Key terms in process control and mineral processing.

Advanced Process Control (APC) is any control strategy that goes beyond basic PID (proportional-integral-derivative) feedback loops. APC systems use mathematical models of the process to predict future behaviour and make coordinated control moves across multiple variables simultaneously. In mineral processing, APC typically refers to Model Predictive Control (MPC) applied to grinding or flotation circuits. The business case for APC is that it holds the plant closer to its operating constraints than a human operator can sustain continuously — translating directly into higher throughput and lower energy per tonne.

Model Predictive Control (MPC) is the most widely deployed form of Advanced Process Control in mineral processing. An MPC controller uses a process model to predict how the circuit will respond to a set of control moves over a future time horizon, then solves an optimisation problem every control cycle to find the best sequence of moves — subject to hard constraints on equipment limits and setpoint bounds. The result is a controller that manages multiple interacting variables (feed rate, water addition, mill speed, cyclone pressure) simultaneously and anticipates their interactions before they cascade. Aeterna PROCESSOPT is an MPC system purpose-built for SABC grinding circuits.

A process model is a mathematical representation of how a plant or circuit responds to changes in its inputs — feed rate, water addition, ore hardness, reagent dosage, and so on. In the context of MPC, the process model is what the controller uses to predict future states and plan its moves. Process models range from empirical (fitted to historical data) to first-principles (derived from physical and chemical laws). Aeterna's digital twin is a continuously calibrated, hybrid process model that combines first-principles structure with live measurement re-anchoring every control cycle.

A digital twin is a live, continuously updated software model of a physical asset or process. In mineral processing, a grinding circuit digital twin holds mass balance, mill filling, power draw, breakage rates, and classification — all calibrated against live sensor measurements. Unlike a static simulation, a digital twin re-anchors its parameters as new data arrives, so it tracks what the plant is actually doing rather than what it was doing when the model was first built. Aeterna's digital twin is the process model the MPC controller acts on.

Mineral processing (also called mineral beneficiation or ore dressing) is the sequence of physical and chemical operations that extract valuable minerals from mined ore. A typical flow includes comminution (crushing and grinding), classification (cyclones, screens), and concentration (flotation, leaching, gravity separation). The goal is to produce a concentrate with a high enough grade and recovery to be economically processed into a final metal product. Advanced process control and smart instruments target the comminution and flotation stages, where small efficiency gains translate into significant revenue.

Process optimisation in mineral processing means operating the circuit continuously at its most productive and efficient point — maximum throughput at target grind size, minimum energy per tonne, and best recovery for the ore type. MPC achieves this by holding the plant against its operating constraints in real time: where operators conservatively back off from the limit to avoid trips and overloads, MPC holds the setpoint at the constraint and corrects drift the moment it occurs. The cumulative effect across a full operating year is typically a 1–3% throughput uplift in SABC circuits.

A smart instrument (also called an intelligent instrument) is a sensor or measurement device that combines raw data acquisition with on-board processing — typically machine learning or computer vision inference — to produce a higher-level measurement that was previously unavailable or required manual sampling. In mineral processing, Aeterna's smart instruments include the Rock Analyzer (rock size from a conveyor camera), the Froth Analyzer (froth texture and bubble size from a flotation cell camera), and the Ball Counter (steel ball consumption from a conveyor camera). Each instrument runs inference on edge hardware at the plant, with no cloud connection required during operation.

The Rock Analyzer — also marketed as RockSense — is an intelligent instrument mounted above the SAG feed conveyor. It uses an industrial camera and edge inference to measure the size distribution of incoming ore in real time: P80 feed size, top size, and the proportion of critical-size material. This feed characterisation signal updates the process model's breakage parameters before the ore even reaches the mill, giving the MPC controller time to adjust feed rate, water addition, and mill speed proactively rather than reactively.

The Froth Analyzer — also marketed as FrothSense — is an intelligent instrument that monitors the surface of flotation cells using an industrial camera mounted above the cell. Edge inference models measure froth texture, bubble size distribution, froth velocity, and stability in real time — variables that are visually apparent to an experienced operator but difficult to quantify consistently. These measurements are earlier indicators of recovery changes than any downstream assay, giving the process controller time to adjust reagent dosage, air rate, and pulp level before recovery drops.

SABC (Semi-Autogenous Ball mill Crusher) is the most common grinding circuit configuration in hard-rock mining. Ore passes through a large semi-autogenous grinding (SAG) mill, then a ball mill for finer grinding, with a pebble crusher in closed circuit with the SAG. Product from the ball mill passes through hydrocyclones for classification before reporting to flotation or leaching. SABC circuits are energy-intensive and highly interactive — changes in SAG feed size, water addition, and ball charge all affect each other — which makes them well-suited to Model Predictive Control.

Moving Horizon Estimation (MHE) is an advanced state estimation technique that uses an optimisation problem over a sliding window of past measurements to estimate unmeasured process states and model parameters. Unlike a Kalman filter, MHE can handle nonlinear systems and hard constraints on state and parameter values. In process control, MHE is used to continuously adapt the process model to changing plant conditions — ore hardness shifts, equipment wear, or seasonal changes — keeping the model anchored to reality as the plant evolves. Aeterna uses MHE to manage model drift between planned model reviews.