Key Takeaways
- AI thickener control is a published, validated strategy. Reinforcement-learning controllers using proximal policy optimization (PPO) have reported flocculant savings of 10 to 15% in the literature.
- There is no single dominant “AI thickener” product yet. The capability ships as an add-on within general advanced process control (APC) platforms from ABB, Metso and Yokogawa, not as a standalone thickening tool.
- The AI target is flocculant dosing and underflow density: getting the right amount of chemical in at the right time as feed conditions (%solids, mineralogy) fluctuate.
- It follows the same closed-loop AI and APC pattern as grinding and flotation control, as part of a broader plant-wide advanced control adoption.
TL;DR
AI thickener control applies a model-predictive or reinforcement-learning controller to flocculant dosing and underflow density. It is available today as a module within APC platforms (ABB, Metso, Yokogawa) or through AI-native platforms like Imubit, rather than as a dedicated thickening product.
What Is Thickener Control?
A thickener is a large settling tank that separates water from slurry. Solids settle to the bottom and leave as a dense underflow, while clear water overflows and returns to the plant. Flocculant, a chemical additive, makes fine particles clump together so they settle faster.
Thickener control means holding underflow density, solids interface height and overflow clarity on target. The main levers are flocculant dosage and the underflow pump rate. Getting dosing right matters a lot. Too little flocculant means poor solid-liquid separation. Too much wastes an expensive reagent and can degrade underflow quality. Because thickened tailings often go straight to the tailings facility, thickener performance also affects water recovery and tailings storage.
How Do I Optimize Thickening With AI?
Thickening is a genuinely dynamic control problem. Incoming feed characteristics, such as %solids, particle size distribution and mineralogy from a changing ore blend, shift continuously. A fixed dosing rate tuned for average conditions leaves money on the table either way.
What the research shows
Research has applied reinforcement learning to this exact problem. A 2024 study in Minerals Engineering trained a PPO controller on a simulated conventional thickener. The controller adjusted flocculant dosage and underflow rate to hold slurry density and interface height, and it handled disturbances while using flocculant efficiently. Literature on this approach reports flocculant savings in the 10 to 15% range, which is a meaningful reagent-cost reduction at plant scale.
How thickener control is sold today
Be clear-eyed: this is a validated control strategy more than a shrink-wrapped product. No single vendor sells “AI thickener control” as a discrete product the way TOMRA sells ore sorters. In practice, the capability shows up as a module within the same APC platforms already handling grinding and flotation. ABB Ability Expert Optimizer, Metso’s control suite and Yokogawa’s process automation all offer thickener and dewatering optimisation within a broader plant-wide deployment. AI-native platforms like Imubit apply the same closed-loop neural-network approach here as they do to grinding.
Bundle thickener control with wider APC
The practical adoption path mirrors grinding circuit optimization and AI flotation control. This is rarely a standalone thickening project. If you already work with an APC vendor or an AI-native closed-loop platform for comminution or flotation, ask what their thickener and dewatering module covers. You will typically get more value bundling it in than buying a separate point solution.
Where Thickener Control Pays Off
The benefits reach beyond reagent cost. Denser underflow means more water returns to the plant, which matters at sites with tight water balances. It also means less water sent to the tailings facility, which supports tailings dam monitoring and dam safety. Stable thickener operation also reduces upsets that ripple back into flotation.
FAQs
What does a thickener do in mineral processing?
A thickener separates solids from water in a slurry. Solids settle to the bottom and are pumped out as a dense underflow, while clarified water overflows and is reused in the plant. Thickeners are used both for concentrate dewatering and for thickening tailings before storage.
How does AI reduce flocculant use?
AI adjusts flocculant dosage continuously as feed conditions change, instead of using a fixed rate set for average conditions. It learns from sensor data how dosage affects settling and underflow density. As a result, it adds only as much flocculant as the current feed needs.
Is there an off-the-shelf AI thickener control product?
Not as a standalone product. AI thickener control is usually delivered as a module within advanced process control platforms from vendors such as ABB, Metso and Yokogawa, or through AI-native optimisation platforms. Most sites add it as part of a wider plant control project.
What sensors does thickener control need?
Typical inputs include feed flow and density, flocculant flow, underflow density, bed or interface level, and overflow turbidity. Rake torque is also important because it warns of overloading. Reliable instruments and a plant historian are the foundation for any AI controller.
Try It With Geocluster
Sorting out which APC vendor’s thickener module is backed by real deployment data, rather than marketing, is exactly the kind of evidence-gathering Geocluster is built to support. So is judging whether a reinforcement-learning approach is mature enough for your plant.