Circuito AI vs ANDRITZ BrainWave: AI-Native vs MPC
Circuito AI vs ANDRITZ BrainWave for mineral processing: reinforcement learning vs MPC, published results, and LATAM deployment models compared.
TL;DR
ANDRITZ BrainWave is a mature, well-proven model predictive control (MPC) platform backed by a multi-billion-euro conglomerate with strong Latin American presence and published results at named copper mines. Circuito AI is an AI-native platform purpose-built for mineral processing, using reinforcement learning that continuously retrains against changing ore conditions. Both deploy on-premises and integrate with any DCS. The fundamental question for your operation is whether you need a proven but static control model, or an adaptive AI system that learns autonomously as your process evolves.
Introduction
Mining operations evaluating advanced process control (APC) for grinding and flotation circuits will encounter both ANDRITZ and Circuito AI on their shortlists. Both are serious platforms with on-premises deployment, vendor-neutral DCS integration, and documented metallurgical improvements. Both can operate in advisory mode before transitioning to closed-loop control.
The similarity, however, is largely architectural. Beneath the surface, these two platforms represent fundamentally different philosophies. ANDRITZ BrainWave is rooted in patented adaptive MPC — a mature control paradigm refined over decades. Circuito AI is built from the ground up on reinforcement learning and neural networks — an AI-native architecture designed for the nonlinear, constantly shifting reality of mineral processing.
Technology: Patented MPC vs Reinforcement Learning
ANDRITZ BrainWave
BrainWave is ANDRITZ’s flagship advanced process control product. It uses a patented adaptive MPC approach based on Laguerre network technology. In practice, this means BrainWave builds a mathematical model of your process, defines constraints and setpoints, and then uses that model to calculate optimal control moves within a predicted horizon.
The system also includes ACE (Advanced Control Expert), a modular supervisory control layer, and IDEAS, a dynamic simulation engine that serves as a digital twin for offline testing and operator training. ANDRITZ has publicly stated that they are training reinforcement learning agents on these digital twins as a “next level” capability, though this remains in development rather than standard deployment.
BrainWave’s MPC approach is well understood by control engineers and has a strong track record. It deploys quickly — often in weeks — and produces measurable results in variability reduction.
Circuito AI
Circuito AI takes a different approach entirely. The platform is built on reinforcement learning (RL) and neural networks — the same class of algorithms that learned to play chess and Go at superhuman levels, now applied to mineral processing control.
Rather than requiring engineers to build a first-principles model of the process, Circuito AI’s AI learns the process dynamics directly from operational data. It writes setpoints to the DCS at 10-second intervals, continuously optimizing in real time. Critically, the models retrain automatically as ore characteristics, reagent conditions, and equipment wear change over time.
This distinction matters most in operations where ore variability is high. A traditional MPC model is tuned for a specific set of process conditions. When those conditions drift — new ore blend, seasonal water quality changes, liner wear in mills — the model’s accuracy degrades until an engineer retunes it. Circuito AI’s reinforcement learning agent adapts to these changes autonomously, without manual intervention.
For a deeper look at Circuito AI’s architecture and how it handles ore variability, see our platform overview.
AI Maturity: Roadmap vs Production
This is perhaps the most important distinction for technical evaluators.
ANDRITZ has publicly positioned AI and machine learning as the “next level” of their digital transformation strategy. Dr. Sohail Nazari, who leads their digital transformation efforts, has described a progression from traditional BrainWave MPC to AI-enhanced control using reinforcement learning trained on their IDEAS digital twin. This is promising work, but it represents a future capability rather than what ships today. When you deploy BrainWave at your site in 2026, you are deploying MPC with supervisory logic — a proven approach, but not an AI-native one.
Circuito AI deploys AI-native from day one. Reinforcement learning is not a roadmap item or an add-on module; it is the core control engine. The platform has accumulated over 20,000 production hours of AI-driven control across operating mines. Every deployment uses the same RL architecture, and every deployment continuously retrains.
The practical implication: ask your vendor whether their AI is in production at operating mines writing setpoints to a DCS, or whether it is in a simulation environment being validated against a digital twin. The answer reveals where each platform stands on the maturity curve.
Published Results in LATAM
ANDRITZ: Named mines, published numbers
ANDRITZ deserves credit for transparency in their LATAM results. They have published performance data from named operations:
- Collahuasi (Chile): +2.3% copper recovery improvement in flotation
- Los Pelambres (Chile): +1.8% copper recovery improvement in flotation
- Constancia (Peru): 54-63% reduction in flotation level variability
In grinding circuits, ANDRITZ has published variability reductions of 32-39% in P80 and 37-52% in hydrocyclone overflow density across various deployments. They cite approximately 196 deployments globally over the past decade, and won Goldcorp/Newmont’s Disrupt Mining competition in 2019.
These are real numbers at real mines in Latin America. Any honest comparison must acknowledge this.
Circuito AI: Different metrics, equally meaningful
Circuito AI’s published results come from different geographies and measure different KPIs:
- Ball mill grinding: +2.6% throughput increase
- SAG/AG grinding: +1.4% throughput increase
- Flotation: Proprietary optimization methodology with results verified through rigorous A/B testing
Named clients include Norilsk Nickel and Russian Copper Company — large-scale operations processing complex ores. Circuito AI brings this operational experience to Latin American mining operations.
The difference in reported metrics is worth noting. ANDRITZ emphasizes recovery improvement (percentage points of Cu recovery) and variability reduction. Circuito AI emphasizes throughput gains and adaptive performance. Both are economically meaningful — a 2.6% throughput increase at a large SAG mill can be worth millions of dollars annually, just as a 2.3% recovery improvement at Collahuasi represents significant additional copper production.
For detailed performance data and methodology, visit our results page.
Process Coverage
ANDRITZ
ANDRITZ offers broad coverage across the mineral processing flowsheet:
- Grinding circuits (SAG, ball mill, hydrocyclone)
- Flotation (rougher, cleaner, scavenger)
- Thickening and dewatering
- Plant-wide supervisory control via ACE
- IDEAS digital twin for simulation and training
This breadth reflects ANDRITZ’s position as a diversified industrial technology company. Their mining division also benefits from cross-pollination with pulp and paper process control, where BrainWave has extensive deployment history.
Circuito AI
Circuito AI focuses specifically on concentrator circuits:
- Grinding optimization (SAG, AG, ball mills)
- Flotation optimization (described as a unique global capability)
- Crushing circuits
Rather than covering the entire plant with supervisory logic, Circuito AI goes deeper into the concentrator — the section of the plant where ore variability has the greatest impact on recovery and throughput. The 10-second control frequency reflects this focus: the AI is making real-time decisions at the speed needed for dynamic flotation and grinding control, not just supervisory setpoint adjustments.
Company Profile and Strategic Fit
ANDRITZ
ANDRITZ is a publicly traded Austrian industrial group with over 8 billion euros in annual group revenue. Founded in 1852, the company operates across four major divisions: pulp and paper, metals, hydropower, and separation (which includes mining). Mining is one part of a large, diversified portfolio.
This scale brings advantages: deep engineering resources, global service networks, financial stability, and the ability to bundle process control with equipment supply. It also means mining competes for R&D budget and executive attention with other divisions.
Circuito AI
Circuito AI is 100% focused on mining and mineral processing. The team consists of mining PhDs and process control specialists who work exclusively on concentrator optimization. No other divisions, no other industries — every engineering hour goes into mineral processing AI.
For operations that want a partner whose entire business depends on their mining AI delivering results, this focus can be a decisive factor. For operations that prefer the security of a multi-billion-euro parent company, ANDRITZ offers that stability.
LATAM Presence and Support
ANDRITZ
ANDRITZ has well-established operations in Latin America, with offices in Sao Paulo (Brazil), Santiago (Chile), and Talcahuano (Chile). They have local engineers, existing service contracts at major mines, and years of relationships with LATAM mining companies. Their Collahuasi, Los Pelambres, and Constancia deployments demonstrate an active LATAM install base.
This is a genuine competitive advantage. Local presence means faster support response, easier site visits, and cultural familiarity with LATAM mining operations.
Circuito AI
Circuito AI is entering Latin America through Circuito AI, the exclusive regional partner for all commercial operations across Mexico, Central America, the Caribbean, and South America. Circuito AI combines Circuito AI’s AI platform with regional mining expertise and on-the-ground commercial capability.
While newer to the region, Circuito AI is building its LATAM presence specifically for mining AI — not as one business unit among many, but as the sole focus.
Who Each Platform Is Best For
ANDRITZ BrainWave is the right choice if your operation:
- Values a proven MPC track record at named LATAM copper mines
- Needs broad plant-wide supervisory control beyond the concentrator
- Prefers the backing of a publicly traded, multi-billion-euro parent company
- Wants to bundle process control with ANDRITZ equipment or services
- Has process conditions that are relatively stable and predictable
- Has existing ANDRITZ relationships or equipment on site
Circuito AI is the right choice if your operation:
- Needs AI that adapts autonomously to changing ore, reagents, and equipment conditions
- Wants a platform purpose-built for mineral processing, not adapted from pulp and paper
- Prioritizes real-time adaptive control (10-second frequency) over supervisory MPC
- Values A/B verified results methodology
- Is looking for a partner whose entire business depends on mining AI performance
- Wants to deploy AI-native technology today, not wait for an AI roadmap to mature
Summary Comparison
| Dimension | ANDRITZ BrainWave | Circuito AI |
|---|---|---|
| Core technology | Patented adaptive MPC (Laguerre) | Reinforcement learning + neural networks |
| AI in production | MPC in production; RL on roadmap | AI-native from day one |
| Control frequency | Supervisory (minutes) | Real-time (10 seconds) |
| Model adaptation | Manual retuning by engineers | Continuous automatic retraining |
| DCS integration | Vendor-neutral | Vendor-neutral |
| Deployment | On-premises | On-premises |
| Grinding results | P80 variability -32 to -39% | +2.6% throughput (ball mill) |
| Flotation results | +2.3% Cu recovery (Collahuasi) | Proprietary methodology, A/B verified |
| LATAM presence | Offices in Brazil and Chile | Entering via Circuito AI |
| Company focus | Multi-industry conglomerate | 100% mineral processing |
| Global deployments | ~196 over 10 years | 20,000+ production hours |
| Parent company | ANDRITZ AG (public, 8B+ EUR revenue) | Independently held |
| Pricing model | Project-based + service contracts | Contact for details |
The Question Worth Asking
When evaluating either platform, consider this: when was the last time your process control model learned something new from your ore?
Static MPC models, however well-tuned at commissioning, gradually degrade as ore characteristics shift. Seasonal changes, new mining zones, reagent variability, and equipment wear all introduce drift that requires manual retuning. An AI-native platform that retrains continuously treats these changes not as disruptions but as new data to learn from.
Both ANDRITZ and Circuito AI are credible platforms with documented results. The choice between them ultimately reflects a deeper question about your operation’s technology strategy: do you want to optimize with the best of traditional control, or do you want to deploy the next generation of process intelligence?
If you are evaluating AI optimization for your LATAM mining operation, contact Circuito AI to discuss how Circuito AI’s adaptive AI compares to your current or prospective APC solution.