Real-time Flotation Optimization: Application of Machine Learning Models to Improve Process Efficiency and Stability

Shared Learning from Flotation 2022

In the mining industry, controlling flotation variables and stabilizing column cells have always been challenging aspects of plant operations, particularly for metal recovery. With the increasing pressure towards more efficient and sustainable processes, and the constant increase in the cost of chemical inputs and energy, innovation in the field of real-time process control is under the spotlight.

Flotation is one of the most widely used methods to separate and concentrate ores, and this is achieved by altering the surfaces of minerals to a hydrophobic or hydrophilic* condition through the use of chemical additives.  However, the industrial flotation process is much more complex with a high degree of uncertainty requiring more than laboratory work to predict its behavior. Even modeling on longer timescales for design purposes is challenging while predicting the real-time operational performance of flotation is more difficult still.

At, we have developed world-class modeling approaches to alleviate mining process bottlenecks such as flotation

Flotation Modelling Approach


In order to showcase the advances we have made in this field, we recently presented our methodology for the optimization of flotation at the Flotation 2022 Conference.

 In Flotation, achieving throughput, grade, and recovery targets are challenging as operators face a reactive and manual approach, often responding to the appearance of the froth, which is compounded by constant changes in feed properties. In addition, limited real-time visibility on circuits and low feedback on how to set control setpoints for optimal recovery threatens flotation performance.

One barrier to data-driven discovery is that existing machine learning methods often do not satisfy the needs of scientific models, since these models must also respect or incorporate the laws of physics as well as process and operation constraints. 


In this way, proposes to analyze flotation behavior within a columnar cell using a method based on neural networks with: 

1) data measured directly
2) other data calculated with the process hydrodynamics
3) physics equations phenomena, all embedded in the main neural model.

The typical variables measured in a columnar cell are the air flow (Jg), pulp flow (Jl), foam height (hf), and other variables such as % solids (Cp), particle size – P80, type and dose of reactants, which are obtained from the circuit. But these variables alone are not enough to give accurate modeling of the process, it is necessary to add hydrodynamic variables, such as the bubble size (db), gas holdup (Eg), and the bubble surface area flux (Sb) that are difficult to measure in real-time. In order to manage the entire process, it is still important to account for recoveries and concentration laws, therefore, adding distribution models of residence time (DTR) and kinetic models is needed to better model flotation behavior.

The results achieved confirm that this type of modeling can be applied to analyze flotation columnar cell behavior and make predictions of grades or other variables in real-time. Our Application uses a method based on a machine learning (ML) process model combined with an explainable and real-time configurable optimization engine.

More about Flotation Optimization Application.

It’s Time to Digitalize

Stimulating digitization to improve business performance, sustainability and reduce risk is part of‘s core values. Through the use of Scientific AI*, we have built digital applications to facilitate decision-making processes in one of the harshest environments. The new machine learning models developed by have the capacity to learn complex non-linear dynamics, making precise predictions possible. By doing so, our apps can drive process improvements, identify and remove bottlenecks, increase asset availability, reliability and utilization even for a single operating unit (like flotation, thickeners, stockpiles, grinding, heap leaching, pumping, etc) or work in tandem to optimize overall mine-to-market efficiency.

Our suite of Applications helps optimize processes in real-time and guide operators, metallurgists and lower-level control systems on how to gain optimal throughput and recovery. With the assistance of our flotation virtual sensors (embedded with ML models) it is possible to achieve higher metal recovery through better prediction of material inputs (ore mineralogy and PSD), deliver OPEX savings, and also reduce localized events by highlighting underperforming cells/columns.

*The Froth flotation is a process that selectively separates materials based upon whether they are water-repelling (hydrophobic) or have an affinity for water (hydrophilic).

*Scientific AI is the fusion of physics models with machine learning techniques. It was developed by to further increase net metal production through the use of our AI solutions.

The Application Portfolio