constant manual attention.
Solution
A spatiotemporal deep-learning model was developed to forecast froth motion and stability. Predictions were fed into control actions to pre-empt fluctuations in the circuit.
Results
The approach produced more stable flotation performance and improved recovery by anticipating froth behavior before it became unstable.
Data Inputs
Froth video sequences; process tags