This article describes how the Flotation Optimizer works. For an introduction to the Optimizer, read this article first.
The Flotation Optimizer does hundreds of simulations of a Flotation circuit using Digital Process Models, deciding on the solution that best aligns with the Value Driver, and provides the resulting Control Variable setpoints as Recommended Setpoints (RSPs).
We will unpack these parts in this article.
1. Digital Process Models
Under the hood of the Flotation App is a combination of models that digitally capture the complex relationships/correlations between the following variables of the Flotation circuit:
Feed (and disturbance) variables: the variables describing the feed, as well as other measured but uncontrolled variables.
Control variables: process variables that can be changed & stabilised by the lower-level control systems
Performance variables: Flotation circuit KPIs like product grade, mass pull and metal recovery.
The outputs of the Digital Process Models are delivered to operations teams as Virtual Sensors, giving them valuable visibility on their process that they haven't had before. The core function of these Digital Process Models, though, is to give the Flotation Optimizer sufficient "knowledge" (or artificial intelligence) of the process to be able to find the optimum.
There are several Digital Process Models in the Flotation App's modelling suite. These include:
Hydrodynamic models, modelling the relationship between air and slurry flows in flotation cells on the hydrodynamic properties like gas hold-up, bubble size and bubble surface area flux. More details here: Overview of Hydrodynamic Virtual Sensors & More in-depth Hydrodynamics theory,
Separation & mass balance models, modelling the kinetic relationship between feed properties and cell hydrodynamics on mineral separation performance in a flotation circuit.
Reagent performance models, modelling the relationship between feed properties, reagent dosing and flotation circuit performance. More details on how this is used by the Optimizer here.
2. Value Driver
The Value Driver is the core configuration of the Flotation Optimizer, and together makes up what is often called the objective function of the Optimizer. It consists of:
Performance Variable Rewards
Each individual Reward is associated with a specific Performance Variable. E.g. maximise metal recovery, or keep the final concentrate metal grade above x g/t. Each Reward is given a particular weight, to reflect its relative importance or priority among other Rewards. When several Rewards are then put together, they should define & reflect the purpose of that Flotation circuit. They define what the operations team mean with "good performance". This is set and adjusted together with the Metallurgist or Process Engineer, so that they can define what they want from their plant.Control Variable Limits
Above all a Flotation circuit should operate in a manner that is safe and sustainable (meaning the process can keep operating at that point in a stable way). This means that the Recommended Setpoints it provides to the control systems or Operator should be constrained. Each Control Variable is given a min & max limit, and the Optimizer will not look for solutions outside of these limits.
The Control Variables are also given a Rate of Change limit (user configurable), limiting how far the next Recommended Setpoints are allowed to be from the current Control Variable values or most recent Recommended Setpoints. This makes sure that there aren't any big jumps in how the circuit operates.
3. Optimizer
Once the Digital Process Models have been configured and calibrated, and the Value Driver has been set up, the Flotation Optimizer can optimize the Flotation circuit. This means: given the feed properties, it finds the Control variable setpoints that will give the best possible Performance. It does this by running hundreds of simulations at any given moment to find the best possible solution.
Next read: