The best software choice for a mineral processing digital twin depends on whether your core focus is metallurgical mass/energy balancing, thermodynamic chemistry, or real-time control room integration . There is no single "one-size-fits-all" tool, but industry leaders dominate specific layers of plant modeling.

Top Software Solutions for Mineral Processing

Choosing the Right Approach for Your Plant

1. For Circuit Design & Metallurgical Simulation

• USIM PAC: Built specifically for mineral processing and extractive metallurgy, it lets you model variable ore feeds, track liberation sizes through grinding/flotation circuits, and predict grade-recovery curves.
• JKSimMet / Bruno (Metso): The industry standard for comminution (crushing and grinding) circuit modeling. While narrower in scope than a full plant twin, they provide rigorous physics-based breakage and liberation models.

2. For Chemical & Thermodynamic Balances

• HSC Chemistry: If your processing plant involves leaching, pressure oxidation, solvent extraction-electrowinning (SX-EW), or smelting, HSC provides unmatched thermo-chemical database modeling that connects directly to process flowsheets.

3. For Live Operational & Real-Time Digital Twins

• Metallurgical Systems or Honeywell: True operational digital twins require hooking models up to your plant historian (PI System, Aspentech, etc.). These platforms ingest live sensor streams, run real-time mass reconciliations, and expose bottlenecks or declining recovery rates on a central dashboard.

If your goal is a true digital twin of a mineral-processing plant, rather than simply a steady-state flowsheet model, I would not pick software based on “best simulator” alone. You need to combine mineral-processing process models + plant historian/OT data + dynamic simulation + optimization/analytics + visualization.

My shortlist

1. My first choice for the mineral-processing model: METSIM

METSIM is unusually well suited to mining because it was designed around metallurgical and chemical processes rather than primarily oil & gas.

It supports steady-state and dynamic simulation, mass/energy balances, comminution, flotation, gravity separation, dense media, leaching, solvent extraction, CIP/CIL, thickeners, filters, crushers, mills, hydrocyclones, pumps, etc. It also has a Mine Module that can link drill-hole/block-model information and mine schedules into the plant simulation.

That makes it particularly attractive if your twin needs to answer questions such as:

• What happens to recovery when ore hardness changes?
• What is the impact of feed grade/mineralogy?
• How should SAG/ball-mill operating conditions change?
• What happens if flotation residence time changes?
• How does a different ore blend propagate through the plant?
• What is the effect on concentrate grade/recovery and tailings?
• What happens to water/energy consumption?
• How does the mine plan affect plant performance?

If you're starting from scratch, METSIM would be high on my list for the physics/metallurgy layer.

2. Best overall digital-twin platform: AVEVA

If by “digital twin” you mean:

Live plant data → validated process model → real-time calculations → what-if simulation → optimization → operator/engineer visualization then I'd look very seriously at AVEVA.

AVEVA Process Simulation supports steady-state and dynamic simulation and, importantly, can integrate real-time data from AVEVA PI System to create an online digital twin. AVEVA's broader CONNECT architecture is intended to connect data, models, analytics and visualization.

That is a much closer match to a plant digital twin than simply buying a flowsheet simulator.

The caveat is that METSIM has considerably more obvious mineral/metallurgical specialization. So I'd think of the choice this way:

METSIM = excellent model of the ore/process

AVEVA = excellent industrial infrastructure around the model

3. One particularly interesting mining-specific option: Metallurgical Systems

Metallurgical Systems has a product called MI Core, specifically aimed at metallurgical digital twins.

This is worth evaluating if you're dealing with an existing operating plant rather than designing a greenfield facility. Their platform is designed around plant-wide mass/energy balances, real-time operational data, metallurgical accounting, dynamic simulation, scenario testing and sustainability metrics.

I'd put this on the shortlist for a producing mine where you already have years of historian, laboratory and metallurgical-accounting data.

4. AspenTech is excellent—but I'd rank it differently for mining

AspenTech is extremely strong for process simulation and digital twins. Aspen OnLine, for example, explicitly connects live operating data with process simulation and supports calibrated predictive models and cloud digital twins.

Aspen also has a strong hybrid-model philosophy: first-principles models combined with AI/ML.

But for a mineral concentrator, I'd generally investigate METSIM and the mining-specific solutions first. Aspen becomes particularly compelling if your plant has significant hydrometallurgy, pyrometallurgy, chemical processing, refining, utilities or downstream processing.

The architecture I'd recommend

I wouldn't try to make one piece of software do everything.

I'd build the twin approximately like this:

                    DIGITAL TWIN
                         │
          ┌──────────────┴──────────────┐
          │                             │
     PHYSICS MODEL                 DATA MODEL
          │                             │
      METSIM /                    PI System /
      specialist                  historian /
      models                      lab data
          │                             │
          └──────────────┬──────────────┘
                         │
                  CALIBRATION ENGINE
                         │
                ┌────────┴────────┐
                │                 │
          First-principles       ML/AI
             models              models
                │                 │
                └────────┬────────┘
                         │
                    OPTIMIZATION
                         │
          ┌──────────────┼──────────────┐
          │              │              │
       What-if       Prediction      Operator
       analysis     / forecasting     dashboard

The key is not to make the digital twin purely AI-driven.

For mineral processing, I'd use a hybrid twin:

First-principles metallurgical model + real plant data + machine learning + optimization. That gives you much better behavior when you encounter ore types or operating conditions that weren't represented heavily in the historical data.

What I would do for your project

If this were my project, I'd run a short technical evaluation of three architectures:

A. METSIM-centered

METSIM + historian + SQL/data platform + Python/ML + visualization

Best when metallurgical accuracy is the priority.

B. AVEVA-centered

AVEVA Process Simulation + PI System + CONNECT + analytics

Best when the goal is eventually a plant-wide operational digital twin integrated with OT/IT infrastructure. AVEVA specifically supports bringing simulations online with operational data.

C. Mining-specific operational twin

Metallurgical Systems MI Core + existing plant data infrastructure

Best candidate to investigate if your primary objective is operational metallurgical accounting, reconciliation, near-real-time plant performance and optimization.

My recommendation

If you tell me what kind of mineral plant you have—for example copper concentrator, gold, lithium, iron ore, nickel, phosphate, etc.—and roughly the flowsheet (crushing → grinding → flotation → thickening, leach/SX-EW, DMS, etc.), I can give you a much more specific recommendation.

I can also lay out a digital-twin architecture for your plant, including the process simulator, historian, data model, sensors/lab data, AI layer, optimization layer, and suggested software stack, and compare the likely CAPEX/licensing complexity of METSIM vs AVEVA vs AspenTech vs mining-specific platforms.

To create a digital twin of a mineral processing plant, the best software options depend heavily on whether you need a metallurgical process twin (for mass/energy balances, chemistry, and yields) or a 3D spatial/asset twin (for equipment maintenance and structural monitoring).

Because mineral processing involves complex dynamic chemistry, multi-phase fluid flows, and grinding circuits, traditional manufacturing twins do not fit.

🌟 Top 3 Process Simulation & Metallurgical Digital Twins

These platforms excel at modeling the actual physics, thermodynamics, and chemical processing of the ore.

• AVEVA (SysCAD / PRO/II / Dynamic Simulation):
  • Best For : Comprehensive pit-to-port process twins backed by Schneider Electric’s robust industrial architecture.
  • Standout Features : Exceptional at integrating heavy process data with real-time operational technology (OT) to track transient plant conditions.
• Honeywell (UniSim Design & Process Digital Twin):
  • Best For : Advanced Process Control (APC) and hybrid modeling.
  • Standout Features : Injects machine learning alongside first-principles physics models to optimize complex setups like hydrocyclone clusters and flotation cells.
• ANDRITZ (Metris Digital Twin):
  • Best For : Live plant optimization and virtual commissioning.
  • Standout Features : Acts as an "OPC Client" directly inside your DCS network, calculating soft sensors for variables that are too expensive to measure physically.

📊 Niche Metallurgical & Discrete Event Simulation

If your focus is strictly on metallurgical accounting or material flow bottlenecks:

• Metallurgical Systems (Metsim / Metris Platform Integration):
  • Best For : Plant-wide mass and energy balances.
  • Standout Features : Built specifically for mineral processors to achieve real-time ESG compliance and accurate metallurgical accounting.
• Simio or AnyLogic:
  • Best For : Discrete Event Simulation (DES).
  • Standout Features : Ideal for scheduling, tracking stockpiles, and optimizing crushers, conveyors, and downstream logistics bottlenecks.

🔎 Top Asset & 3D Spatial Twins

If you want to view a 3D model of your plant to monitor equipment wear (like SAG mill liners) or manage maintenance schedules:

• Ansys (Twin Builder):
  • Best For : High-fidelity physics validation.
  • Standout Features : Simulates deep mechanical stresses and structural breakdown on high-impact machinery like crushers.
• Siemens (Xcelerator / Tecnomatix):
  • Best For : Heavy automated infrastructure and hardware ecosystems.
  • Standout Features : Leverages the "Digital Twin Composer" on AWS alongside NVIDIA Omniverse for unparalleled industrial 3D visualization.
• Bentley Systems or Dassault Systèmes (3DEXPERIENCE):
  • Best For : Integrating surface infrastructure with subsurface geological models.

💡 Framework for Selection

To help narrow down the ideal platform architecture for your project, could you share:

• What specific minerals or metals does your processing plant handle?
• What is your primary goal (e.g., optimizing chemical recovery yields, predictive maintenance on heavy crushers, or metallurgical accounting)?
• What Control System / DCS (e.g., Honeywell, Siemens, ABB) does the physical plant currently run on?

Yes. For a mineral-processing-plant digital twin, I would not start with a generic process simulator like Aspen HYSYS or AVEVA and try to force it into a mining application. The best choice depends heavily on what you're processing and how deep you want the twin to go.

My shortlist

1. My first choice: Metso HSC + Geminex

For an actual mineral concentrator digital twin, I'd investigate Metso first.

HSC Chemistry has specific mineral-processing models covering things like comminution, classification, separation and dewatering, with particle-based simulation. Its HSC-Sim environment lets you construct flowsheets from unit-operation models and incorporate plant/test data.

More importantly, Metso has Geminex, which is specifically positioned as a metallurgical digital twin rather than merely a flowsheet simulator. It combines process models with operational data, automatically adapts model parameters to live plant data, and can run scenarios against the actual process.

That's a major distinction:

Traditional simulation:

Ore → [models] → predicted plant performance

Digital twin:

Real plant → historian/PI → calibrated models → current plant state → prediction/optimization → operator

Metso describes Geminex as using HSC process models together with real plant data and calibrated process models.

If your plant is something like:

crushing → grinding → classification → flotation → thickening → filtration

I'd put HSC/Geminex at the top of my evaluation list.

---

2. USIM PAC — probably the strongest alternative

BRGM / CASPEO's USIM PAC is particularly interesting if you want a sophisticated mineral-processing simulator without tying yourself to an equipment OEM.

It is explicitly designed for mineral-processing plants and can cover crushing, grinding, magnetic and gravity separation, flotation, leaching, concentration and refining within a common platform. It also supports digital twins, plant performance analysis, bottleneck identification, mass/water/energy balances, equipment sizing and optimization.

I'd seriously evaluate USIM PAC if you want:

• A detailed plant model
• Strong mineralogical/particle-based modeling
• Mass balancing and reconciliation
• What-if studies
• Bottleneck analysis
• Optimization
• A vendor-neutral platform

---

3. METSIM

METSIM is another very established choice in mining/metallurgy.

It supports:

• Steady-state mass and energy balances
• Equipment sizing
• Existing-plant parameterization
• Dynamic simulation
• Complex custom models
• Unlimited unit operations and streams

It also has its own programming language environment, which makes it extremely extensible.

I'd lean toward METSIM if your plant has substantial hydrometallurgy, pyrometallurgy or chemical-processing complexity, rather than a conventional concentrator.

---

4. Siemens gPROMS — best if you want a very sophisticated physics-based twin

Siemens's gPROMS is a different beast.

It's an equation-based modeling environment designed for high-fidelity models, dynamic simulation, parameter estimation, optimization and digital twins. Siemens specifically supports connecting digital process twins to real-time plant data for monitoring, soft sensing and real-time optimization.

I'd consider it if your ambition is:

"We want a highly accurate physics-based computational representation of the plant that can eventually optimize the plant in real time." But it generally requires substantially more modeling expertise than a mining-specific package.

---

5. AVEVA Process Simulation

AVEVA is worth considering if the digital twin needs to become part of a broader industrial data/operations architecture.

AVEVA Process Simulation supports high-fidelity process models, dynamic simulation, design, operator training and operations, and can integrate real-time operating data through the PI System.

The catch is that AVEVA is much more general process-industry oriented. For a flotation/concentrator model, I'd expect to do considerably more custom modeling than with HSC or USIM PAC.

---

What I'd build for your plant

The important thing is that the simulator isn't the digital twin.

I'd architect the system roughly like this:

                    ┌─────────────────────────┐
                    │     Mine / Ore Data     │
                    │ grade, mineralogy, PSD  │
                    └────────────┬────────────┘
                                 │
                                 ▼
┌──────────────┐       ┌──────────────────────┐
│ Plant PLC/DCS│──────▶│ Historian / PI System│
└──────────────┘       └──────────┬───────────┘
                                  │
                    ┌─────────────▼─────────────┐
                    │       DIGITAL TWIN        │
                    │                           │
                    │  Mass balance              │
                    │  Mineralogy                │
                    │  Unit operation models     │
                    │  Equipment models          │
                    │  State estimation          │
                    │  Model calibration         │
                    └─────────────┬─────────────┘
                                  │
                 ┌────────────────┼────────────────┐
                 ▼                ▼                ▼
           What-if studies   Optimization    Prediction
                 │                │                │
                 └────────────────┼────────────────┘
                                  ▼
                         Operator / Engineer
                              Dashboard

The most important component is the calibration/state-estimation layer. A beautiful flowsheet model that isn't continuously reconciled against actual plant behavior isn't really a useful operational twin.

The approach I'd recommend

I'd do it in five stages:

1. Build a rigorous steady-state model

• Every major unit operation
• Mass/water balance
• Assays
• Particle-size distributions
• Mineralogy
• Recovery
• Equipment operating parameters

2. Reconcile it against historical plant data

• Historian data
• Lab assays
• Flow meters
• Density
• Power
• Pressures
• Levels
• Instrument uncertainty

3. Turn it into a dynamic model

• Residence times
• Inventories
• Delays
• Equipment dynamics
• Ore variability
• Startup/shutdown behavior

4. Connect it to live plant data

• PI/historian
• OPC UA
• DCS/PLC
• Laboratory information
• MES/production systems

5. Add optimization

• Throughput
• Recovery
• Grade
• Energy
• Water
• Reagent consumption
• Equipment constraints

At that point you've moved from a process simulator → calibrated plant model → operational digital twin → optimization system.

One particularly important question

The ore/mineralogy and process type will determine my recommendation much more than the fact that it's a mining plant.

If you tell me what commodity you're processing (copper, gold, iron, lithium, phosphate, etc.), your flowsheet (e.g. SAG/ball mill + flotation), approximate throughput, and what you want the twin to actually do—for example operator training, troubleshooting, throughput optimization, predictive maintenance, recovery optimization, or autonomous control—I can give you a much more specific software/architecture recommendation, including HSC vs USIM PAC vs METSIM vs gPROMS and what I'd use for each layer of the twin.

I'd consider it if your ambition is:

"We want a highly accurate physics-based computational representation of the plant that can eventually optimize the plant in real time." But it generally requires substantially more modeling expertise than a mining-specific package.

---

5. AVEVA Process Simulation

AVEVA is worth considering if the digital twin needs to become part of a broader industrial data/operations architecture.

AVEVA Process Simulation supports high-fidelity process models, dynamic simulation, design, operator training and operations, and can integrate real-time operating data through the PI System.

The catch is that AVEVA is much more general process-industry oriented. For a flotation/concentrator model, I'd expect to do considerably more custom modeling than with HSC or USIM PAC.

---

What I'd build for your plant

The important thing is that the simulator isn't the digital twin.

I'd architect the system roughly like this:

                    ┌─────────────────────────┐
                    │     Mine / Ore Data     │
                    │ grade, mineralogy, PSD  │
                    └────────────┬────────────┘
                                 │
                                 ▼
┌──────────────┐       ┌──────────────────────┐
│ Plant PLC/DCS│──────▶│ Historian / PI System│
└──────────────┘       └──────────┬───────────┘
                                  │
                    ┌─────────────▼─────────────┐
                    │       DIGITAL TWIN        │
                    │                           │
                    │  Mass balance              │
                    │  Mineralogy                │
                    │  Unit operation models     │
                    │  Equipment models          │
                    │  State estimation          │
                    │  Model calibration         │
                    └─────────────┬─────────────┘
                                  │
                 ┌────────────────┼────────────────┐
                 ▼                ▼                ▼
           What-if studies   Optimization    Prediction
                 │                │                │
                 └────────────────┼────────────────┘
                                  ▼
                         Operator / Engineer
                              Dashboard

The most important component is the calibration/state-estimation layer. A beautiful flowsheet model that isn't continuously reconciled against actual plant behavior isn't really a useful operational twin.

The approach I'd recommend

I'd do it in five stages:

1. Build a rigorous steady-state model

• Every major unit operation
• Mass/water balance
• Assays
• Particle-size distributions
• Mineralogy
• Recovery
• Equipment operating parameters

2. Reconcile it against historical plant data

• Historian data
• Lab assays
• Flow meters
• Density
• Power
• Pressures
• Levels
• Instrument uncertainty

3. Turn it into a dynamic model

• Residence times
• Inventories
• Delays
• Equipment dynamics
• Ore variability
• Startup/shutdown behavior

4. Connect it to live plant data

• PI/historian
• OPC UA
• DCS/PLC
• Laboratory information
• MES/production systems

5. Add optimization

• Throughput
• Recovery
• Grade
• Energy
• Water
• Reagent consumption
• Equipment constraints

At that point you've moved from a process simulator → calibrated plant model → operational digital twin → optimization system.

One particularly important question

The ore/mineralogy and process type will determine my recommendation much more than the fact that it's a mining plant.

If you tell me what commodity you're processing (copper, gold, iron, lithium, phosphate, etc.), your flowsheet (e.g. SAG/ball mill + flotation), approximate throughput, and what you want the twin to actually do—for example operator training, troubleshooting, throughput optimization, predictive maintenance, recovery optimization, or autonomous control—I can give you a much more specific software/architecture recommendation, including HSC vs USIM PAC vs METSIM vs gPROMS and what I'd use for each layer of the twin.

To create a digital twin of a mineral processing plant, the best software options depend heavily on whether you need a metallurgical process twin (for mass/energy balances, chemistry, and yields) or a 3D spatial/asset twin (for equipment maintenance and structural monitoring).

Because mineral processing involves complex dynamic chemistry, multi-phase fluid flows, and grinding circuits, traditional manufacturing twins do not fit.

🌟 Top 3 Process Simulation & Metallurgical Digital Twins

These platforms excel at modeling the actual physics, thermodynamics, and chemical processing of the ore.

• AVEVA (SysCAD / PRO/II / Dynamic Simulation):
  • Best For : Comprehensive pit-to-port process twins backed by Schneider Electric’s robust industrial architecture.
  • Standout Features : Exceptional at integrating heavy process data with real-time operational technology (OT) to track transient plant conditions.
• Honeywell (UniSim Design & Process Digital Twin):
  • Best For : Advanced Process Control (APC) and hybrid modeling.
  • Standout Features : Injects machine learning alongside first-principles physics models to optimize complex setups like hydrocyclone clusters and flotation cells.
• ANDRITZ (Metris Digital Twin):
  • Best For : Live plant optimization and virtual commissioning.
  • Standout Features : Acts as an "OPC Client" directly inside your DCS network, calculating soft sensors for variables that are too expensive to measure physically.

📊 Niche Metallurgical & Discrete Event Simulation

If your focus is strictly on metallurgical accounting or material flow bottlenecks:

• Metallurgical Systems (Metsim / Metris Platform Integration):
  • Best For : Plant-wide mass and energy balances.
  • Standout Features : Built specifically for mineral processors to achieve real-time ESG compliance and accurate metallurgical accounting.
• Simio or AnyLogic:
  • Best For : Discrete Event Simulation (DES).
  • Standout Features : Ideal for scheduling, tracking stockpiles, and optimizing crushers, conveyors, and downstream logistics bottlenecks.

🔎 Top Asset & 3D Spatial Twins

If you want to view a 3D model of your plant to monitor equipment wear (like SAG mill liners) or manage maintenance schedules:

• Ansys (Twin Builder):
  • Best For : High-fidelity physics validation.
  • Standout Features : Simulates deep mechanical stresses and structural breakdown on high-impact machinery like crushers.
• Siemens (Xcelerator / Tecnomatix):
  • Best For : Heavy automated infrastructure and hardware ecosystems.
  • Standout Features : Leverages the "Digital Twin Composer" on AWS alongside NVIDIA Omniverse for unparalleled industrial 3D visualization.
• Bentley Systems or Dassault Systèmes (3DEXPERIENCE):
  • Best For : Integrating surface infrastructure with subsurface geological models.

💡 Framework for Selection

To help narrow down the ideal platform architecture for your project, could you share:

• What specific minerals or metals does your processing plant handle?
• What is your primary goal (e.g., optimizing chemical recovery yields, predictive maintenance on heavy crushers, or metallurgical accounting)?
• What Control System / DCS (e.g., Honeywell, Siemens, ABB) does the physical plant currently run on?

If your goal is a true operational digital twin of an existing mineral-processing plant, rather than simply a flowsheet simulator, I would put Metso HSC Sim + Geminex at the top of the shortlist.

The important distinction is that a digital twin is usually not one piece of software. You want a physics-based process model, connected to live plant data, continuously calibrated, with optimization/scenario capability on top.

My shortlist

1. Metso HSC Sim + Geminex — my first choice

metso.com is unusually well aligned with mineral processing. HSC supports particle-based mineral-processing models covering comminution, classification, separation and dewatering, as well as broader metallurgical calculations. Its simulation environment supports steady-state and dynamic flowsheets and model calibration against experimental data.

More importantly, Geminex is specifically designed as a metallurgical digital twin, rather than simply being a simulator. It connects process models to real plant data, automatically adapts model parameters, and can simulate operating scenarios and optimization decisions.

Metso has actually deployed this approach at Newmont's Lihir operation, where the twin uses real-time flow, density and reagent data and combines physics-based models with machine learning.

That's a very close match to what you're describing.

I'd choose this if you have a conventional concentrator/mineral-processing plant—crushing, grinding, classification, flotation, gravity, magnetic separation, thickening, filtration, etc.—and want to progress toward an online operational twin.

---

2. USIM PAC — excellent specialist alternative

If your primary requirement is accurate mineral-processing simulation, rather than the broader digital-twin platform, I'd investigate USIM PAC very seriously.

It is particularly attractive when the engineering team wants to build a detailed representation of the flowsheet and investigate:

• Mass balances
• Particle-size distributions
• Comminution
• Classification
• Flotation
• Magnetic/gravity separation
• Solid/liquid separation
• Water circuits
• Alternative flowsheet configurations
• Plant optimization

I'd put it in the same conversation as HSC for the process-modeling core, but I'd want to investigate its particular real-time data/online-twin architecture before selecting it as the overall digital-twin platform.

---

3. AspenTech — powerful if your plant has a lot of chemical processing

aspentech.com becomes more attractive if your "mineral processing plant" includes substantial hydrometallurgy, leaching, solvent extraction, precipitation, refining, acid circuits, gas systems, or other chemical processes.

Aspen is extremely mature for rigorous process simulation and online digital twins. AspenTech explicitly supports deploying Aspen Plus/HYSYS models online for real-time monitoring, what-if analysis, troubleshooting and optimization.

The caveat is that Aspen is fundamentally stronger in chemical/process engineering than in particle-based mineral-processing phenomena.

So for:

Ore → crusher → mill → cyclone → flotation → concentrate

I'd favor HSC/USIM/SysCAD.

For:

Ore → leach → SX → precipitation → refinery

I'd seriously consider Aspen.

---

4. Siemens gPROMS — excellent for a custom high-fidelity twin

siemens.com is another interesting option. Siemens explicitly positions gPROMS around high-fidelity models and digital process twins spanning design through operations.

I'd consider it if your organization has strong modeling/controls expertise and wants to build a highly customized first-principles model, rather than relying primarily on pre-built mineral-processing unit models.

It's powerful, but probably more engineering effort than you need for a conventional concentrator.

---

The architecture I'd recommend

The biggest mistake I'd avoid is thinking:

"Let's buy digital-twin software and connect it to the historian." Instead, I'd build the system in layers:

                     DIGITAL TWIN
                          │
             ┌────────────┴────────────┐
             │                         │
       Optimization              Visualization
       / What-if                   / KPIs
             │                         │
             └────────────┬────────────┘
                          │
                 PROCESS MODEL
                          │
          ┌───────────────┼───────────────┐
          │               │               │
      Comminution     Flotation       Water/
      & classification  & recovery    reagents
          │               │               │
          └───────────────┼───────────────┘
                          │
                   MODEL CALIBRATION
                          │
                 ┌────────┴────────┐
                 │                 │
             Historian          Lab/LIMS
                 │                 │
                 └────────┬────────┘
                          │
                    PLANT DATA
                          │
       PLC / DCS / SCADA / sensors / analyzers

Phase 1 — Build the steady-state model

Start with your actual flowsheet and establish:

• Ore characteristics
• Feed rates
• PSDs
• Mineralogy
• Water flows
• Reagent additions
• Equipment operating conditions
• Recoveries
• Grade
• Mass balance
• Metallurgical balance

The model needs to reproduce historical plant campaigns before you worry about fancy AI.

Phase 2 — Calibrate it against plant data

This is probably the most important stage.

I'd take several months of historical data covering:

• Normal operation
• High/low throughput
• Different ore domains
• Different grind sizes
• Different reagent regimes
• Equipment constraints
• Upsets

Then quantify how accurately the model predicts actual plant behavior.

Phase 3 — Connect the live plant

Connect:

DCS/PLC → historian → data layer → twin

Typical inputs might include:

• Belt scales
• Density meters
• Flow meters
• Particle-size analyzers
• Online XRF
• Power draw
• Pressure
• Pump speeds
• Valve positions
• Reagent flows
• pH
• ORP
• Level measurements
• Laboratory assays

Phase 4 — Make it predictive

Now the twin can answer things like:

"If we increase mill throughput by 8%, what happens to P80, flotation recovery, concentrate grade and tailings?" or:

"Given this ore blend, what grind size and reagent regime maximize recovery while keeping energy consumption below X?" That's where the twin becomes genuinely valuable.

Phase 5 — Optimization / APC

Eventually you can have:

Twin → optimizer → recommended setpoints → APC/DCS

rather than simply:

Twin → pretty dashboard

That's the point at which you're getting real operational value.

---

One particularly important decision

I'd decide whether you want a:

A. Process digital twin

"What is happening inside my processing plant, and what will happen if I change something?" or a:

B. Mine-to-metal digital twin

"Given the orebody, mine plan, ore blend and plant constraints, what is the optimum way to run the entire operation?" If it's A, I'd lean strongly toward Metso HSC/Geminex, USIM PAC, or SysCAD.

If it's B, I'd start looking at an architecture that also incorporates geology, block models, mine planning, scheduling and ore tracking. Hexagon, for example, has MinePlan capabilities covering geological/block modeling and mine planning, while Metso's Geminex can model the metallurgical portion.

My recommendation

For a real operating mineral-processing plant, my initial shortlist would be:

1. Metso HSC Sim + Geminex — strongest overall choice

2. USIM PAC — strongest alternative for mineral-process simulation

3. SysCAD — excellent engineering/process alternative

4. AspenTech — particularly compelling for hydrometallurgy/chemical processing

5. Siemens gPROMS — best when you want a highly customized high-fidelity model

And I would not start by selecting the software. I'd first define the twin's intended decisions/KPIs and architecture, then evaluate vendors against a representative section of your actual plant.

If you tell me what mineral you're processing (e.g. copper, gold, iron ore, lithium), the major unit operations, and whether you want the twin for operator decision support, optimization/APC, debottlenecking, or mine-to-mill planning, I can narrow this down to 2–3 platforms and give you a proposed digital-twin architecture and implementation roadmap.

Now the twin can answer things like:

"If we increase mill throughput by 8%, what happens to P80, flotation recovery, concentrate grade and tailings?" or:

"Given this ore blend, what grind size and reagent regime maximize recovery while keeping energy consumption below X?" That's where the twin becomes genuinely valuable.

Phase 5 — Optimization / APC

Eventually you can have:

Twin → optimizer → recommended setpoints → APC/DCS

rather than simply:

Twin → pretty dashboard

That's the point at which you're getting real operational value.

---

One particularly important decision

I'd decide whether you want a:

A. Process digital twin

"What is happening inside my processing plant, and what will happen if I change something?" or a:

B. Mine-to-metal digital twin

"Given the orebody, mine plan, ore blend and plant constraints, what is the optimum way to run the entire operation?" If it's A, I'd lean strongly toward Metso HSC/Geminex, USIM PAC, or SysCAD.

If it's B, I'd start looking at an architecture that also incorporates geology, block models, mine planning, scheduling and ore tracking. Hexagon, for example, has MinePlan capabilities covering geological/block modeling and mine planning, while Metso's Geminex can model the metallurgical portion.

My recommendation

For a real operating mineral-processing plant, my initial shortlist would be:

For a mineral processing plant digital twin, there usually isn’t one “best” software. The best choice depends on whether your goal is:

• Flowsheet design and metallurgical optimization
• Real-time plant digital twin (live data + reconciliation + optimization)
• Dynamic simulation for controls/training
• 3D asset/plant visualization
• Mine-to-mill integration

For most concentrators (crushing → grinding → flotation → thickening → filtration), I would typically build a stack rather than rely on one product.

Recommended shortlist for mineral processing digital twins

My recommendation by plant maturity

1. If you want a live operating digital twin

A strong architecture would be:

Plant data layer

• Historian (e.g., PI System)
• SCADA/DCS data
• LIMS assays
• MES/production data

↓

Metallurgical model layer

• MI Core® or USIM PAC/METSIM

↓

Analytics/optimization layer

• Machine learning models
• Constraint optimization
• Operator dashboards

↓

Visualization layer

• 3D plant model / asset visualization

This gives you:

• Real-time material balance
• Recovery prediction
• Grade/recovery optimization
• Detection of losses to tailings
• “What-if” scenarios before changing operating conditions

Digital twins in mining are most valuable when they remain connected to live plant data; a static simulation model alone is usually just a process model, not a true twin.

2. If your goal is plant debottlenecking and expansion studies

I would look first at:

1. USIM PAC
2. METSIM
3. JKSimMet (for grinding-heavy circuits)

These are excellent for questions like:

• “What happens if feed tonnage increases 15%?”
• “Should we add another cyclone?”
• “Where is recovery being lost?”
• “What is the optimum grind size?”

USIM PAC is particularly aligned with mineral processing because it models many unit operations from crushing through flotation and hydrometallurgy.

3. If you want autonomous optimization (“self-driving plant”)

You will likely need:

• A physics-based process model
• Real-time data reconciliation
• AI/ML optimization
• A control interface

A purely AI model often struggles because ore bodies and plant conditions change; a hybrid approach (first-principles model + ML) is usually more robust.

A practical implementation roadmap

Phase 1 — 3–6 months: Build the foundation

• Clean historian tags
• Validate assays
• Create a plant mass balance
• Digitize flowsheets
• Calibrate models

Phase 2 — 6–12 months: Create the twin

• Connect live operating data
• Run continuous reconciliation
• Add equipment models
• Validate predictions against production

Phase 3 — 12–24 months: Optimize

• Predict recovery
• Optimize reagent addition
• Optimize throughput vs grind
• Recommend operator actions
• Add predictive maintenance

If I were selecting for a typical copper/gold concentrator

My first evaluation would be:

1. MI Core® + existing historian → best fit for a true operational metallurgical twin
2. USIM PAC → best process simulation backbone
3. METSIM → if hydrometallurgy or complex chemistry dominates
4. AVEVA ecosystem → if the company wants a broader enterprise digital transformation platform

The key question is: Are you trying to build a digital twin for engineering studies, or a live operational twin that continuously mirrors and optimizes the plant? The software choice changes significantly depending on that.

The best choice for a mineral processing plant digital twin depends on whether you focus primarily on metallurgical mass/energy balancing, real-time control room integration, or granular particle/equipment flow dynamics.

METSIM, Honeywell Forge/UniSim , and Metallurgical Systems stand out as the top industry-leading platforms for plant-wide process simulation and digital twinning.

Top Process Simulation & Digital Twin Software

Key Selection Criteria

• For Plant-Wide Metallurgical Accounting & Balances: Use METSIM or Metallurgical Systems to build accurate mass and water balances across crushing, grinding, flotation, and thickening circuits.
• For Real-Time Operational Twins & Control: Large automation vendors like Honeywell, ABB , and Schneider Electric offer platforms that tie live SCADA/IoT data streams directly into thermodynamic or data-driven machine learning models to predict unmeasured variables.
• For Equipment-Specific Mechanical/Flow Design: Leverage Ansys Rocky if your digital twin project requires deep physics-based modeling of particle interactions in transfer chutes, mills, or screens.

To help narrow down the right platform, let me know:

• Is your plant currently operational (needing live SCADA integration) or in design phase?
• Are you focusing more on hydrometallurgical/pyrometallurgical chemistry or physical comminution/flotation circuits?

Creating a digital twin for a mineral processing plant requires bridging two different worlds: steady-state/dynamic metallurgical process simulation (crushing, grinding, flotation, leaching) and real-time industrial IoT data integration.

The "best" software depends on whether your focus is strictly on metallurgical mass/energy balancing or building a live, plant-wide operational digital twin.

Phase 1: Dedicated Mineral Processing & Metallurgical Simulators

If your primary goal is modeling ore behavior, particle size distribution (PSD), liberation, and chemical recovery, these are the industry-standard process simulation engines:

• METSIM : The industry heavyweight for mass and energy balance calculations. It handles complex hydrometallurgical, pyro-metallurgical, and conventional mineral processing circuits exceptionally well. It is widely used for flowsheet design, process optimization, and evaluating varying ore types.
• USIM PAC : Specifically designed for the mineral and metallurgical industries. It covers everything from comminution (crushing/grinding) to refining, featuring robust libraries for particle size, mineral composition, and liberation classes. Excellent for data reconciliation and plant auditing.
• JKSimMet / JKSimFloat : The gold standard explicitly for comminution (crushing, SAG/ball milling) and flotation circuit modeling and optimization. If your digital twin's heart beats loudest in the grinding and flotation bays, these tools provide the underlying breakage and liberation physics.
• HSC Chemistry : Outstanding thermodynamic calculation and chemical reaction simulation tools, especially useful if your plant involves complex leaching, pressure oxidation, or solvent extraction-electrowinning (SX-EW).

Phase 2: Enterprise-Grade Real-Time Digital Twin Platforms

If you want to move beyond offline engineering simulations into a live, continuously synchronized digital twin tied to your plant’s SCADA/DCS data, look at enterprise industrial platforms:

• AVEVA Process Simulation + PI System : Backed by Schneider Electric, AVEVA excels at live operational twins. The AVEVA PI System acts as the high-speed data backbone capturing millions of sensor streams (tonnage, water pull, reagent feed), while AVEVA Process Simulation builds the underlying thermodynamic and mass-balance models.
• Siemens Xcelerator : Siemens provides robust automation, electrical, and process digital twin capabilities, merging PLC/SCADA data with physics-based simulation models for equipment health, energy tracking, and throughput optimization.
• ANDRITZ IDEAS : Frequently used for dynamic simulation, control system checkout (virtual commissioning), and advanced operator training simulators (OTS) that can run in parallel with the live plant.

Which path should you choose?

1. The Hybrid Approach (Most Common): Build your foundational physics and metallurgical models using specialized tools like METSIM or USIM PAC to understand steady-state bottlenecks, reagent consumption, and grade recoveries. Then, export or re-platform those core logic formulas into an enterprise IoT data historian layer like AVEVA or Siemens to feed live sensor data into the model for real-time operator guidance.
2. If focusing tightly on Circuit Optimization: Start with JKSimMet/JKSimFloat or USIM PAC to dial in the grinding and separation physics before scaling up to a plant-wide twin.

To help narrow down the ideal software stack, could you share:

• Is your plant greenfield (yet to be built) or brownfield (existing with live sensor data)?
• What are your primary units—is it heavily comminution/flotation focused, or does it include complex hydrometallurgy/leaching?

If your goal is a true digital twin of an operating mineral-processing plant, rather than simply a flowsheet for design studies, I would not choose software based on “best process simulator” alone. You really need to separate the problem into (1) mineral-processing process modeling, (2) dynamic/operational simulation, and (3) the data/industrial-digital-twin layer.

For most mining plants, my shortlist would be:

My #1 choice: METSIM

For a conventional mineral-processing operation—crushing → grinding → classification → flotation → thickening/filtration → concentrate/tailings, or a hydromet plant—I would seriously evaluate METSIM first.

METSIM is unusually well aligned with what you're describing because it isn't just a steady-state flowsheet simulator. It supports:

• steady-state mass and energy balances
• parameterization against existing plant data
• dynamic simulation
• PID and other process-control logic
• equipment sizing
• comminution and classification
• flotation
• solid/liquid separation
• leaching, SX, EW, CIL/CIP, etc.
• mine/block-model inputs
• stockpile and material tracking
• operating-cost calculations

Its current documentation specifically describes parameterizing models from existing plant operation and dynamic simulation, which are two of the things I'd consider essential for turning a process model into a plant twin.

It also has a Mine module capable of connecting drill-hole/block-model/mining-sequence information to stockpiles and the plant, allowing simulation over the life of the mine.

USIM PAC may actually be better for an ore-to-product twin

I'd put USIM PAC extremely high on your evaluation list, particularly if your main objective is plant reconciliation, optimization, ore variability and performance prediction.

USIM PAC was developed specifically around mineral processing and can represent the material characteristics that matter enormously in mining—particle size, mineral composition, physical properties, etc.—rather than treating the feed simply as conventional chemical-process streams. BRGM describes it as being usable for digital twins, plant monitoring, bottleneck identification, material/water/energy balances, equipment sizing and optimization.

That makes it particularly interesting if your twin needs to answer questions such as:

“Given today's ore mineralogy and PSD, what will the plant produce six hours from now?” rather than simply:

“What is the mass balance around this flowsheet?”

HSC Chemistry is worth considering for difficult chemistry/mineralogy

HSC-Sim is particularly strong when your process has significant thermodynamics, mineralogy, hydrometallurgy or complex chemical reactions.

Metso describes HSC-Sim as a flexible process simulator used for mineralogical and hydrometallurgical mass-balance simulations, with the ability to represent thousands of particle/variable types.

I'd be especially interested in it for:

• hydrometallurgy
• leaching
• precipitation
• roasting
• aqueous chemistry
• mineralogical modeling
• complex equilibrium problems

Where Siemens/AVEVA enter the picture

This is the important distinction: METSIM/USIM PAC are closer to the mineral-processing brain of the twin. Siemens/AVEVA are closer to the industrial digital-twin nervous system.

For example, Siemens gPROMS is designed for high-fidelity process models, optimization and digital process twins across the concept-to-operations lifecycle.

Similarly, AVEVA Process Simulation is positioned around design, simulation, training and operations, with the high-fidelity process model serving as the foundation for a digital twin.

So I wouldn't necessarily make them compete directly with METSIM. A sophisticated architecture could use both.

What I'd build

If this were my project, I'd architect the twin roughly like this:

Plant instrumentation / historian

↓

Data layer SCADA + DCS + historian + laboratory/LIMS + MES + maintenance + mine/ore data

↓

Mineral-processing digital twin

USIM PAC or METSIM → material characterization → mass balance → equipment models → recovery → throughput → water balance → reagent consumption → product quality → tailings

↓

Dynamic model

→ residence times → surge capacity → control loops → disturbances → startup/shutdown → ore-feed changes

↓

Optimization / ML layer

→ throughput optimization → recovery optimization → reagent optimization → energy optimization → predictive control → what-if scenarios

↓

Operator/engineering interface

→ live plant state → predicted state → alarms → recommendations → scenario simulator

That architecture is much more valuable than simply putting a 3D model of the plant on a screen and calling it a digital twin.

The critical decision

Before selecting the software, I'd define what you want the twin to predict.

For example:

1. Mass-balance twin — “Where is all the material going?”
2. Performance twin — “What recovery/grade/throughput should I expect?”
3. Dynamic twin — “What happens over the next 5–60 minutes when feed or operating conditions change?”
4. Optimization twin — “What operating setpoints maximize tonnes/recovery/profit?”
5. Predictive twin — “What is likely to happen before the plant actually does it?”
6. Operator training twin — “Can operators practice abnormal situations?”
7. Asset twin — “Will this pump/mill/thickener fail or degrade?”
8. Mine-to-mill twin — “How does ore from today's mining sequence propagate through the entire plant?”

For a serious mining implementation, I'd aim for #2–#5, with #8 if you have good geological and mine-plan data.

My recommendation

If you gave me no additional information and told me to start evaluating vendors tomorrow, I'd run a proof-of-concept bake-off between METSIM and USIM PAC, and then evaluate Siemens/AVEVA as the broader enterprise/dynamic-twin layer rather than assuming one product has to do everything.

The POC should use your actual plant data, not a vendor demonstration. Give each vendor the same 3–6 months of historian + lab data and ask them to reproduce:

• feed tonnes and PSD
• mineral/element assays
• mill power and throughput
• cyclone performance
• flotation recovery and grade
• concentrate production
• tailings losses
• water balance
• reagent consumption
• major equipment constraints
• plant disturbances

Then test whether the model can predict a period of plant operation that wasn't used for calibration. That is the real test of a digital twin.

If you tell me what mineral you're processing (copper, gold, lithium, iron ore, etc.), the flowsheet, major equipment, DCS/historian you use (OSIsoft PI, Ignition, AVEVA, Siemens, Rockwell, etc.), and whether your goal is optimization vs operator training vs predictive control, I can narrow this down to a specific recommended software stack and architecture for your plant.

So I wouldn't necessarily make them compete directly with METSIM. A sophisticated architecture could use both.

What I'd build

If this were my project, I'd architect the twin roughly like this:

Plant instrumentation / historian

↓

Data layer SCADA + DCS + historian + laboratory/LIMS + MES + maintenance + mine/ore data

↓

Mineral-processing digital twin

USIM PAC or METSIM → material characterization → mass balance → equipment models → recovery → throughput → water balance → reagent consumption → product quality → tailings

↓

Dynamic model

→ residence times → surge capacity → control loops → disturbances → startup/shutdown → ore-feed changes

↓

Optimization / ML layer

→ throughput optimization → recovery optimization → reagent optimization → energy optimization → predictive control → what-if scenarios

↓

Operator/engineering interface

→ live plant state → predicted state → alarms → recommendations → scenario simulator

That architecture is much more valuable than simply putting a 3D model of the plant on a screen and calling it a digital twin.

The critical decision

Before selecting the software, I'd define what you want the twin to predict.

For example:

1. Mass-balance twin — “Where is all the material going?”
2. Performance twin — “What recovery/grade/throughput should I expect?”
3. Dynamic twin — “What happens over the next 5–60 minutes when feed or operating conditions change?”
4. Optimization twin — “What operating setpoints maximize tonnes/recovery/profit?”
5. Predictive twin — “What is likely to happen before the plant actually does it?”
6. Operator training twin — “Can operators practice abnormal situations?”
7. Asset twin — “Will this pump/mill/thickener fail or degrade?”
8. Mine-to-mill twin — “How does ore from today's mining sequence propagate through the entire plant?”

For a serious mining implementation, I'd aim for #2–#5, with #8 if you have good geological and mine-plan data.

My recommendation

If you gave me no additional information and told me to start evaluating vendors tomorrow, I'd run a proof-of-concept bake-off between METSIM and USIM PAC, and then evaluate Siemens/AVEVA as the broader enterprise/dynamic-twin layer rather than assuming one product has to do everything.

The POC should use your actual plant data, not a vendor demonstration. Give each vendor the same 3–6 months of historian + lab data and ask them to reproduce:

• feed tonnes and PSD
• mineral/element assays
• mill power and throughput
• cyclone performance
• flotation recovery and grade
• concentrate production
• tailings losses
• water balance
• reagent consumption
• major equipment constraints
• plant disturbances

Then test whether the model can predict a period of plant operation that wasn't used for calibration. That is the real test of a digital twin.

If you tell me what mineral you're processing (copper, gold, lithium, iron ore, etc.), the flowsheet, major equipment, DCS/historian you use (OSIsoft PI, Ignition, AVEVA, Siemens, Rockwell, etc.), and whether your goal is optimization vs operator training vs predictive control, I can narrow this down to a specific recommended software stack and architecture for your plant.