Key Benefits

  • Maximum flotation recovery
  • Reduced waste
  • Lower OPEX
  • Quick ROI

Smart vision technology for comprehensive froth analysis

We’re experiencing a period of skyrocketing demand for minerals, at a time when ore grades are declining and environmental impact is a primary concern. Across the world, mine operators are being asked to do more with less.


Fortunately, we’re also living in an age of incredible advances in technology. Digitally-enabled machines. AI. Machine learning. All working together to make our equipment smarter, more efficient and more insightful. Because the road to maximum recovery is paved with data. The more we know, the more efficiently we can process materials.


When it comes to froth analysis, traditional vision systems have focused on measuring froth velocity and froth stability. These measurements remain important, but they don’t always provide enough information to enable enhanced recovery. That’s why our smart vision technology offers a third and cutting-edge feature: the ability to analyze froth in its entirety and classify the froth by types and conditions.


How is this possible? We’ve integrated AI and machine learning-driven smart technology into our froth camera. Deep neural networks (DNN) – algorithms, if you like – are employed for object detection. So that when you train your camera on your flotation cell, it is possible to identify virtually any object, as well as the property of the object.

Multiple cost benefits of improved flotation control

Maximum material recovery

Our smart vision system uses Deep Neural Networks to optimize process and reagent control. Although the technology is complex, the principle is simple: you can see more, and therefore you can recover more.


Reduced waste

Enhanced recovery means less of your valuable product goes into tailings. That’s good for your profit margin – and for the environment.

Lower OPEX

Because our vision system doesn’t alter your process, you are getting an increased output for the same input – meaning a lower cost per tonne.


Quick ROI

All these benefits quickly pay for the cost of the system, making this a no-brainer for mine operators looking for affordable ways to increase productivity.

Automated optimisation using AI-driven froth cameras

Froth camera using Deep Neural Networks (DNN)

Our DNN-based system allows analysis far beyond what is possible with traditional camera analysis. We use this system to measure froth velocity and to infer the froth grade, which then allows us to quantify the mass flow rate of each flotation cell and make decisions on which cell levels should be increased or decreased based on current grade/recovery curve and the exact behaviour of each individually optimised flotation cell.


This allows us to avoid the costly errors with traditional vision systems that prevent maximum recovery. Specifically, cell lip interference and pulping are two issues that traditional camera systems can’t accommodate for. Traditional camera systems rely on velocity measurements and can’t account for cell lip interference when the froth pushes in waves, or flows towards a cell lip, but doesn’t spill over due to low levels in the cell or build-up on the lip. Our system knows intuitively when the velocity measurement contributes to accurate results and simply leaves it out when it doesn’t.


Flotation circuit optimisation with ECS/ProcessExpert® 

Optimisation of the flotation circuit begins with mastering level control. Disruptions happen, but with model predictive control (MPC) we can plan for them – enabling precise level control, waves or no waves. This ensures the flotation cells can achieve the desired pull rates quickly and in a stable manner whenever the expert system commands a set point change. 


Our expert system, ECS/ProcessExpert®, uses the signals from the camera, model predictive control (MPC) and a fuzzy logic process control to optimize operation. This combination allows us to perfectly stabilize flotation cell level control, thus reducing variability in the cell, and maximise recovery at the desired grade. 

Stories from FLSmidth